subsurface/parse-xml.c

2191 lines
54 KiB
C
Raw Normal View History

#include <stdio.h>
#include <ctype.h>
#include <string.h>
#include <stdlib.h>
#include <errno.h>
#include <unistd.h>
#include <assert.h>
#define __USE_XOPEN
#include <time.h>
#include <libxml/parser.h>
#include <libxml/parserInternals.h>
#include <libxml/tree.h>
#include <libxslt/transform.h>
#include <libdivecomputer/parser.h>
#include "gettext.h"
#include "dive.h"
Assemble the actual Suunto serial number It turns out that the serial number returned by libdivecomputer isn't really the serial number as interpreted by the vendor. Those tend to be strings, but libdivecomputer gives us a 32bit number. Some experimenting showed that for the Suunto devies tested the serial number is encoded in that 32bit number: It so happens that the Suunto serial number strings are strings that have all numbers, but they aren't *one* number. They are four bytes representing two numbers each, and the "23500027" string is actually the four bytes 23 50 00 27 (0x17 0x32 0x00 0x1b). And libdivecomputer has incorrectly parsed those four bytes as one number, not as the encoded serial number string it is. So the value 389152795 is actually hex 0x1732001b, which is 0x17 0x32 0x00 0x1b, which is - 23 50 00 27. This should be done by libdivecomputer, but hey, in the meantime this at least shows the concept. And helps test the XML save/restore code. It depends on the two patches that create the whole "device.c" infrastructure, of course. With this, my dive file ends up having the settings section look like this: <divecomputerid model='Suunto Vyper Air' deviceid='d4629110' serial='01201094' firmware='1.1.22'/> <divecomputerid model='Suunto HelO2' deviceid='995dd566' serial='23500027' firmware='1.0.4'/> where the format of the firmware version is something I guessed at, but it was the obvious choice (again, it's byte-based, I'm ignoring the high byte that is zero for both of my Suuntos). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2013-01-10 00:14:21 +00:00
#include "device.h"
int verbose, quit;
int metric = 1;
static xmlDoc *test_xslt_transforms(xmlDoc *doc, const char **params);
/* the dive table holds the overall dive list; target table points at
* the table we are currently filling */
struct dive_table dive_table;
struct dive_table *target_table = NULL;
/*
* Add a dive into the dive_table array
*/
static void record_dive_to_table(struct dive *dive, struct dive_table *table)
{
assert(table != NULL);
int nr = table->nr, allocated = table->allocated;
struct dive **dives = table->dives;
if (nr >= allocated) {
allocated = (nr + 32) * 3 / 2;
dives = realloc(dives, allocated * sizeof(struct dive *));
if (!dives)
exit(1);
table->dives = dives;
table->allocated = allocated;
}
dives[nr] = fixup_dive(dive);
table->nr = nr + 1;
}
void record_dive(struct dive *dive)
{
record_dive_to_table(dive, &dive_table);
}
static void start_match(const char *type, const char *name, char *buffer)
{
if (verbose > 2)
printf("Matching %s '%s' (%s)\n",
type, name, buffer);
}
static void nonmatch(const char *type, const char *name, char *buffer)
{
if (verbose > 1)
printf("Unable to match %s '%s' (%s)\n",
type, name, buffer);
}
typedef void (*matchfn_t)(char *buffer, void *);
static int match(const char *pattern, int plen,
const char *name,
matchfn_t fn, char *buf, void *data)
{
switch (name[plen]) {
case '\0':
case '.':
break;
default:
return 0;
}
if (memcmp(pattern, name, plen))
return 0;
fn(buf, data);
return 1;
}
struct units xml_parsing_units;
const struct units SI_units = SI_UNITS;
const struct units IMPERIAL_units = IMPERIAL_UNITS;
/*
* Dive info as it is being built up..
*/
static struct divecomputer *cur_dc;
static struct dive *cur_dive;
static dive_trip_t *cur_trip = NULL;
static struct sample *cur_sample;
static struct picture *cur_picture;
static struct {
int active;
duration_t time;
int type, flags, value;
const char *name;
} cur_event;
static struct {
struct {
const char *model;
uint32_t deviceid;
const char *nickname, *serial_nr, *firmware;
} dc;
} cur_settings;
static bool in_settings = false;
static bool in_userid = false;
static struct tm cur_tm;
static int cur_cylinder_index, cur_ws_index;
static int lastndl, laststoptime, laststopdepth, lastcns, lastpo2, lastindeco;
First step in cleaning up cylinder pressure sensor logic This clarifies/changes the meaning of our "cylinderindex" entry in our samples. It has been rather confused, because different dive computers have done things differently, and the naming really hasn't helped. There are two totally different - and independent - cylinder "indexes": - the pressure sensor index, which indicates which cylinder the sensor data is from. - the "active cylinder" index, which indicates which cylinder we actually breathe from. These two values really are totally independent, and have nothing what-so-ever to do with each other. The sensor index may well be fixed: many dive computers only support a single pressure sensor (whether wireless or wired), and the sensor index is thus always zero. Other dive computers may support multiple pressure sensors, and the gas switch event may - or may not - indicate that the sensor changed too. A dive computer might give the sensor data for *all* cylinders it can read, regardless of which one is the one we're actively breathing. In fact, some dive computers might give sensor data for not just *your* cylinder, but your buddies. This patch renames "cylinderindex" in the samples as "sensor", making it quite clear that it's about which sensor index the pressure data in the sample is about. The way we figure out which is the currently active gas is with an explicit has change event. If a computer (like the Uemis Zurich) joins the two concepts together, then a sensor change should also create a gas switch event. This patch also changes the Uemis importer to do that. Finally, it should be noted that the plot info works totally separately from the sample data, and is about what we actually *display*, not about the sample pressures etc. In the plot info, the "cylinderindex" does in fact mean the currently active cylinder, and while it is initially set to match the sensor information from the samples, we then walk the gas change events and fix it up - and if the active cylinder differs from the sensor cylinder, we clear the sensor data. [Dirk Hohndel: this conflicted with some of my recent changes - I think I merged things correctly...] Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2012-12-31 04:00:51 +00:00
static int lastcylinderindex, lastsensor;
/*
* If we don't have an explicit dive computer,
* we use the implicit one that every dive has..
*/
static struct divecomputer *get_dc(void)
{
return cur_dc ?: &cur_dive->dc;
}
static enum import_source {
UNKNOWN,
LIBDIVECOMPUTER,
DIVINGLOG,
UDDF,
} import_source;
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
static void divedate(char *buffer, timestamp_t *when)
{
int d, m, y;
int hh, mm, ss;
hh = 0;
mm = 0;
ss = 0;
if (sscanf(buffer, "%d.%d.%d %d:%d:%d", &d, &m, &y, &hh, &mm, &ss) >= 3) {
/* This is ok, and we got at least the date */
} else if (sscanf(buffer, "%d-%d-%d %d:%d:%d", &y, &m, &d, &hh, &mm, &ss) >= 3) {
/* This is also ok */
} else {
fprintf(stderr, "Unable to parse date '%s'\n", buffer);
return;
}
cur_tm.tm_year = y;
cur_tm.tm_mon = m - 1;
cur_tm.tm_mday = d;
cur_tm.tm_hour = hh;
cur_tm.tm_min = mm;
cur_tm.tm_sec = ss;
*when = utc_mktime(&cur_tm);
}
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
static void divetime(char *buffer, timestamp_t *when)
{
int h, m, s = 0;
if (sscanf(buffer, "%d:%d:%d", &h, &m, &s) >= 2) {
cur_tm.tm_hour = h;
cur_tm.tm_min = m;
cur_tm.tm_sec = s;
*when = utc_mktime(&cur_tm);
}
}
/* Libdivecomputer: "2011-03-20 10:22:38" */
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
static void divedatetime(char *buffer, timestamp_t *when)
{
int y, m, d;
int hr, min, sec;
if (sscanf(buffer, "%d-%d-%d %d:%d:%d",
&y, &m, &d, &hr, &min, &sec) == 6) {
cur_tm.tm_year = y;
cur_tm.tm_mon = m - 1;
cur_tm.tm_mday = d;
cur_tm.tm_hour = hr;
cur_tm.tm_min = min;
cur_tm.tm_sec = sec;
*when = utc_mktime(&cur_tm);
}
}
enum ParseState {
FINDSTART,
FINDEND
};
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
static void divetags(char *buffer, struct tag_entry **tags)
{
int i = 0, start = 0, end = 0;
enum ParseState state = FINDEND;
int len = buffer ? strlen(buffer) : 0;
while (i < len) {
if (buffer[i] == ',') {
if (state == FINDSTART) {
/* Detect empty tags */
} else if (state == FINDEND) {
/* Found end of tag */
if (i > 0 && buffer[i - 1] != '\\') {
buffer[i] = '\0';
state = FINDSTART;
taglist_add_tag(tags, buffer + start);
} else {
state = FINDSTART;
}
}
} else if (buffer[i] == ' ') {
/* Handled */
} else {
/* Found start of tag */
if (state == FINDSTART) {
state = FINDEND;
start = i;
} else if (state == FINDEND) {
end = i;
}
}
i++;
}
if (state == FINDEND) {
if (end < start)
end = len - 1;
if (len > 0) {
buffer[end + 1] = '\0';
taglist_add_tag(tags, buffer + start);
}
}
}
enum number_type {
NEITHER,
FLOAT
};
static enum number_type parse_float(const char *buffer, double *res, const char **endp)
{
double val;
static bool first_time = true;
errno = 0;
val = ascii_strtod(buffer, endp);
if (errno || *endp == buffer)
return NEITHER;
if (**endp == ',') {
if (IS_FP_SAME(val, rint(val))) {
/* we really want to send an error if this is a Subsurface native file
* as this is likely indication of a bug - but right now we don't have
* that information available */
if (first_time) {
fprintf(stderr, "Floating point value with decimal comma (%s)?\n", buffer);
first_time = false;
}
/* Try again in permissive mode*/
val = strtod_flags(buffer, endp, 0);
}
}
*res = val;
return FLOAT;
}
union int_or_float {
double fp;
};
static enum number_type integer_or_float(char *buffer, union int_or_float *res)
{
const char *end;
return parse_float(buffer, &res->fp, &end);
}
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
static void pressure(char *buffer, pressure_t *pressure)
{
double mbar = 0.0;
union int_or_float val;
switch (integer_or_float(buffer, &val)) {
case FLOAT:
/* Just ignore zero values */
if (!val.fp)
break;
switch (xml_parsing_units.pressure) {
case PASCAL:
mbar = val.fp / 100;
break;
case BAR:
/* Assume mbar, but if it's really small, it's bar */
mbar = val.fp;
if (fabs(mbar) < 5000)
mbar = mbar * 1000;
break;
case PSI:
mbar = psi_to_mbar(val.fp);
break;
}
if (fabs(mbar) > 5 && fabs(mbar) < 5000000) {
pressure->mbar = rint(mbar);
break;
}
/* fallthrough */
default:
printf("Strange pressure reading %s\n", buffer);
}
}
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
static void salinity(char *buffer, int *salinity)
{
union int_or_float val;
switch (integer_or_float(buffer, &val)) {
case FLOAT:
*salinity = rint(val.fp * 10.0);
break;
default:
printf("Strange salinity reading %s\n", buffer);
}
}
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
static void depth(char *buffer, depth_t *depth)
{
union int_or_float val;
switch (integer_or_float(buffer, &val)) {
case FLOAT:
switch (xml_parsing_units.length) {
case METERS:
depth->mm = rint(val.fp * 1000);
break;
case FEET:
depth->mm = feet_to_mm(val.fp);
break;
}
break;
default:
printf("Strange depth reading %s\n", buffer);
}
}
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
static void weight(char *buffer, weight_t *weight)
{
union int_or_float val;
switch (integer_or_float(buffer, &val)) {
case FLOAT:
switch (xml_parsing_units.weight) {
case KG:
weight->grams = rint(val.fp * 1000);
break;
case LBS:
weight->grams = lbs_to_grams(val.fp);
break;
}
break;
default:
printf("Strange weight reading %s\n", buffer);
}
}
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
static void temperature(char *buffer, temperature_t *temperature)
{
union int_or_float val;
switch (integer_or_float(buffer, &val)) {
case FLOAT:
switch (xml_parsing_units.temperature) {
case KELVIN:
temperature->mkelvin = val.fp * 1000;
break;
case CELSIUS:
temperature->mkelvin = C_to_mkelvin(val.fp);
break;
case FAHRENHEIT:
temperature->mkelvin = F_to_mkelvin(val.fp);
break;
}
break;
default:
printf("Strange temperature reading %s\n", buffer);
}
/* temperatures outside -40C .. +70C should be ignored */
if (temperature->mkelvin < ZERO_C_IN_MKELVIN - 40000 ||
temperature->mkelvin > ZERO_C_IN_MKELVIN + 70000)
temperature->mkelvin = 0;
}
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
static void sampletime(char *buffer, duration_t *time)
{
int i;
int min, sec;
i = sscanf(buffer, "%d:%d", &min, &sec);
switch (i) {
case 1:
sec = min;
min = 0;
/* fallthrough */
case 2:
time->seconds = sec + min * 60;
break;
default:
printf("Strange sample time reading %s\n", buffer);
}
}
static void offsettime(char *buffer, offset_t *time)
{
duration_t uoffset;
int sign = 1;
if (*buffer == '-') {
sign = -1;
buffer++;
}
/* yes, this could indeed fail if we have an offset > 34yrs
* - too bad */
sampletime(buffer, &uoffset);
time->seconds = sign * uoffset.seconds;
}
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
static void duration(char *buffer, duration_t *time)
{
/* DivingLog 5.08 (and maybe other versions) appear to sometimes
* store the dive time as 44.00 instead of 44:00;
* This attempts to parse this in a fairly robust way */
if (!strchr(buffer, ':') && strchr(buffer, '.')) {
char *mybuffer = strdup(buffer);
char *dot = strchr(mybuffer, '.');
*dot = ':';
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
sampletime(mybuffer, time);
free(mybuffer);
} else {
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
sampletime(buffer, time);
}
}
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
static void percent(char *buffer, fraction_t *fraction)
{
double val;
const char *end;
switch (parse_float(buffer, &val, &end)) {
case FLOAT:
/* Turn fractions into percent unless explicit.. */
if (val <= 1.0) {
while (isspace(*end))
end++;
if (*end != '%')
val *= 100;
}
/* Then turn percent into our integer permille format */
if (val >= 0 && val <= 100.0) {
fraction->permille = rint(val * 10);
break;
}
default:
printf(translate("gettextFromC", "Strange percentage reading %s\n"), buffer);
break;
}
}
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
static void gasmix(char *buffer, fraction_t *fraction)
{
/* libdivecomputer does negative percentages. */
if (*buffer == '-')
return;
if (cur_cylinder_index < MAX_CYLINDERS)
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
percent(buffer, fraction);
}
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
static void gasmix_nitrogen(char *buffer, struct gasmix *gasmix)
{
/* Ignore n2 percentages. There's no value in them. */
}
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
static void cylindersize(char *buffer, volume_t *volume)
{
union int_or_float val;
switch (integer_or_float(buffer, &val)) {
case FLOAT:
volume->mliter = rint(val.fp * 1000);
break;
default:
printf("Strange volume reading %s\n", buffer);
break;
}
}
/* Trim a character string by removing leading and trailing white space characters.
* Parameter: a pointer to a null-terminated character string (buffer);
* Return value: length of the trimmed string, excluding the terminal 0x0 byte
* The original pointer (buffer) remains valid after this function has been called
* and points to the trimmed string */
int trimspace(char *buffer) {
int i, size, start, end;
size = strlen(buffer);
for(start = 0; isspace(buffer[start]); start++)
if (start >= size) return 0; // Find 1st character following leading whitespace
for(end = size - 1; isspace(buffer[end]); end--) // Find last character before trailing whitespace
if (end <= 0) return 0;
for(i = start; i <= end; i++) // Move the nonspace characters to the start of the string
buffer[i-start] = buffer[i];
size = end - start + 1;
buffer[size] = 0x0; // then terminate the string
return size; // return string length
}
static void utf8_string(char *buffer, void *_res)
{
char **res = _res;
int size;
size = trimspace(buffer);
if(size)
*res = strdup(buffer);
}
/* Extract the dive computer type from the xml text buffer */
static void get_dc_type(char *buffer, enum dive_comp_type *i)
{
if((trimspace(buffer)) && (strcmp(buffer,"CCR") == 0))
*i = CCR; // if the xml string = "CCR", set dc-type to CCR
} // otherwise the default dc-type is used (OC)
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
#define MATCH(pattern, fn, dest) ({ \
/* Silly type compatibility test */ \
if (0) (fn)("test", dest); \
match(pattern, strlen(pattern), name, (matchfn_t) (fn), buf, dest); })
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
static void get_index(char *buffer, int *i)
{
*i = atoi(buffer);
}
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
static void get_uint8(char *buffer, uint8_t *i)
{
*i = atoi(buffer);
}
static void get_bearing(char *buffer, bearing_t *bearing)
{
bearing->degrees = atoi(buffer);
}
static void get_rating(char *buffer, int *i)
{
int j = atoi(buffer);
if (j >= 0 && j <= 5) {
*i = j;
}
}
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
static void double_to_o2pressure(char *buffer, o2pressure_t *i)
{
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
i->mbar = rint(ascii_strtod(buffer, NULL) * 1000.0);
}
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
static void hex_value(char *buffer, uint32_t *i)
{
*i = strtoul(buffer, NULL, 16);
}
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
static void get_tripflag(char *buffer, tripflag_t *tf)
First cut of explicit trip tracking This code establishes the explicit trip data structures and loads and saves them in the XML data. No attempts are made to edit / modify the trips, yet. Loading XML files without trip data creates the trips based on timing as before. Saving out the same, unmodified data will create 'trip' entries in the XML file with a 'number' that reflects the number of dives in that trip. The trip tag also stores the beginning time of the first dive in the trip and the location of the trip (which we display in the summary entries in the UI). The logic allows for dives that aren't part of a dive trip. All other dives simply belong to the "previous" dive trip - i.e. the dive trip with the latest start time that is earlier or equal to the start time of this dive. This logic significantly simplifies the tracking of trips compared to other approaches that I have tried. The automatic grouping into trips now is an option that defaults to off (as it makes changes to the XML file - and people who don't want this feature shouldn't have trips added to their XML files that they then need to manually remove). For now you have to select this option, then exit the program and start it again. Still to do is to trigger the trip generation at run time. We also need a way to mark dives as not part of trips and to allow options to combine trips, split trips, edit trip location data, etc. The code has only had some limited testing when opening multiple files. The code is known to fail if a location name contains unquoted special characters like an "'". This commit also fixes a visual inconsistency in the preferences dialog where the font selector button didn't have a frame around it that told you what this option was about. Inspired-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2012-08-22 05:04:24 +00:00
{
*tf = strcmp(buffer, "NOTRIP") ? TF_NONE : NO_TRIP;
First cut of explicit trip tracking This code establishes the explicit trip data structures and loads and saves them in the XML data. No attempts are made to edit / modify the trips, yet. Loading XML files without trip data creates the trips based on timing as before. Saving out the same, unmodified data will create 'trip' entries in the XML file with a 'number' that reflects the number of dives in that trip. The trip tag also stores the beginning time of the first dive in the trip and the location of the trip (which we display in the summary entries in the UI). The logic allows for dives that aren't part of a dive trip. All other dives simply belong to the "previous" dive trip - i.e. the dive trip with the latest start time that is earlier or equal to the start time of this dive. This logic significantly simplifies the tracking of trips compared to other approaches that I have tried. The automatic grouping into trips now is an option that defaults to off (as it makes changes to the XML file - and people who don't want this feature shouldn't have trips added to their XML files that they then need to manually remove). For now you have to select this option, then exit the program and start it again. Still to do is to trigger the trip generation at run time. We also need a way to mark dives as not part of trips and to allow options to combine trips, split trips, edit trip location data, etc. The code has only had some limited testing when opening multiple files. The code is known to fail if a location name contains unquoted special characters like an "'". This commit also fixes a visual inconsistency in the preferences dialog where the font selector button didn't have a frame around it that told you what this option was about. Inspired-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2012-08-22 05:04:24 +00:00
}
/*
* Divinglog is crazy. The temperatures are in celsius. EXCEPT
* for the sample temperatures, that are in Fahrenheit.
* WTF?
*
* Oh, and I think Diving Log *internally* probably kept them
* in celsius, because I'm seeing entries like
*
* <Temp>32.0</Temp>
*
* in there. Which is freezing, aka 0 degC. I bet the "0" is
* what Diving Log uses for "no temperature".
*
* So throw away crap like that.
*
* It gets worse. Sometimes the sample temperatures are in
* Celsius, which apparently happens if you are in a SI
* locale. So we now do:
*
* - temperatures < 32.0 == Celsius
* - temperature == 32.0 -> garbage, it's a missing temperature (zero converted from C to F)
* - temperatures > 32.0 == Fahrenheit
*/
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
static void fahrenheit(char *buffer, temperature_t *temperature)
{
union int_or_float val;
switch (integer_or_float(buffer, &val)) {
case FLOAT:
if (IS_FP_SAME(val.fp, 32.0))
break;
if (val.fp < 32.0)
temperature->mkelvin = C_to_mkelvin(val.fp);
else
temperature->mkelvin = F_to_mkelvin(val.fp);
break;
default:
fprintf(stderr, "Crazy Diving Log temperature reading %s\n", buffer);
}
}
/*
* Did I mention how bat-shit crazy divinglog is? The sample
* pressures are in PSI. But the tank working pressure is in
* bar. WTF^2?
*
* Crazy stuff like this is why subsurface has everything in
* these inconvenient typed structures, and you have to say
* "pressure->mbar" to get the actual value. Exactly so that
* you can never have unit confusion.
*
* It gets worse: sometimes apparently the pressures are in
* bar, sometimes in psi. Dirk suspects that this may be a
* DivingLog Uemis importer bug, and that they are always
* supposed to be in bar, but that the importer got the
* sample importing wrong.
*
* Sadly, there's no way to really tell. So I think we just
* have to have some arbitrary cut-off point where we assume
* that smaller values mean bar.. Not good.
*/
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
static void psi_or_bar(char *buffer, pressure_t *pressure)
{
union int_or_float val;
switch (integer_or_float(buffer, &val)) {
case FLOAT:
if (val.fp > 400)
pressure->mbar = psi_to_mbar(val.fp);
else
pressure->mbar = rint(val.fp * 1000);
break;
default:
fprintf(stderr, "Crazy Diving Log PSI reading %s\n", buffer);
}
}
static int divinglog_fill_sample(struct sample *sample, const char *name, char *buf)
{
return MATCH("time.p", sampletime, &sample->time) ||
MATCH("depth.p", depth, &sample->depth) ||
MATCH("temp.p", fahrenheit, &sample->temperature) ||
MATCH("press1.p", psi_or_bar, &sample->cylinderpressure) ||
0;
}
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
static void uddf_gasswitch(char *buffer, struct sample *sample)
{
int idx = atoi(buffer);
int seconds = sample->time.seconds;
struct dive *dive = cur_dive;
struct divecomputer *dc = get_dc();
add_gas_switch_event(dive, dc, seconds, idx);
}
static int uddf_fill_sample(struct sample *sample, const char *name, char *buf)
{
return MATCH("divetime", sampletime, &sample->time) ||
MATCH("depth", depth, &sample->depth) ||
MATCH("temperature", temperature, &sample->temperature) ||
MATCH("tankpressure", pressure, &sample->cylinderpressure) ||
MATCH("ref.switchmix", uddf_gasswitch, sample) ||
0;
}
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
static void eventtime(char *buffer, duration_t *duration)
{
sampletime(buffer, duration);
if (cur_sample)
duration->seconds += cur_sample->time.seconds;
}
static void try_to_match_autogroup(const char *name, char *buf)
{
int autogroupvalue;
start_match("autogroup", name, buf);
if (MATCH("state.autogroup", get_index, &autogroupvalue)) {
set_autogroup(autogroupvalue);
return;
}
nonmatch("autogroup", name, buf);
}
void add_gas_switch_event(struct dive *dive, struct divecomputer *dc, int seconds, int idx)
{
/* The gas switch event format is insane. It will be fixed, I think */
int o2 = get_o2(&dive->cylinder[idx].gasmix);
int he = get_he(&dive->cylinder[idx].gasmix);
int value;
o2 = (o2 + 5) / 10;
he = (he + 5) / 10;
value = o2 + (he << 16);
add_event(dc, seconds, he ? SAMPLE_EVENT_GASCHANGE2 : SAMPLE_EVENT_GASCHANGE, 0, value, "gaschange");
}
static void get_cylinderindex(char *buffer, uint8_t *i)
{
*i = atoi(buffer);
if (lastcylinderindex != *i) {
add_gas_switch_event(cur_dive, get_dc(), cur_sample->time.seconds, *i);
lastcylinderindex = *i;
}
}
static void get_sensor(char *buffer, uint8_t *i)
{
*i = atoi(buffer);
lastsensor = *i;
}
static void try_to_fill_dc_settings(const char *name, char *buf)
{
start_match("divecomputerid", name, buf);
if (MATCH("model.divecomputerid", utf8_string, &cur_settings.dc.model))
return;
if (MATCH("deviceid.divecomputerid", hex_value, &cur_settings.dc.deviceid))
return;
if (MATCH("nickname.divecomputerid", utf8_string, &cur_settings.dc.nickname))
return;
if (MATCH("serial.divecomputerid", utf8_string, &cur_settings.dc.serial_nr))
Assemble the actual Suunto serial number It turns out that the serial number returned by libdivecomputer isn't really the serial number as interpreted by the vendor. Those tend to be strings, but libdivecomputer gives us a 32bit number. Some experimenting showed that for the Suunto devies tested the serial number is encoded in that 32bit number: It so happens that the Suunto serial number strings are strings that have all numbers, but they aren't *one* number. They are four bytes representing two numbers each, and the "23500027" string is actually the four bytes 23 50 00 27 (0x17 0x32 0x00 0x1b). And libdivecomputer has incorrectly parsed those four bytes as one number, not as the encoded serial number string it is. So the value 389152795 is actually hex 0x1732001b, which is 0x17 0x32 0x00 0x1b, which is - 23 50 00 27. This should be done by libdivecomputer, but hey, in the meantime this at least shows the concept. And helps test the XML save/restore code. It depends on the two patches that create the whole "device.c" infrastructure, of course. With this, my dive file ends up having the settings section look like this: <divecomputerid model='Suunto Vyper Air' deviceid='d4629110' serial='01201094' firmware='1.1.22'/> <divecomputerid model='Suunto HelO2' deviceid='995dd566' serial='23500027' firmware='1.0.4'/> where the format of the firmware version is something I guessed at, but it was the obvious choice (again, it's byte-based, I'm ignoring the high byte that is zero for both of my Suuntos). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2013-01-10 00:14:21 +00:00
return;
if (MATCH("firmware.divecomputerid", utf8_string, &cur_settings.dc.firmware))
Assemble the actual Suunto serial number It turns out that the serial number returned by libdivecomputer isn't really the serial number as interpreted by the vendor. Those tend to be strings, but libdivecomputer gives us a 32bit number. Some experimenting showed that for the Suunto devies tested the serial number is encoded in that 32bit number: It so happens that the Suunto serial number strings are strings that have all numbers, but they aren't *one* number. They are four bytes representing two numbers each, and the "23500027" string is actually the four bytes 23 50 00 27 (0x17 0x32 0x00 0x1b). And libdivecomputer has incorrectly parsed those four bytes as one number, not as the encoded serial number string it is. So the value 389152795 is actually hex 0x1732001b, which is 0x17 0x32 0x00 0x1b, which is - 23 50 00 27. This should be done by libdivecomputer, but hey, in the meantime this at least shows the concept. And helps test the XML save/restore code. It depends on the two patches that create the whole "device.c" infrastructure, of course. With this, my dive file ends up having the settings section look like this: <divecomputerid model='Suunto Vyper Air' deviceid='d4629110' serial='01201094' firmware='1.1.22'/> <divecomputerid model='Suunto HelO2' deviceid='995dd566' serial='23500027' firmware='1.0.4'/> where the format of the firmware version is something I guessed at, but it was the obvious choice (again, it's byte-based, I'm ignoring the high byte that is zero for both of my Suuntos). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2013-01-10 00:14:21 +00:00
return;
nonmatch("divecomputerid", name, buf);
}
static void try_to_fill_event(const char *name, char *buf)
{
start_match("event", name, buf);
if (MATCH("event", utf8_string, &cur_event.name))
return;
if (MATCH("name", utf8_string, &cur_event.name))
return;
if (MATCH("time", eventtime, &cur_event.time))
return;
if (MATCH("type", get_index, &cur_event.type))
return;
if (MATCH("flags", get_index, &cur_event.flags))
return;
if (MATCH("value", get_index, &cur_event.value))
return;
nonmatch("event", name, buf);
}
static int match_dc_data_fields(struct divecomputer *dc, const char *name, char *buf)
{
if (MATCH("maxdepth", depth, &dc->maxdepth))
return 1;
if (MATCH("meandepth", depth, &dc->meandepth))
return 1;
if (MATCH("max.depth", depth, &dc->maxdepth))
return 1;
if (MATCH("mean.depth", depth, &dc->meandepth))
return 1;
if (MATCH("duration", duration, &dc->duration))
return 1;
if (MATCH("divetime", duration, &dc->duration))
return 1;
if (MATCH("divetimesec", duration, &dc->duration))
return 1;
if (MATCH("surfacetime", duration, &dc->surfacetime))
return 1;
if (MATCH("airtemp", temperature, &dc->airtemp))
return 1;
if (MATCH("watertemp", temperature, &dc->watertemp))
return 1;
if (MATCH("air.temperature", temperature, &dc->airtemp))
return 1;
if (MATCH("water.temperature", temperature, &dc->watertemp))
return 1;
if (MATCH("pressure.surface", pressure, &dc->surface_pressure))
return 1;
if (MATCH("salinity.water", salinity, &dc->salinity))
return 1;
return 0;
}
/* We're in the top-level dive xml. Try to convert whatever value to a dive value */
static void try_to_fill_dc(struct divecomputer *dc, const char *name, char *buf)
{
start_match("divecomputer", name, buf);
if (MATCH("date", divedate, &dc->when))
return;
if (MATCH("time", divetime, &dc->when))
return;
if (MATCH("model", utf8_string, &dc->model))
return;
if (MATCH("deviceid", hex_value, &dc->deviceid))
return;
if (MATCH("diveid", hex_value, &dc->diveid))
return;
if (MATCH("dctype", get_dc_type, &dc->dctype))
return;
if (MATCH("no_o2sensors", get_sensor, &dc->no_o2sensors))
return;
if (match_dc_data_fields(dc, name, buf))
return;
nonmatch("divecomputer", name, buf);
}
/* We're in samples - try to convert the random xml value to something useful */
static void try_to_fill_sample(struct sample *sample, const char *name, char *buf)
{
int in_deco;
start_match("sample", name, buf);
if (MATCH("pressure.sample", pressure, &sample->cylinderpressure))
return;
if (MATCH("cylpress.sample", pressure, &sample->cylinderpressure))
return;
if (MATCH("pdiluent.sample", pressure, &sample->diluentpressure))
return;
if (MATCH("cylinderindex.sample", get_cylinderindex, &sample->sensor))
First step in cleaning up cylinder pressure sensor logic This clarifies/changes the meaning of our "cylinderindex" entry in our samples. It has been rather confused, because different dive computers have done things differently, and the naming really hasn't helped. There are two totally different - and independent - cylinder "indexes": - the pressure sensor index, which indicates which cylinder the sensor data is from. - the "active cylinder" index, which indicates which cylinder we actually breathe from. These two values really are totally independent, and have nothing what-so-ever to do with each other. The sensor index may well be fixed: many dive computers only support a single pressure sensor (whether wireless or wired), and the sensor index is thus always zero. Other dive computers may support multiple pressure sensors, and the gas switch event may - or may not - indicate that the sensor changed too. A dive computer might give the sensor data for *all* cylinders it can read, regardless of which one is the one we're actively breathing. In fact, some dive computers might give sensor data for not just *your* cylinder, but your buddies. This patch renames "cylinderindex" in the samples as "sensor", making it quite clear that it's about which sensor index the pressure data in the sample is about. The way we figure out which is the currently active gas is with an explicit has change event. If a computer (like the Uemis Zurich) joins the two concepts together, then a sensor change should also create a gas switch event. This patch also changes the Uemis importer to do that. Finally, it should be noted that the plot info works totally separately from the sample data, and is about what we actually *display*, not about the sample pressures etc. In the plot info, the "cylinderindex" does in fact mean the currently active cylinder, and while it is initially set to match the sensor information from the samples, we then walk the gas change events and fix it up - and if the active cylinder differs from the sensor cylinder, we clear the sensor data. [Dirk Hohndel: this conflicted with some of my recent changes - I think I merged things correctly...] Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2012-12-31 04:00:51 +00:00
return;
if (MATCH("sensor.sample", get_sensor, &sample->sensor))
return;
if (MATCH("depth.sample", depth, &sample->depth))
return;
if (MATCH("temp.sample", temperature, &sample->temperature))
return;
if (MATCH("temperature.sample", temperature, &sample->temperature))
return;
if (MATCH("sampletime.sample", sampletime, &sample->time))
return;
if (MATCH("time.sample", sampletime, &sample->time))
return;
if (MATCH("ndl.sample", sampletime, &sample->ndl))
return;
if (MATCH("tts.sample", sampletime, &sample->tts))
return;
if (MATCH("in_deco.sample", get_index, &in_deco)) {
sample->in_deco = (in_deco == 1);
return;
}
if (MATCH("stoptime.sample", sampletime, &sample->stoptime))
return;
if (MATCH("stopdepth.sample", depth, &sample->stopdepth))
return;
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
if (MATCH("cns.sample", get_uint8, &sample->cns))
return;
if (MATCH("sensor1.sample", double_to_o2pressure, &sample->o2sensor[0])) // CCR O2 sensor data
return;
if (MATCH("sensor2.sample", double_to_o2pressure, &sample->o2sensor[1]))
return;
if (MATCH("sensor3.sample", double_to_o2pressure, &sample->o2sensor[2])) // up to 3 CCR sensors
return;
if (MATCH("setpoint.sample", double_to_o2pressure, &sample->o2setpoint))
return;
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
if (MATCH("po2.sample", double_to_o2pressure, &sample->po2))
return;
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
if (MATCH("heartbeat", get_uint8, &sample->heartbeat))
return;
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
if (MATCH("bearing", get_bearing, &sample->bearing))
return;
switch (import_source) {
case DIVINGLOG:
if (divinglog_fill_sample(sample, name, buf))
return;
break;
case UDDF:
if (uddf_fill_sample(sample, name, buf))
return;
break;
default:
break;
}
nonmatch("sample", name, buf);
}
void try_to_fill_userid(const char *name, char *buf)
{
if (prefs.save_userid_local)
set_userid(buf);
}
static const char *country, *city;
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
static void divinglog_place(char *place, char **location)
{
char buffer[1024], *p;
int len;
len = snprintf(buffer, sizeof(buffer),
"%s%s%s%s%s",
place,
city ? ", " : "",
city ? city : "",
country ? ", " : "",
country ? country : "");
p = malloc(len + 1);
memcpy(p, buffer, len + 1);
*location = p;
city = NULL;
country = NULL;
}
static int divinglog_dive_match(struct dive *dive, const char *name, char *buf)
{
return MATCH("divedate", divedate, &dive->when) ||
MATCH("entrytime", divetime, &dive->when) ||
MATCH("divetime", duration, &dive->dc.duration) ||
MATCH("depth", depth, &dive->dc.maxdepth) ||
MATCH("depthavg", depth, &dive->dc.meandepth) ||
MATCH("tanktype", utf8_string, &dive->cylinder[0].type.description) ||
MATCH("tanksize", cylindersize, &dive->cylinder[0].type.size) ||
MATCH("presw", pressure, &dive->cylinder[0].type.workingpressure) ||
MATCH("press", pressure, &dive->cylinder[0].start) ||
MATCH("prese", pressure, &dive->cylinder[0].end) ||
MATCH("comments", utf8_string, &dive->notes) ||
MATCH("names.buddy", utf8_string, &dive->buddy) ||
MATCH("name.country", utf8_string, &country) ||
MATCH("name.city", utf8_string, &city) ||
MATCH("name.place", divinglog_place, &dive->location) ||
0;
}
/*
* Uddf specifies ISO 8601 time format.
*
* There are many variations on that. This handles the useful cases.
*/
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
static void uddf_datetime(char *buffer, timestamp_t *when)
{
char c;
int y, m, d, hh, mm, ss;
struct tm tm = { 0 };
int i;
i = sscanf(buffer, "%d-%d-%d%c%d:%d:%d", &y, &m, &d, &c, &hh, &mm, &ss);
if (i == 7)
goto success;
ss = 0;
if (i == 6)
goto success;
i = sscanf(buffer, "%04d%02d%02d%c%02d%02d%02d", &y, &m, &d, &c, &hh, &mm, &ss);
if (i == 7)
goto success;
ss = 0;
if (i == 6)
goto success;
bad_date:
printf("Bad date time %s\n", buffer);
return;
success:
if (c != 'T' && c != ' ')
goto bad_date;
tm.tm_year = y;
tm.tm_mon = m - 1;
tm.tm_mday = d;
tm.tm_hour = hh;
tm.tm_min = mm;
tm.tm_sec = ss;
*when = utc_mktime(&tm);
}
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
#define uddf_datedata(name, offset) \
static void uddf_##name(char *buffer, timestamp_t *when) \
{ \
cur_tm.tm_##name = atoi(buffer) + offset; \
*when = utc_mktime(&cur_tm); \
}
uddf_datedata(year, 0)
uddf_datedata(mon, -1)
uddf_datedata(mday, 0)
uddf_datedata(hour, 0)
uddf_datedata(min, 0)
static int uddf_dive_match(struct dive *dive, const char *name, char *buf)
{
return MATCH("datetime", uddf_datetime, &dive->when) ||
MATCH("diveduration", duration, &dive->dc.duration) ||
MATCH("greatestdepth", depth, &dive->dc.maxdepth) ||
MATCH("year.date", uddf_year, &dive->when) ||
MATCH("month.date", uddf_mon, &dive->when) ||
MATCH("day.date", uddf_mday, &dive->when) ||
MATCH("hour.time", uddf_hour, &dive->when) ||
MATCH("minute.time", uddf_min, &dive->when) ||
0;
}
/*
* This parses "floating point" into micro-degrees.
* We don't do exponentials etc, if somebody does
* gps locations in that format, they are insane.
*/
degrees_t parse_degrees(char *buf, char **end)
{
int sign = 1, decimals = 6, value = 0;
degrees_t ret;
while (isspace(*buf))
buf++;
switch (*buf) {
case '-':
sign = -1;
/* fallthrough */
case '+':
buf++;
}
while (isdigit(*buf)) {
value = 10 * value + *buf - '0';
buf++;
}
/* Get the first six decimals if they exist */
if (*buf == '.')
buf++;
do {
value *= 10;
if (isdigit(*buf)) {
value += *buf - '0';
buf++;
}
} while (--decimals);
/* Rounding */
switch (*buf) {
case '5' ... '9':
value++;
}
while (isdigit(*buf))
buf++;
*end = buf;
ret.udeg = value * sign;
return ret;
}
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
static void gps_lat(char *buffer, struct dive *dive)
{
char *end;
dive->latitude = parse_degrees(buffer, &end);
}
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
static void gps_long(char *buffer, struct dive *dive)
{
char *end;
dive->longitude = parse_degrees(buffer, &end);
}
Make parse-xml callbacks be type-safe .. and fix the type breakage brought in by commit eaf6d564874a ("CCR code: Change to sample structure") The XML parsing callbacks pass a "void *" around, because the helper function that matches the XML node names ("match()") does so for all the different dive/sample/dc member nodes that all have different types. But that also hid the fact that it very much depended on the various types being regular "int" etc, rather than the denser types that were introduced so that the CCR data wouldn't expand memory use excessively. As a result, XML loading would overwrite other members, and possibly even the allocation, when it wrote an "int" value to something that only was a 8-bit allocation. I left the "utf8_string()" without type checking - so it still uses "void *_res" for the result type, with the cast happening inside the function. That's because the result destination ends up being a bit mixed-up wrt "const char **" and just plain "char **". Note that the thing we modify itself isn't const (it's not "char *const *"), but the pointer, but we basically sometimes assign a "const char *", and sometimes a "char *". I considered making two different versions of the callback, but it just wasn't worth it. So "utf8_string()" users still aren't type-checked, and you'd better give it a pointer to something that is some kind of "char *" This patch doesn't really change the calling convention of the matching function itself, but it makes the wrapper macro ("MATCH()") take a properly type-checked function pointer instead (with a dummy call to do type checking), and then casts the pointer to the "void *" type for the actual real call. The function pointer call is not really portable (although it works on all sane architectures, particularly since the cast only changes one argument from one type of pointer to another), and to make matters worse uses the gcc statement-expression extension. But all the compilers we use seem to support that gcc'ism, so in practice this gives us type-safety with no downsides. (If we ever want to use MSVC to compile subsurface, I suspect we'll have to ifdef out the statement expression use and not type-check things. Or perhaps re-write the thing as a ternary expression instead, or something). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-06-07 21:41:07 +00:00
static void gps_location(char *buffer, struct dive *dive)
{
char *end;
dive->latitude = parse_degrees(buffer, &end);
dive->longitude = parse_degrees(end, &end);
}
static void gps_picture_location(char *buffer, struct picture *pic)
{
char *end;
pic->latitude = parse_degrees(buffer, &end);
pic->longitude = parse_degrees(end, &end);
}
/* We're in the top-level dive xml. Try to convert whatever value to a dive value */
static void try_to_fill_dive(struct dive *dive, const char *name, char *buf)
{
start_match("dive", name, buf);
switch (import_source) {
case DIVINGLOG:
if (divinglog_dive_match(dive, name, buf))
return;
break;
case UDDF:
if (uddf_dive_match(dive, name, buf))
return;
break;
default:
break;
}
if (MATCH("number", get_index, &dive->number))
return;
Get rid of crazy empty tag_list element at the start So this is totally unrelated to the git repository format, except for the fact that I noticed it while writing the git saving code. The subsurface divetag list handling is being stupid, and has a initial dummy entry at the head of the list for no good reason. I say "no good reason", because there *is* a reason for it: it allows code to avoid the special case of empty list and adding entries to before the first entry etc etc. But that reason is a really *bad* reason, because it's valid only because people don't understand basic list manipulation and pointers to pointers. So get rid of the dummy element, and do things right instead - by passing a *pointer* to the list, instead of the list. And then when traversing the list and looking for a place to insert things, don't go to the next entry - just update the "pointer to pointer" to point to the address of the next entry. Each entry in a C linked list is no different than the list itself, so you can use the pointer to the pointer to the next entry as a pointer to the list. This is a pet peeve of mine. The real beauty of pointers can never be understood unless you understand the indirection they allow. People who grew up with Pascal and were corrupted by that mindset are mentally stunted. Niklaus Wirth has a lot to answer for! But never fear. You too can overcome that mental limitation, it just needs some brain exercise. Reading this patch may help. In particular, contemplate the new "taglist_add_divetag()". Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-03-10 17:18:13 +00:00
if (MATCH("tags", divetags, &dive->tag_list))
return;
if (MATCH("tripflag", get_tripflag, &dive->tripflag))
First cut of explicit trip tracking This code establishes the explicit trip data structures and loads and saves them in the XML data. No attempts are made to edit / modify the trips, yet. Loading XML files without trip data creates the trips based on timing as before. Saving out the same, unmodified data will create 'trip' entries in the XML file with a 'number' that reflects the number of dives in that trip. The trip tag also stores the beginning time of the first dive in the trip and the location of the trip (which we display in the summary entries in the UI). The logic allows for dives that aren't part of a dive trip. All other dives simply belong to the "previous" dive trip - i.e. the dive trip with the latest start time that is earlier or equal to the start time of this dive. This logic significantly simplifies the tracking of trips compared to other approaches that I have tried. The automatic grouping into trips now is an option that defaults to off (as it makes changes to the XML file - and people who don't want this feature shouldn't have trips added to their XML files that they then need to manually remove). For now you have to select this option, then exit the program and start it again. Still to do is to trigger the trip generation at run time. We also need a way to mark dives as not part of trips and to allow options to combine trips, split trips, edit trip location data, etc. The code has only had some limited testing when opening multiple files. The code is known to fail if a location name contains unquoted special characters like an "'". This commit also fixes a visual inconsistency in the preferences dialog where the font selector button didn't have a frame around it that told you what this option was about. Inspired-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2012-08-22 05:04:24 +00:00
return;
if (MATCH("date", divedate, &dive->when))
return;
if (MATCH("time", divetime, &dive->when))
return;
if (MATCH("datetime", divedatetime, &dive->when))
return;
/*
* Legacy format note: per-dive depths and duration get saved
* in the first dive computer entry
*/
if (match_dc_data_fields(&dive->dc, name, buf))
return;
if (MATCH("filename.picture", utf8_string, &cur_picture->filename))
return;
if (MATCH("offset.picture", offsettime, &cur_picture->offset))
return;
if (MATCH("gps.picture", gps_picture_location, cur_picture))
return;
if (MATCH("cylinderstartpressure", pressure, &dive->cylinder[0].start))
return;
if (MATCH("cylinderendpressure", pressure, &dive->cylinder[0].end))
return;
if (MATCH("gps", gps_location, dive))
return;
if (MATCH("Place", gps_location, dive))
return;
if (MATCH("latitude", gps_lat, dive))
return;
if (MATCH("sitelat", gps_lat, dive))
return;
if (MATCH("lat", gps_lat, dive))
return;
if (MATCH("longitude", gps_long, dive))
return;
if (MATCH("sitelon", gps_long, dive))
return;
if (MATCH("lon", gps_long, dive))
return;
if (MATCH("location", utf8_string, &dive->location))
return;
if (MATCH("name.dive", utf8_string, &dive->location))
return;
if (MATCH("suit", utf8_string, &dive->suit))
return;
if (MATCH("divesuit", utf8_string, &dive->suit))
return;
if (MATCH("notes", utf8_string, &dive->notes))
return;
if (MATCH("divemaster", utf8_string, &dive->divemaster))
return;
if (MATCH("buddy", utf8_string, &dive->buddy))
return;
if (MATCH("rating.dive", get_rating, &dive->rating))
return;
if (MATCH("visibility.dive", get_rating, &dive->visibility))
return;
if (MATCH("size.cylinder", cylindersize, &dive->cylinder[cur_cylinder_index].type.size))
return;
if (MATCH("workpressure.cylinder", pressure, &dive->cylinder[cur_cylinder_index].type.workingpressure))
return;
if (MATCH("description.cylinder", utf8_string, &dive->cylinder[cur_cylinder_index].type.description))
return;
if (MATCH("start.cylinder", pressure, &dive->cylinder[cur_cylinder_index].start))
return;
if (MATCH("end.cylinder", pressure, &dive->cylinder[cur_cylinder_index].end))
return;
if (MATCH("description.weightsystem", utf8_string, &dive->weightsystem[cur_ws_index].description))
return;
if (MATCH("weight.weightsystem", weight, &dive->weightsystem[cur_ws_index].weight))
return;
if (MATCH("weight", weight, &dive->weightsystem[cur_ws_index].weight))
return;
if (MATCH("o2", gasmix, &dive->cylinder[cur_cylinder_index].gasmix.o2))
return;
if (MATCH("o2percent", gasmix, &dive->cylinder[cur_cylinder_index].gasmix.o2))
return;
if (MATCH("n2", gasmix_nitrogen, &dive->cylinder[cur_cylinder_index].gasmix))
return;
if (MATCH("he", gasmix, &dive->cylinder[cur_cylinder_index].gasmix.he))
return;
if (MATCH("air.divetemperature", temperature, &dive->airtemp))
return;
if (MATCH("water.divetemperature", temperature, &dive->watertemp))
return;
nonmatch("dive", name, buf);
}
First cut of explicit trip tracking This code establishes the explicit trip data structures and loads and saves them in the XML data. No attempts are made to edit / modify the trips, yet. Loading XML files without trip data creates the trips based on timing as before. Saving out the same, unmodified data will create 'trip' entries in the XML file with a 'number' that reflects the number of dives in that trip. The trip tag also stores the beginning time of the first dive in the trip and the location of the trip (which we display in the summary entries in the UI). The logic allows for dives that aren't part of a dive trip. All other dives simply belong to the "previous" dive trip - i.e. the dive trip with the latest start time that is earlier or equal to the start time of this dive. This logic significantly simplifies the tracking of trips compared to other approaches that I have tried. The automatic grouping into trips now is an option that defaults to off (as it makes changes to the XML file - and people who don't want this feature shouldn't have trips added to their XML files that they then need to manually remove). For now you have to select this option, then exit the program and start it again. Still to do is to trigger the trip generation at run time. We also need a way to mark dives as not part of trips and to allow options to combine trips, split trips, edit trip location data, etc. The code has only had some limited testing when opening multiple files. The code is known to fail if a location name contains unquoted special characters like an "'". This commit also fixes a visual inconsistency in the preferences dialog where the font selector button didn't have a frame around it that told you what this option was about. Inspired-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2012-08-22 05:04:24 +00:00
/* We're in the top-level trip xml. Try to convert whatever value to a trip value */
static void try_to_fill_trip(dive_trip_t **dive_trip_p, const char *name, char *buf)
First cut of explicit trip tracking This code establishes the explicit trip data structures and loads and saves them in the XML data. No attempts are made to edit / modify the trips, yet. Loading XML files without trip data creates the trips based on timing as before. Saving out the same, unmodified data will create 'trip' entries in the XML file with a 'number' that reflects the number of dives in that trip. The trip tag also stores the beginning time of the first dive in the trip and the location of the trip (which we display in the summary entries in the UI). The logic allows for dives that aren't part of a dive trip. All other dives simply belong to the "previous" dive trip - i.e. the dive trip with the latest start time that is earlier or equal to the start time of this dive. This logic significantly simplifies the tracking of trips compared to other approaches that I have tried. The automatic grouping into trips now is an option that defaults to off (as it makes changes to the XML file - and people who don't want this feature shouldn't have trips added to their XML files that they then need to manually remove). For now you have to select this option, then exit the program and start it again. Still to do is to trigger the trip generation at run time. We also need a way to mark dives as not part of trips and to allow options to combine trips, split trips, edit trip location data, etc. The code has only had some limited testing when opening multiple files. The code is known to fail if a location name contains unquoted special characters like an "'". This commit also fixes a visual inconsistency in the preferences dialog where the font selector button didn't have a frame around it that told you what this option was about. Inspired-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2012-08-22 05:04:24 +00:00
{
start_match("trip", name, buf);
dive_trip_t *dive_trip = *dive_trip_p;
First cut of explicit trip tracking This code establishes the explicit trip data structures and loads and saves them in the XML data. No attempts are made to edit / modify the trips, yet. Loading XML files without trip data creates the trips based on timing as before. Saving out the same, unmodified data will create 'trip' entries in the XML file with a 'number' that reflects the number of dives in that trip. The trip tag also stores the beginning time of the first dive in the trip and the location of the trip (which we display in the summary entries in the UI). The logic allows for dives that aren't part of a dive trip. All other dives simply belong to the "previous" dive trip - i.e. the dive trip with the latest start time that is earlier or equal to the start time of this dive. This logic significantly simplifies the tracking of trips compared to other approaches that I have tried. The automatic grouping into trips now is an option that defaults to off (as it makes changes to the XML file - and people who don't want this feature shouldn't have trips added to their XML files that they then need to manually remove). For now you have to select this option, then exit the program and start it again. Still to do is to trigger the trip generation at run time. We also need a way to mark dives as not part of trips and to allow options to combine trips, split trips, edit trip location data, etc. The code has only had some limited testing when opening multiple files. The code is known to fail if a location name contains unquoted special characters like an "'". This commit also fixes a visual inconsistency in the preferences dialog where the font selector button didn't have a frame around it that told you what this option was about. Inspired-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2012-08-22 05:04:24 +00:00
if (MATCH("date", divedate, &dive_trip->when))
return;
if (MATCH("time", divetime, &dive_trip->when))
First cut of explicit trip tracking This code establishes the explicit trip data structures and loads and saves them in the XML data. No attempts are made to edit / modify the trips, yet. Loading XML files without trip data creates the trips based on timing as before. Saving out the same, unmodified data will create 'trip' entries in the XML file with a 'number' that reflects the number of dives in that trip. The trip tag also stores the beginning time of the first dive in the trip and the location of the trip (which we display in the summary entries in the UI). The logic allows for dives that aren't part of a dive trip. All other dives simply belong to the "previous" dive trip - i.e. the dive trip with the latest start time that is earlier or equal to the start time of this dive. This logic significantly simplifies the tracking of trips compared to other approaches that I have tried. The automatic grouping into trips now is an option that defaults to off (as it makes changes to the XML file - and people who don't want this feature shouldn't have trips added to their XML files that they then need to manually remove). For now you have to select this option, then exit the program and start it again. Still to do is to trigger the trip generation at run time. We also need a way to mark dives as not part of trips and to allow options to combine trips, split trips, edit trip location data, etc. The code has only had some limited testing when opening multiple files. The code is known to fail if a location name contains unquoted special characters like an "'". This commit also fixes a visual inconsistency in the preferences dialog where the font selector button didn't have a frame around it that told you what this option was about. Inspired-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2012-08-22 05:04:24 +00:00
return;
if (MATCH("location", utf8_string, &dive_trip->location))
First cut of explicit trip tracking This code establishes the explicit trip data structures and loads and saves them in the XML data. No attempts are made to edit / modify the trips, yet. Loading XML files without trip data creates the trips based on timing as before. Saving out the same, unmodified data will create 'trip' entries in the XML file with a 'number' that reflects the number of dives in that trip. The trip tag also stores the beginning time of the first dive in the trip and the location of the trip (which we display in the summary entries in the UI). The logic allows for dives that aren't part of a dive trip. All other dives simply belong to the "previous" dive trip - i.e. the dive trip with the latest start time that is earlier or equal to the start time of this dive. This logic significantly simplifies the tracking of trips compared to other approaches that I have tried. The automatic grouping into trips now is an option that defaults to off (as it makes changes to the XML file - and people who don't want this feature shouldn't have trips added to their XML files that they then need to manually remove). For now you have to select this option, then exit the program and start it again. Still to do is to trigger the trip generation at run time. We also need a way to mark dives as not part of trips and to allow options to combine trips, split trips, edit trip location data, etc. The code has only had some limited testing when opening multiple files. The code is known to fail if a location name contains unquoted special characters like an "'". This commit also fixes a visual inconsistency in the preferences dialog where the font selector button didn't have a frame around it that told you what this option was about. Inspired-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2012-08-22 05:04:24 +00:00
return;
if (MATCH("notes", utf8_string, &dive_trip->notes))
First cut of explicit trip tracking This code establishes the explicit trip data structures and loads and saves them in the XML data. No attempts are made to edit / modify the trips, yet. Loading XML files without trip data creates the trips based on timing as before. Saving out the same, unmodified data will create 'trip' entries in the XML file with a 'number' that reflects the number of dives in that trip. The trip tag also stores the beginning time of the first dive in the trip and the location of the trip (which we display in the summary entries in the UI). The logic allows for dives that aren't part of a dive trip. All other dives simply belong to the "previous" dive trip - i.e. the dive trip with the latest start time that is earlier or equal to the start time of this dive. This logic significantly simplifies the tracking of trips compared to other approaches that I have tried. The automatic grouping into trips now is an option that defaults to off (as it makes changes to the XML file - and people who don't want this feature shouldn't have trips added to their XML files that they then need to manually remove). For now you have to select this option, then exit the program and start it again. Still to do is to trigger the trip generation at run time. We also need a way to mark dives as not part of trips and to allow options to combine trips, split trips, edit trip location data, etc. The code has only had some limited testing when opening multiple files. The code is known to fail if a location name contains unquoted special characters like an "'". This commit also fixes a visual inconsistency in the preferences dialog where the font selector button didn't have a frame around it that told you what this option was about. Inspired-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2012-08-22 05:04:24 +00:00
return;
nonmatch("trip", name, buf);
}
/*
* While in some formats file boundaries are dive boundaries, in many
* others (as for example in our native format) there are
* multiple dives per file, so there can be other events too that
* trigger a "new dive" marker and you may get some nesting due
* to that. Just ignore nesting levels.
* On the flipside it is possible that we start an XML file that ends
* up having no dives in it at all - don't create a bogus empty dive
* for those. It's not entirely clear what is the minimum set of data
* to make a dive valid, but if it has no location, no date and no
* samples I'm pretty sure it's useless.
*/
static bool is_dive(void)
{
return (cur_dive &&
(cur_dive->location || cur_dive->when || cur_dive->dc.samples));
}
static void reset_dc_info(struct divecomputer *dc)
{
lastcns = lastpo2 = lastndl = laststoptime = laststopdepth = lastindeco = 0;
First step in cleaning up cylinder pressure sensor logic This clarifies/changes the meaning of our "cylinderindex" entry in our samples. It has been rather confused, because different dive computers have done things differently, and the naming really hasn't helped. There are two totally different - and independent - cylinder "indexes": - the pressure sensor index, which indicates which cylinder the sensor data is from. - the "active cylinder" index, which indicates which cylinder we actually breathe from. These two values really are totally independent, and have nothing what-so-ever to do with each other. The sensor index may well be fixed: many dive computers only support a single pressure sensor (whether wireless or wired), and the sensor index is thus always zero. Other dive computers may support multiple pressure sensors, and the gas switch event may - or may not - indicate that the sensor changed too. A dive computer might give the sensor data for *all* cylinders it can read, regardless of which one is the one we're actively breathing. In fact, some dive computers might give sensor data for not just *your* cylinder, but your buddies. This patch renames "cylinderindex" in the samples as "sensor", making it quite clear that it's about which sensor index the pressure data in the sample is about. The way we figure out which is the currently active gas is with an explicit has change event. If a computer (like the Uemis Zurich) joins the two concepts together, then a sensor change should also create a gas switch event. This patch also changes the Uemis importer to do that. Finally, it should be noted that the plot info works totally separately from the sample data, and is about what we actually *display*, not about the sample pressures etc. In the plot info, the "cylinderindex" does in fact mean the currently active cylinder, and while it is initially set to match the sensor information from the samples, we then walk the gas change events and fix it up - and if the active cylinder differs from the sensor cylinder, we clear the sensor data. [Dirk Hohndel: this conflicted with some of my recent changes - I think I merged things correctly...] Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2012-12-31 04:00:51 +00:00
lastsensor = lastcylinderindex = 0;
}
static void reset_dc_settings(void)
{
free((void *)cur_settings.dc.model);
free((void *)cur_settings.dc.nickname);
Assemble the actual Suunto serial number It turns out that the serial number returned by libdivecomputer isn't really the serial number as interpreted by the vendor. Those tend to be strings, but libdivecomputer gives us a 32bit number. Some experimenting showed that for the Suunto devies tested the serial number is encoded in that 32bit number: It so happens that the Suunto serial number strings are strings that have all numbers, but they aren't *one* number. They are four bytes representing two numbers each, and the "23500027" string is actually the four bytes 23 50 00 27 (0x17 0x32 0x00 0x1b). And libdivecomputer has incorrectly parsed those four bytes as one number, not as the encoded serial number string it is. So the value 389152795 is actually hex 0x1732001b, which is 0x17 0x32 0x00 0x1b, which is - 23 50 00 27. This should be done by libdivecomputer, but hey, in the meantime this at least shows the concept. And helps test the XML save/restore code. It depends on the two patches that create the whole "device.c" infrastructure, of course. With this, my dive file ends up having the settings section look like this: <divecomputerid model='Suunto Vyper Air' deviceid='d4629110' serial='01201094' firmware='1.1.22'/> <divecomputerid model='Suunto HelO2' deviceid='995dd566' serial='23500027' firmware='1.0.4'/> where the format of the firmware version is something I guessed at, but it was the obvious choice (again, it's byte-based, I'm ignoring the high byte that is zero for both of my Suuntos). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2013-01-10 00:14:21 +00:00
free((void *)cur_settings.dc.serial_nr);
free((void *)cur_settings.dc.firmware);
cur_settings.dc.model = NULL;
cur_settings.dc.nickname = NULL;
Assemble the actual Suunto serial number It turns out that the serial number returned by libdivecomputer isn't really the serial number as interpreted by the vendor. Those tend to be strings, but libdivecomputer gives us a 32bit number. Some experimenting showed that for the Suunto devies tested the serial number is encoded in that 32bit number: It so happens that the Suunto serial number strings are strings that have all numbers, but they aren't *one* number. They are four bytes representing two numbers each, and the "23500027" string is actually the four bytes 23 50 00 27 (0x17 0x32 0x00 0x1b). And libdivecomputer has incorrectly parsed those four bytes as one number, not as the encoded serial number string it is. So the value 389152795 is actually hex 0x1732001b, which is 0x17 0x32 0x00 0x1b, which is - 23 50 00 27. This should be done by libdivecomputer, but hey, in the meantime this at least shows the concept. And helps test the XML save/restore code. It depends on the two patches that create the whole "device.c" infrastructure, of course. With this, my dive file ends up having the settings section look like this: <divecomputerid model='Suunto Vyper Air' deviceid='d4629110' serial='01201094' firmware='1.1.22'/> <divecomputerid model='Suunto HelO2' deviceid='995dd566' serial='23500027' firmware='1.0.4'/> where the format of the firmware version is something I guessed at, but it was the obvious choice (again, it's byte-based, I'm ignoring the high byte that is zero for both of my Suuntos). Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2013-01-10 00:14:21 +00:00
cur_settings.dc.serial_nr = NULL;
cur_settings.dc.firmware = NULL;
cur_settings.dc.deviceid = 0;
}
static void settings_start(void)
{
in_settings = true;
}
static void settings_end(void)
{
in_settings = false;
}
static void dc_settings_start(void)
{
reset_dc_settings();
}
static void dc_settings_end(void)
{
create_device_node(cur_settings.dc.model, cur_settings.dc.deviceid, cur_settings.dc.serial_nr,
cur_settings.dc.firmware, cur_settings.dc.nickname);
reset_dc_settings();
}
static void dive_start(void)
{
if (cur_dive)
return;
cur_dive = alloc_dive();
reset_dc_info(&cur_dive->dc);
memset(&cur_tm, 0, sizeof(cur_tm));
if (cur_trip) {
2012-11-10 18:51:03 +00:00
add_dive_to_trip(cur_dive, cur_trip);
cur_dive->tripflag = IN_TRIP;
}
}
static void dive_end(void)
{
if (!cur_dive)
return;
if (!is_dive())
free(cur_dive);
else
record_dive_to_table(cur_dive, target_table);
cur_dive = NULL;
cur_dc = NULL;
cur_cylinder_index = 0;
cur_ws_index = 0;
}
First cut of explicit trip tracking This code establishes the explicit trip data structures and loads and saves them in the XML data. No attempts are made to edit / modify the trips, yet. Loading XML files without trip data creates the trips based on timing as before. Saving out the same, unmodified data will create 'trip' entries in the XML file with a 'number' that reflects the number of dives in that trip. The trip tag also stores the beginning time of the first dive in the trip and the location of the trip (which we display in the summary entries in the UI). The logic allows for dives that aren't part of a dive trip. All other dives simply belong to the "previous" dive trip - i.e. the dive trip with the latest start time that is earlier or equal to the start time of this dive. This logic significantly simplifies the tracking of trips compared to other approaches that I have tried. The automatic grouping into trips now is an option that defaults to off (as it makes changes to the XML file - and people who don't want this feature shouldn't have trips added to their XML files that they then need to manually remove). For now you have to select this option, then exit the program and start it again. Still to do is to trigger the trip generation at run time. We also need a way to mark dives as not part of trips and to allow options to combine trips, split trips, edit trip location data, etc. The code has only had some limited testing when opening multiple files. The code is known to fail if a location name contains unquoted special characters like an "'". This commit also fixes a visual inconsistency in the preferences dialog where the font selector button didn't have a frame around it that told you what this option was about. Inspired-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2012-08-22 05:04:24 +00:00
static void trip_start(void)
{
if (cur_trip)
return;
dive_end();
cur_trip = calloc(1, sizeof(dive_trip_t));
First cut of explicit trip tracking This code establishes the explicit trip data structures and loads and saves them in the XML data. No attempts are made to edit / modify the trips, yet. Loading XML files without trip data creates the trips based on timing as before. Saving out the same, unmodified data will create 'trip' entries in the XML file with a 'number' that reflects the number of dives in that trip. The trip tag also stores the beginning time of the first dive in the trip and the location of the trip (which we display in the summary entries in the UI). The logic allows for dives that aren't part of a dive trip. All other dives simply belong to the "previous" dive trip - i.e. the dive trip with the latest start time that is earlier or equal to the start time of this dive. This logic significantly simplifies the tracking of trips compared to other approaches that I have tried. The automatic grouping into trips now is an option that defaults to off (as it makes changes to the XML file - and people who don't want this feature shouldn't have trips added to their XML files that they then need to manually remove). For now you have to select this option, then exit the program and start it again. Still to do is to trigger the trip generation at run time. We also need a way to mark dives as not part of trips and to allow options to combine trips, split trips, edit trip location data, etc. The code has only had some limited testing when opening multiple files. The code is known to fail if a location name contains unquoted special characters like an "'". This commit also fixes a visual inconsistency in the preferences dialog where the font selector button didn't have a frame around it that told you what this option was about. Inspired-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2012-08-22 05:04:24 +00:00
memset(&cur_tm, 0, sizeof(cur_tm));
}
static void trip_end(void)
{
if (!cur_trip)
return;
insert_trip(&cur_trip);
First cut of explicit trip tracking This code establishes the explicit trip data structures and loads and saves them in the XML data. No attempts are made to edit / modify the trips, yet. Loading XML files without trip data creates the trips based on timing as before. Saving out the same, unmodified data will create 'trip' entries in the XML file with a 'number' that reflects the number of dives in that trip. The trip tag also stores the beginning time of the first dive in the trip and the location of the trip (which we display in the summary entries in the UI). The logic allows for dives that aren't part of a dive trip. All other dives simply belong to the "previous" dive trip - i.e. the dive trip with the latest start time that is earlier or equal to the start time of this dive. This logic significantly simplifies the tracking of trips compared to other approaches that I have tried. The automatic grouping into trips now is an option that defaults to off (as it makes changes to the XML file - and people who don't want this feature shouldn't have trips added to their XML files that they then need to manually remove). For now you have to select this option, then exit the program and start it again. Still to do is to trigger the trip generation at run time. We also need a way to mark dives as not part of trips and to allow options to combine trips, split trips, edit trip location data, etc. The code has only had some limited testing when opening multiple files. The code is known to fail if a location name contains unquoted special characters like an "'". This commit also fixes a visual inconsistency in the preferences dialog where the font selector button didn't have a frame around it that told you what this option was about. Inspired-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2012-08-22 05:04:24 +00:00
cur_trip = NULL;
}
static void event_start(void)
{
memset(&cur_event, 0, sizeof(cur_event));
cur_event.active = 1;
}
static void event_end(void)
{
struct divecomputer *dc = get_dc();
if (cur_event.name) {
if (strcmp(cur_event.name, "surface") != 0) {
/* 123 is a magic event that we used for a while to encode images in dives */
if (cur_event.type == 123) {
struct picture *pic = alloc_picture();
pic->filename = strdup(cur_event.name);
/* theoretically this could fail - but we didn't support multi year offsets */
pic->offset.seconds = cur_event.time.seconds;
dive_add_picture(cur_dive, pic);
} else {
/* At some point gas change events did not have any type. Thus we need to add
* one on import, if we encounter the type one missing.
*/
if (cur_event.type == 0 && strcmp(cur_event.name, "gaschange") == 0)
cur_event.type = cur_event.value >> 16 > 0 ? SAMPLE_EVENT_GASCHANGE2 : SAMPLE_EVENT_GASCHANGE;
add_event(dc, cur_event.time.seconds,
cur_event.type, cur_event.flags,
cur_event.value, cur_event.name);
}
}
free((void *)cur_event.name);
}
cur_event.active = 0;
}
static void picture_start(void)
{
cur_picture = alloc_picture();
}
static void picture_end(void)
{
dive_add_picture(cur_dive, cur_picture);
cur_picture = NULL;
}
static void cylinder_start(void)
{
}
static void cylinder_end(void)
{
cur_cylinder_index++;
}
static void ws_start(void)
{
}
static void ws_end(void)
{
cur_ws_index++;
}
static void sample_start(void)
{
cur_sample = prepare_sample(get_dc());
cur_sample->ndl.seconds = lastndl;
cur_sample->in_deco = lastindeco;
cur_sample->stoptime.seconds = laststoptime;
cur_sample->stopdepth.mm = laststopdepth;
cur_sample->cns = lastcns;
cur_sample->po2.mbar = lastpo2;
First step in cleaning up cylinder pressure sensor logic This clarifies/changes the meaning of our "cylinderindex" entry in our samples. It has been rather confused, because different dive computers have done things differently, and the naming really hasn't helped. There are two totally different - and independent - cylinder "indexes": - the pressure sensor index, which indicates which cylinder the sensor data is from. - the "active cylinder" index, which indicates which cylinder we actually breathe from. These two values really are totally independent, and have nothing what-so-ever to do with each other. The sensor index may well be fixed: many dive computers only support a single pressure sensor (whether wireless or wired), and the sensor index is thus always zero. Other dive computers may support multiple pressure sensors, and the gas switch event may - or may not - indicate that the sensor changed too. A dive computer might give the sensor data for *all* cylinders it can read, regardless of which one is the one we're actively breathing. In fact, some dive computers might give sensor data for not just *your* cylinder, but your buddies. This patch renames "cylinderindex" in the samples as "sensor", making it quite clear that it's about which sensor index the pressure data in the sample is about. The way we figure out which is the currently active gas is with an explicit has change event. If a computer (like the Uemis Zurich) joins the two concepts together, then a sensor change should also create a gas switch event. This patch also changes the Uemis importer to do that. Finally, it should be noted that the plot info works totally separately from the sample data, and is about what we actually *display*, not about the sample pressures etc. In the plot info, the "cylinderindex" does in fact mean the currently active cylinder, and while it is initially set to match the sensor information from the samples, we then walk the gas change events and fix it up - and if the active cylinder differs from the sensor cylinder, we clear the sensor data. [Dirk Hohndel: this conflicted with some of my recent changes - I think I merged things correctly...] Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2012-12-31 04:00:51 +00:00
cur_sample->sensor = lastsensor;
}
static void sample_end(void)
{
if (!cur_dive)
return;
finish_sample(get_dc());
lastndl = cur_sample->ndl.seconds;
lastindeco = cur_sample->in_deco;
laststoptime = cur_sample->stoptime.seconds;
laststopdepth = cur_sample->stopdepth.mm;
lastcns = cur_sample->cns;
lastpo2 = cur_sample->po2.mbar;
cur_sample = NULL;
}
static void divecomputer_start(void)
{
struct divecomputer *dc;
/* Start from the previous dive computer */
dc = &cur_dive->dc;
while (dc->next)
dc = dc->next;
/* Did we already fill that in? */
if (dc->samples || dc->model || dc->when) {
struct divecomputer *newdc = calloc(1, sizeof(*newdc));
if (newdc) {
dc->next = newdc;
dc = newdc;
}
}
/* .. this is the one we'll use */
cur_dc = dc;
reset_dc_info(dc);
}
static void divecomputer_end(void)
{
if (!cur_dc->when)
cur_dc->when = cur_dive->when;
cur_dc = NULL;
}
static void userid_start(void)
{
in_userid = true;
set_save_userid_local(true); //if the xml contains userid, keep saving it.
}
static void userid_stop(void)
{
in_userid = false;
}
static void entry(const char *name, char *buf)
{
if (in_userid) {
try_to_fill_userid(name, buf);
return;
}
if (in_settings) {
try_to_fill_dc_settings(name, buf);
try_to_match_autogroup(name, buf);
return;
}
if (cur_event.active) {
try_to_fill_event(name, buf);
return;
}
if (cur_sample) {
try_to_fill_sample(cur_sample, name, buf);
return;
}
if (cur_dc) {
try_to_fill_dc(cur_dc, name, buf);
return;
}
if (cur_dive) {
try_to_fill_dive(cur_dive, name, buf);
return;
}
if (cur_trip) {
try_to_fill_trip(&cur_trip, name, buf);
return;
}
}
static const char *nodename(xmlNode *node, char *buf, int len)
{
int levels = 2;
char *p = buf;
if (!node || !node->name)
return "root";
if (node->parent && !strcmp(node->name, "text"))
node = node->parent;
/* Make sure it's always NUL-terminated */
p[--len] = 0;
for (;;) {
const char *name = node->name;
char c;
while ((c = *name++) != 0) {
/* Cheaper 'tolower()' for ASCII */
c = (c >= 'A' && c <= 'Z') ? c - 'A' + 'a' : c;
*p++ = c;
if (!--len)
return buf;
}
*p = 0;
node = node->parent;
if (!node || !node->name)
return buf;
*p++ = '.';
if (!--len)
return buf;
if (!--levels)
return buf;
}
}
#define MAXNAME 32
static void visit_one_node(xmlNode *node)
{
char *content;
static char buffer[MAXNAME];
const char *name;
content = node->content;
if (!content || xmlIsBlankNode(node))
return;
name = nodename(node, buffer, sizeof(buffer));
entry(name, content);
}
static void traverse(xmlNode *root);
static void traverse_properties(xmlNode *node)
{
xmlAttr *p;
for (p = node->properties; p; p = p->next)
traverse(p->children);
}
static void visit(xmlNode *n)
{
visit_one_node(n);
traverse_properties(n);
traverse(n->children);
}
static void DivingLog_importer(void)
{
import_source = DIVINGLOG;
/*
* Diving Log units are really strange.
*
* Temperatures are in C, except in samples,
* when they are in Fahrenheit. Depths are in
* meters, an dpressure is in PSI in the samples,
* but in bar when it comes to working pressure.
*
* Crazy f*%^ morons.
*/
xml_parsing_units = SI_units;
}
static void uddf_importer(void)
{
import_source = UDDF;
xml_parsing_units = SI_units;
xml_parsing_units.pressure = PASCAL;
xml_parsing_units.temperature = KELVIN;
}
/*
* I'm sure this could be done as some fancy DTD rules.
* It's just not worth the headache.
*/
static struct nesting {
const char *name;
void (*start)(void), (*end)(void);
} nesting[] = {
{ "divecomputerid", dc_settings_start, dc_settings_end },
{ "settings", settings_start, settings_end },
{ "dive", dive_start, dive_end },
{ "Dive", dive_start, dive_end },
{ "trip", trip_start, trip_end },
{ "sample", sample_start, sample_end },
{ "waypoint", sample_start, sample_end },
{ "SAMPLE", sample_start, sample_end },
{ "reading", sample_start, sample_end },
{ "event", event_start, event_end },
{ "mix", cylinder_start, cylinder_end },
{ "gasmix", cylinder_start, cylinder_end },
{ "cylinder", cylinder_start, cylinder_end },
{ "weightsystem", ws_start, ws_end },
{ "divecomputer", divecomputer_start, divecomputer_end },
{ "P", sample_start, sample_end },
{ "userid", userid_start, userid_stop},
{ "picture", picture_start, picture_end },
/* Import type recognition */
{ "Divinglog", DivingLog_importer },
{ "uddf", uddf_importer },
{ NULL, }
};
static void traverse(xmlNode *root)
{
xmlNode *n;
for (n = root; n; n = n->next) {
struct nesting *rule = nesting;
if (!n->name) {
visit(n);
continue;
}
do {
if (!strcmp(rule->name, n->name))
break;
rule++;
} while (rule->name);
if (rule->start)
rule->start();
visit(n);
if (rule->end)
rule->end();
}
}
/* Per-file reset */
static void reset_all(void)
{
/*
* We reset the units for each file. You'd think it was
* a per-dive property, but I'm not going to trust people
* to do per-dive setup. If the xml does have per-dive
* data within one file, we might have to reset it per
* dive for that format.
*/
xml_parsing_units = SI_units;
import_source = UNKNOWN;
}
/* divelog.de sends us xml files that claim to be iso-8859-1
* but once we decode the HTML encoded characters they turn
* into UTF-8 instead. So skip the incorrect encoding
* declaration and decode the HTML encoded characters */
const char *preprocess_divelog_de(const char *buffer)
{
char *ret = strstr(buffer, "<DIVELOGSDATA>");
if (ret) {
xmlParserCtxtPtr ctx;
char buf[] = "";
int i;
for (i = 0; i < strlen(ret); ++i)
if (!isascii(ret[i]))
return buffer;
ctx = xmlCreateMemoryParserCtxt(buf, sizeof(buf));
ret = xmlStringLenDecodeEntities(ctx, ret, strlen(ret), XML_SUBSTITUTE_REF, 0, 0, 0);
return ret;
}
return buffer;
}
void parse_xml_buffer(const char *url, const char *buffer, int size,
struct dive_table *table, const char **params)
{
xmlDoc *doc;
const char *res = preprocess_divelog_de(buffer);
target_table = table;
doc = xmlReadMemory(res, strlen(res), url, NULL, 0);
if (res != buffer)
free((char *)res);
if (!doc) {
report_error(translate("gettextFromC", "Failed to parse '%s'"), url);
return;
}
set_save_userid_local(false);
set_userid("");
reset_all();
dive_start();
doc = test_xslt_transforms(doc, params);
traverse(xmlDocGetRootElement(doc));
dive_end();
xmlFreeDoc(doc);
}
extern int dm4_events(void *handle, int columns, char **data, char **column)
{
event_start();
if (data[1])
cur_event.time.seconds = atoi(data[1]);
if (data[2]) {
switch (atoi(data[2])) {
case 1:
/* 1 Mandatory Safety Stop */
cur_event.name = strdup("safety stop (mandatory)");
break;
case 3:
/* 3 Deco */
/* What is Subsurface's term for going to
* deco? */
cur_event.name = strdup("deco");
break;
case 4:
/* 4 Ascent warning */
cur_event.name = strdup("ascent");
break;
case 5:
/* 5 Ceiling broken */
cur_event.name = strdup("violation");
break;
case 6:
/* 6 Mandatory safety stop ceiling error */
cur_event.name = strdup("violation");
break;
case 7:
/* 7 Below deco floor */
cur_event.name = strdup("below floor");
break;
case 8:
/* 8 Dive time alarm */
cur_event.name = strdup("divetime");
break;
case 9:
/* 9 Depth alarm */
cur_event.name = strdup("maxdepth");
break;
case 10:
/* 10 OLF 80% */
case 11:
/* 11 OLF 100% */
cur_event.name = strdup("OLF");
break;
case 12:
/* 12 High pO₂ */
cur_event.name = strdup("PO2");
break;
case 13:
/* 13 Air time */
cur_event.name = strdup("airtime");
break;
case 17:
/* 17 Ascent warning */
cur_event.name = strdup("ascent");
break;
case 18:
/* 18 Ceiling error */
cur_event.name = strdup("ceiling");
break;
case 19:
/* 19 Surfaced */
cur_event.name = strdup("surface");
break;
case 20:
/* 20 Deco */
cur_event.name = strdup("deco");
break;
case 22:
/* 22 Mandatory safety stop violation */
cur_event.name = strdup("violation");
break;
case 257:
/* 257 Dive active */
/* This seems to be given after surface
* when descending again. Ignoring it. */
break;
case 258:
/* 258 Bookmark */
if (data[3]) {
cur_event.name = strdup("heading");
cur_event.value = atoi(data[3]);
} else {
cur_event.name = strdup("bookmark");
}
break;
default:
cur_event.name = strdup("unknown");
cur_event.value = atoi(data[2]);
break;
}
}
event_end();
return 0;
}
extern int dm4_tags(void *handle, int columns, char **data, char **column)
{
if (data[0])
Get rid of crazy empty tag_list element at the start So this is totally unrelated to the git repository format, except for the fact that I noticed it while writing the git saving code. The subsurface divetag list handling is being stupid, and has a initial dummy entry at the head of the list for no good reason. I say "no good reason", because there *is* a reason for it: it allows code to avoid the special case of empty list and adding entries to before the first entry etc etc. But that reason is a really *bad* reason, because it's valid only because people don't understand basic list manipulation and pointers to pointers. So get rid of the dummy element, and do things right instead - by passing a *pointer* to the list, instead of the list. And then when traversing the list and looking for a place to insert things, don't go to the next entry - just update the "pointer to pointer" to point to the address of the next entry. Each entry in a C linked list is no different than the list itself, so you can use the pointer to the pointer to the next entry as a pointer to the list. This is a pet peeve of mine. The real beauty of pointers can never be understood unless you understand the indirection they allow. People who grew up with Pascal and were corrupted by that mindset are mentally stunted. Niklaus Wirth has a lot to answer for! But never fear. You too can overcome that mental limitation, it just needs some brain exercise. Reading this patch may help. In particular, contemplate the new "taglist_add_divetag()". Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2014-03-10 17:18:13 +00:00
taglist_add_tag(&cur_dive->tag_list, data[0]);
return 0;
}
extern int dm4_dive(void *param, int columns, char **data, char **column)
{
int i, interval, retval = 0;
sqlite3 *handle = (sqlite3 *)param;
float *profileBlob;
unsigned char *tempBlob;
int *pressureBlob;
char *err = NULL;
char get_events_template[] = "select * from Mark where DiveId = %d";
char get_tags_template[] = "select Text from DiveTag where DiveId = %d";
char get_events[64];
dive_start();
cur_dive->number = atoi(data[0]);
cur_dive->when = (time_t)(atol(data[1]));
if (data[2])
utf8_string(data[2], &cur_dive->notes);
/*
* DM4 stores Duration and DiveTime. It looks like DiveTime is
* 10 to 60 seconds shorter than Duration. However, I have no
* idea what is the difference and which one should be used.
* Duration = data[3]
* DiveTime = data[15]
*/
if (data[3])
cur_dive->duration.seconds = atoi(data[3]);
if (data[15])
cur_dive->dc.duration.seconds = atoi(data[15]);
/*
* TODO: the deviceid hash should be calculated here.
*/
settings_start();
dc_settings_start();
if (data[4])
utf8_string(data[4], &cur_settings.dc.serial_nr);
if (data[5])
utf8_string(data[5], &cur_settings.dc.model);
cur_settings.dc.deviceid = 0xffffffff;
dc_settings_end();
settings_end();
if (data[6])
cur_dive->dc.maxdepth.mm = atof(data[6]) * 1000;
if (data[8])
cur_dive->dc.airtemp.mkelvin = C_to_mkelvin(atoi(data[8]));
if (data[9])
cur_dive->dc.watertemp.mkelvin = C_to_mkelvin(atoi(data[9]));
/*
* TODO: handle multiple cylinders
*/
cylinder_start();
if (data[22] && atoi(data[22]) > 0)
cur_dive->cylinder[cur_cylinder_index].start.mbar = atoi(data[22]);
else if (data[10] && atoi(data[10]) > 0)
cur_dive->cylinder[cur_cylinder_index].start.mbar = atoi(data[10]);
if (data[23] && atoi(data[23]) > 0)
cur_dive->cylinder[cur_cylinder_index].end.mbar = (atoi(data[23]));
if (data[11] && atoi(data[11]) > 0)
cur_dive->cylinder[cur_cylinder_index].end.mbar = (atoi(data[11]));
if (data[12])
cur_dive->cylinder[cur_cylinder_index].type.size.mliter = (atof(data[12])) * 1000;
if (data[13])
cur_dive->cylinder[cur_cylinder_index].type.workingpressure.mbar = (atoi(data[13]));
if (data[20])
cur_dive->cylinder[cur_cylinder_index].gasmix.o2.permille = atoi(data[20]) * 10;
if (data[21])
cur_dive->cylinder[cur_cylinder_index].gasmix.he.permille = atoi(data[21]) * 10;
cylinder_end();
if (data[14])
cur_dive->dc.surface_pressure.mbar = (atoi(data[14]) * 1000);
interval = data[16] ? atoi(data[16]) : 0;
profileBlob = (float *)data[17];
tempBlob = (unsigned char *)data[18];
pressureBlob = (int *)data[19];
for (i = 0; interval && i * interval < cur_dive->duration.seconds; i++) {
sample_start();
cur_sample->time.seconds = i * interval;
if (profileBlob)
cur_sample->depth.mm = profileBlob[i] * 1000;
else
cur_sample->depth.mm = cur_dive->dc.maxdepth.mm;
if (data[18] && data[18][0])
cur_sample->temperature.mkelvin = C_to_mkelvin(tempBlob[i]);
if (data[19] && data[19][0])
cur_sample->cylinderpressure.mbar = pressureBlob[i];
sample_end();
}
snprintf(get_events, sizeof(get_events) - 1, get_events_template, cur_dive->number);
retval = sqlite3_exec(handle, get_events, &dm4_events, 0, &err);
if (retval != SQLITE_OK) {
fprintf(stderr, "%s", translate("gettextFromC", "Database query get_events failed.\n"));
return 1;
}
snprintf(get_events, sizeof(get_events) - 1, get_tags_template, cur_dive->number);
retval = sqlite3_exec(handle, get_events, &dm4_tags, 0, &err);
if (retval != SQLITE_OK) {
fprintf(stderr, "%s", translate("gettextFromC", "Database query get_tags failed.\n"));
return 1;
}
dive_end();
/*
for (i=0; i<columns;++i) {
fprintf(stderr, "%s\t", column[i]);
}
fprintf(stderr, "\n");
for (i=0; i<columns;++i) {
fprintf(stderr, "%s\t", data[i]);
}
fprintf(stderr, "\n");
//exit(0);
*/
return SQLITE_OK;
}
int parse_dm4_buffer(sqlite3 *handle, const char *url, const char *buffer, int size,
struct dive_table *table)
{
int retval;
char *err = NULL;
target_table = table;
/* StartTime is converted from Suunto's nano seconds to standard
* time. We also need epoch, not seconds since year 1. */
char get_dives[] = "select D.DiveId,StartTime/10000000-62135596800,Note,Duration,SourceSerialNumber,Source,MaxDepth,SampleInterval,StartTemperature,BottomTemperature,D.StartPressure,D.EndPressure,Size,CylinderWorkPressure,SurfacePressure,DiveTime,SampleInterval,ProfileBlob,TemperatureBlob,PressureBlob,Oxygen,Helium,MIX.StartPressure,MIX.EndPressure FROM Dive AS D JOIN DiveMixture AS MIX ON D.DiveId=MIX.DiveId";
retval = sqlite3_exec(handle, get_dives, &dm4_dive, handle, &err);
if (retval != SQLITE_OK) {
fprintf(stderr, translate("gettextFromC", "Database query failed '%s'.\n"), url);
return 1;
}
return 0;
}
extern int shearwater_cylinders(void *handle, int columns, char **data, char **column)
{
cylinder_start();
if (data[0])
cur_dive->cylinder[cur_cylinder_index].gasmix.o2.permille = atof(data[0]) * 1000;
if (data[1])
cur_dive->cylinder[cur_cylinder_index].gasmix.he.permille = atof(data[1]) * 1000;
cylinder_end();
return 0;
}
extern int shearwater_changes(void *handle, int columns, char **data, char **column)
{
event_start();
if (data[0])
cur_event.time.seconds = atoi(data[0]);
if (data[1]) {
cur_event.name = strdup("gaschange");
cur_event.value = atof(data[1]) * 100;
}
event_end();
return 0;
}
extern int shearwater_profile_sample(void *handle, int columns, char **data, char **column)
{
sample_start();
if (data[0])
cur_sample->time.seconds = atoi(data[0]);
if (data[1])
cur_sample->depth.mm = metric ? atof(data[1]) * 1000 : feet_to_mm(atof(data[1]));
if (data[2])
cur_sample->temperature.mkelvin = metric ? C_to_mkelvin(atof(data[2])) : F_to_mkelvin(atof(data[2]));
if (data[3])
cur_sample->po2.mbar = atof(data[3]) * 1000;
if (data[4])
cur_sample->ndl.seconds = atoi(data[4]) * 60;
if (data[5])
cur_sample->cns = atoi(data[5]);
if (data[6])
cur_sample->stopdepth.mm = metric ? atoi(data[6]) * 1000 : feet_to_mm(atoi(data[6]));
/* We don't actually have data[3], but it should appear in the
* SQL query at some point.
if (data[3])
cur_sample->cylinderpressure.mbar = metric ? atoi(data[3]) * 1000 : psi_to_mbar(atoi(data[3]));
*/
sample_end();
return 0;
}
extern int shearwater_dive(void *param, int columns, char **data, char **column)
{
int retval = 0;
sqlite3 *handle = (sqlite3 *)param;
char *err = NULL;
char get_profile_template[] = "select currentTime,currentDepth,waterTemp,averagePPO2,currentNdl,CNSPercent,decoCeiling from dive_log_records where diveLogId = %d";
char get_cylinder_template[] = "select fractionO2,fractionHe from dive_log_records where diveLogId = %d group by fractionO2,fractionHe";
char get_changes_template[] = "select a.currentTime,a.fractionO2,a.fractionHe from dive_log_records as a,dive_log_records as b where a.diveLogId = %d and b.diveLogId = %d and (a.id - 1) = b.id and (a.fractionO2 != b.fractionO2 or a.fractionHe != b.fractionHe) union select min(currentTime),fractionO2,fractionHe from dive_log_records";
char get_buffer[1024];
dive_start();
cur_dive->number = atoi(data[0]);
cur_dive->when = (time_t)(atol(data[1]));
if (data[2])
utf8_string(data[2], &cur_dive->location);
if (data[3])
utf8_string(data[3], &cur_dive->buddy);
if (data[4])
utf8_string(data[4], &cur_dive->notes);
metric = atoi(data[5]) == 1 ? 0 : 1;
/* TODO: verify that metric calculation is correct */
if (data[6])
cur_dive->dc.maxdepth.mm = metric ? atof(data[6]) * 1000 : feet_to_mm(atof(data[6]));
if (data[7])
cur_dive->dc.duration.seconds = atoi(data[7]) * 60;
if (data[8])
cur_dive->dc.surface_pressure.mbar = atoi(data[8]);
/*
* TODO: the deviceid hash should be calculated here.
*/
settings_start();
dc_settings_start();
if (data[9])
utf8_string(data[9], &cur_settings.dc.serial_nr);
if (data[10])
utf8_string(data[10], &cur_settings.dc.model);
cur_settings.dc.deviceid = 0xffffffff;
dc_settings_end();
settings_end();
snprintf(get_buffer, sizeof(get_buffer) - 1, get_cylinder_template, cur_dive->number);
retval = sqlite3_exec(handle, get_buffer, &shearwater_cylinders, 0, &err);
if (retval != SQLITE_OK) {
fprintf(stderr, "%s", translate("gettextFromC", "Database query get_cylinders failed.\n"));
return 1;
}
snprintf(get_buffer, sizeof(get_buffer) - 1, get_changes_template, cur_dive->number, cur_dive->number);
retval = sqlite3_exec(handle, get_buffer, &shearwater_changes, 0, &err);
if (retval != SQLITE_OK) {
fprintf(stderr, "%s", translate("gettextFromC", "Database query get_changes failed.\n"));
return 1;
}
snprintf(get_buffer, sizeof(get_buffer) - 1, get_profile_template, cur_dive->number);
retval = sqlite3_exec(handle, get_buffer, &shearwater_profile_sample, 0, &err);
if (retval != SQLITE_OK) {
fprintf(stderr, "%s", translate("gettextFromC", "Database query get_profile_sample failed.\n"));
return 1;
}
dive_end();
return SQLITE_OK;
}
int parse_shearwater_buffer(sqlite3 *handle, const char *url, const char *buffer, int size,
struct dive_table *table)
{
int retval;
char *err = NULL;
target_table = table;
char get_dives[] = "select i.diveId,timestamp,location||' / '||site,buddy,notes,imperialUnits,maxDepth,maxTime,startSurfacePressure,computerSerial,computerModel FROM dive_info AS i JOIN dive_logs AS l ON i.diveId=l.diveId";
retval = sqlite3_exec(handle, get_dives, &shearwater_dive, handle, &err);
if (retval != SQLITE_OK) {
fprintf(stderr, translate("gettextFromC", "Database query failed '%s'.\n"), url);
return 1;
}
return 0;
}
void parse_xml_init(void)
{
LIBXML_TEST_VERSION
}
void parse_xml_exit(void)
{
xmlCleanupParser();
}
static struct xslt_files {
const char *root;
const char *file;
const char *attribute;
} xslt_files[] = {
{ "SUUNTO", "SuuntoSDM.xslt", NULL },
{ "Dive", "SuuntoDM4.xslt", "xmlns" },
{ "Dive", "shearwater.xslt", "version" },
{ "JDiveLog", "jdivelog2subsurface.xslt", NULL },
{ "dives", "MacDive.xslt", NULL },
{ "DIVELOGSDATA", "divelogs.xslt", NULL },
{ "uddf", "uddf.xslt", NULL },
{ "UDDF", "uddf.xslt", NULL },
{ "profile", "udcf.xslt", NULL },
{ "Divinglog", "DivingLog.xslt", NULL },
{ "csv", "csv2xml.xslt", NULL },
{ "sensuscsv", "sensuscsv.xslt", NULL },
{ "manualcsv", "manualcsv2xml.xslt", NULL },
{ NULL, }
};
static xmlDoc *test_xslt_transforms(xmlDoc *doc, const char **params)
{
struct xslt_files *info = xslt_files;
xmlDoc *transformed;
xsltStylesheetPtr xslt = NULL;
xmlNode *root_element = xmlDocGetRootElement(doc);
char *attribute;
while (info->root) {
if ((strcasecmp(root_element->name, info->root) == 0)) {
if (info->attribute == NULL)
break;
else if (xmlGetProp(root_element, info->attribute) != NULL)
break;
}
info++;
}
if (info->root) {
attribute = xmlGetProp(xmlFirstElementChild(root_element), "name");
if (attribute) {
if (strcasecmp(attribute, "subsurface") == 0) {
free((void *)attribute);
return doc;
}
free((void *)attribute);
}
xmlSubstituteEntitiesDefault(1);
xslt = get_stylesheet(info->file);
if (xslt == NULL) {
report_error(translate("gettextFromC", "Can't open stylesheet %s"), info->file);
return doc;
}
transformed = xsltApplyStylesheet(xslt, doc, params);
xmlFreeDoc(doc);
xsltFreeStylesheet(xslt);
return transformed;
}
return doc;
}