The only time the TankItem is replot is when new data is set.
Therefore, replot() can be folded into setData().
The good thing is that setData() is passed the dive to be
plot. So the data can be extracted from there instead of
the global displayed_dive variable.
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
There is no point in a separate set-axis function if we never
change the axis anyway. Make the axis a const-reference to
show that it can never be changed.
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
Since the initial commit introducing TankItem, there was
a connection that replotted the item if the horizontal axis
sent the sizeChanged() signal. I never managed to create
this signal for the horizontal axis, only for the vertical
axes. Therefore remove this thing. If it turns out that
we need it after all, readd it in a more deterministic
way (i.e. call where it is needed).
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
This one is extremely obscure: TankItem::setData(), which is
called on every replot, was passed the DivePlotDataModel,
even though it doesn't access that model at all.
Instead, it connect()s to the model to stay informed of changes
to the data. First of all, this should obviously be done
once in the constructor, not on every replot.
But also, the setData() function is called on every replot
one lines before sending the model-changed signal.
Thus, the tankitem was always repainted twice.
Just remove the whole connect() thing and go for a more
deterministic model. Should the tankbar not be repainted
anywhere, add the appropriate calls there.
Accordingly rename the "modelDataChanged" slot to "replot".
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
In an effort to reduce the size of dive.h and dive.c, break out
the event related functions. Moreover event-names were handled
by the profile-code, collect that also in the new source files.
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
Instead of accessing the cylinder table directly, use the get_cylinder()
function. This gives less unwieldy expressions. But more importantly,
the function does bound checking. This is crucial for now as the code
hasn't be properly audited since the change to arbitrarily sized
cylinder tables. Accesses of invalid cylinder indexes may lead to
silent data-corruption that is sometimes not even noticed by
valgrind. Returning NULL instead of an invalid pointer will make
debugging much easier.
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
TankItem would happily access a non-existing cylinder and crash.
But freedives for example have no cylinders. Thus, handle that
situation gracefully by exiting early if there is no cylinder.
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
Instead of using fixed size arrays, use a new cylinder_table structure.
The code copies the weightsystem code, but is significantly more complex
because cylinders are such an integral part of the core.
Two functions to access the cylinders were added:
get_cylinder() and get_or_create_cylinder()
The former does a simple array access and supposes that the cylinder
exists. The latter is used by the parser(s) and if a cylinder with
the given id does not exist, cylinders up to that id are generated.
One point will make C programmers cringe: the cylinder structure is
passed by value. This is due to the way the table-macros work. A
refactoring of the table macros is planned. It has to be noted that
the size of a cylinder_t is 64 bytes, i.e. 8 long words on a 64-bit
architecture, so passing on the stack is probably not even significantly
slower than passing as reference.
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
Some widgets copy the full plot info. Free these data on exit to
prevent monstrous valgrind reports.
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
TankItem had a "height" member variable that was never modified.
Turn it into a constant, which is local to the translation unit.
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
The whole plot info data was copied only so that the time of the
last item could be determined later. Instead, simply store the
timestamp.
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
Calculation of the x-position and the width of the tank-bar
was done outside of the function. Move it into the function
to make the caller a bit more readable.
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
This is only used for an "is initialized"-check. But there are
other member variables that are used for that purpose. Remove.
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
It expands to nullptr anyway and is inconsitent with the rest of the
code. Let's remove this anachronism.
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
This is another entry in the series to make more things
"const-clean" with the ultimate goal of merge_dive() take
const pointers.
This concerns functions taking pointers to events and
the fallout from making these const.
The somewhat debatable part of this commit might be
that get_next_event() is split in a two distinct
(const and non-const) versions with different names,
since C doesn't allow overloading. The linker should
recognize that these functions are identical and remove
one of them.
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
In a previous commit, the get_gasmix_* functions were changed to
return by value. For consistency, also pass gasmix by value.
Note that on common 64-bit platforms struct gasmix is the size
of a pointer [2 * 32 bit vs. 64 bit] and therefore uses the
same space on the stack. On 32-bit platforms, the stack use
is probably doubled, but in return a dereference is avoided.
Supporting arbitrary gas-mixes (H2, Ar, ...) will be such an
invasive change that going back to pointers is probably the
least of our worries.
This commit is a step in const-ifying input parameters (passing
by value is the ultimate way of signaling that the input parameter
will not be changed [unless there are references to said parameter]).
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
Currently, get_gasmix_from_event() and get_gasmix() return pointers
to either static or to (possibly changing) dive data. This seems like
a dangerous practice and the returned data should be used immediately.
Instead, return the gasmix by value. This is in preparation of
const-ifying input parameters of a number of core functions, which
will ultimately let the merge() function take const-arguments in
preparation of undo of dive-merging.
On common 64-bit systems gasmix (two "int"s) is the size of a pointer
and can be returned in a register.
On 32-bit systems a pointer to the struct to be filled out will be
passed.
Since get_gasmix() now returns a value, the first invocation is
tested by a NULL-initialized "struct event *". Document this in
a comment.
Signed-off-by: Berthold Stoeger <bstoeger@mail.tuwien.ac.at>
CID 45172
And one non initalized member hAxis could be removed as it is
not used. Surprisingly, Coverity did not see this.
Signed-off-by: Jan Mulder <jlmulder@xs4all.nl>
Go back to the old startegy of retrieving the correct end of the dive
plot by looking at the plot data instead of looking at dc->duration.
Signed-off-by: Stefan Fuchs <sfuchs@gmx.de>
When Linus modified the gas handling code six weeks ago he pointed out
that that had broken the tankbar; with this patch we now simply walk the
gas changes of the displayed dive directly and create the tankbar
rectangles from that information.
See #562
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
This finally handles multiple cylinder pressures, both overlapping and
consecutive, and it seems to work on the nasty cases I've thrown at it.
Want to just track five different cylinders all at once, without any
pesky gas switch events? Sure, you can do that. It will show five
different gas pressures for your five cylinders, and they will go down
as you breathe down the cylinders.
I obviously don't have any real data for that case, but I do have a test
file with five actual cylinders that all have samples over the whole
course of the dive. The end result looks messy as hell, but what did
you expect?
HOWEVER.
The only way to do this sanely was
- actually make the "struct plot_info" have all the cylinder pressures
(so no "sensor index and pressure" - every cylinder has a pressure for
every plot info entry)
This obviously makes the plot_info much bigger. We used to have
MAX_CYLINDERS be a fairly generous 8, which seems sane. The planning
code made that 8 be 20. That seems questionable. But whatever.
The good news is that the plot-info should hopefully get freed, and
only be allocated one dive at a time, so the fact that it is big and
nasty shouldn't be a scaling issue, though.
- the "populate_pressure_information()" function had to be rewritten
quite a bit. The good news is that it's actually simpler now, although
I would not go so far as to really call it simple. It's still
complicated and suble, but now it explicitly just does one cylinder at
a time.
It *used* to have this insanely complicated "keep track of the pressure
ranges for every cylinder at once". I just couldn't stand that model
and keep my sanity, so it now just tracks one cylinder at a time, and
doesn't have an array of live data, instead the caller will just call
it for each cylinder.
- get rid of some of our hackier stuff, like the code that populates the
plot_info data code with the currently selected cylinder number, and
clears out any other pressures. That obviously does *not* work when you
may not have a single primary cylinder any more.
Now, the above sounds like all good things. Yeah, it mostly is.
BUT.
There's a few big downsides from the above:
- there's no sane way to do this as a series of small changes.
The change to make the plot_info take an array of cylinder pressures
rather than the sensor+pressure model really isn't amenable to "fix up
one use at a time". When you switch over to the new data structure
model, you have to switch over to the new way of populating the
pressure ranges. The two just go hand in hand.
- Some of our code *depended* on the "sensor+pressure" model. I fixed all
the ones I could sanely fix. There was one particular case that I just
couldn't sanely fix, and I didn't care enough about it to do something
insane.
So the only _known_ breakage is the "TankItem" profile widget. That's
the bar at the bottom of the profile that shows which cylinder is in
use right now. You'd think that would be trivial to fix up, and yes it
would be - I could just use the regular model of
firstcyl = explicit_first_cylinder(dive, dc)
.. then iterate over the gas change events to see the others ..
but the problem with the "TankItem" widget is that it does its own
model, and it has thrown away the dive and the dive computer
information. It just doesn't even know. It only knows what cylinders
there are, and the plot_info. And it just used to look at the sensor
number in the plot_info, and be done with that. That number no longer
exists.
- I have tested it, and I think the code is better, but hey, it's a
fairly large patch to some of the more complex code in our code base.
That "interpolate missing pressure fields" code really isn't pretty. It
may be prettier, but..
Anyway, without further ado, here's the patch. No sign-off yet, because I
do think people should look and comment. But I think the patch is fine,
and I'll fix anythign that anybody can find, *except* for that TankItem
thing that I will refuse to touch. That class is ugly. It needs to have
access to the actual dive.
Note how it actually does remove more lines than it adds, and that's
despite added comments etc. The code really is simpler, but there may be
cases in there that need more work.
Known missing pieces that don't currently take advantage of concurrent
cylinder pressure data:
- the momentary SAC rate coloring for dives will need more work
- dive merging (but we expect to generally normally not merge dive
computers, which is the main source of sensor data)
- actually taking advantage of different sensor data from different
dive computers
But most of all: Testing. Lots and lots of testing to find all the
corner cases.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
This is a very timid start at making us actually use multiple sensors
without the magical special case for just CCR oxygen tracking.
It mainly does:
- turn the "sample->sensor" index into an array of two indexes, to
match the pressures themselves.
- get rid of dive->{oxygen_cylinder_index,diluent_cylinder_index},
since a CCR dive should now simply set the sample->sensor[] indices
correctly instead.
- in a couple of places, start actually looping over the sensors rather
than special-case the O2 case (although often the small "loops" are
just unrolled, since it's just two cases.
but in many cases we still end up only covering the zero sensor case,
because the CCR O2 sensor code coverage was fairly limited.
It's entirely possible (even likely) that this migth break some existing
case: it tries to be a fairly direct ("stupid") translation of the old
code, but unlike the preparatory patch this does actually does change
some semantics.
For example, right now the git loader code assumes that if the git save
data contains a o2pressure entry, it just hardcodes the O2 sensor index
to 1.
In fact, one issue is going to simply be that our file formats do not
have that multiple sensor format, but instead had very clearly encoded
things as being the CCR O2 pressure sensor.
But this is hopefully close to usable, and I will need feedback (and
maybe test cases) from people who have existing CCR dives with pressure
data.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Add the tankbar to the profile and change the relative positions of the depth
and temperature curves to minimize overlap.
Signed-off-by: Joakim Bygdell <j.bygdell@gmail.com>
When painting the tankbar the function triggers on change in cylinder index,
as a result the new gascolour are changed at the next sample time point.
On a divecomputer with a reasonable fast sample rate the 2-3s offset are hardly
noticable, especially on a longer dive.
For divecomputers with slow sample rate the 10-30s offset are clearly visible.
This is fixed by start painting the new gascolour at the time point of the switch event rather than the time point of the next sample.
Signed-off-by: Joakim Bygdell <j.bygdell@gmail.com>
Having subsurface-core as a directory name really messes with
autocomplete and is obviously redundant. Simmilarly, qt-mobile caused an
autocomplete conflict and also was inconsistent with the desktop-widget
name for the directory containing the "other" UI.
And while cleaning up the resulting change in the path name for include
files, I decided to clean up those even more to make them consistent
overall.
This could have been handled in more commits, but since this requires a
make clean before the build, it seemed more sensible to do it all in one.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
The reason for that is, even if profile widget is made with qpainter
and for that reason it should be a desktop widget, it's being used
on the mobile version because of a lack of QML plotting library that
is fast and reliable.
We discovered that it was faster just to encapsulate our Profile in
a QML class and call it directly.
Signed-off-by: Tomaz Canabrava <tomaz.canabrava@intel.com>
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
2015-10-30 10:36:49 -07:00
Renamed from desktop-widgets/profile/tankitem.cpp (Browse further)
