So far Linus has hated all of my attempts to visualize vertical velocity
through color. This time I'm trying something dramatically new: there is
no PURPLE involved. Maybe that will convince him of the value.
We simply calculate the vertical velocity for the current plot segment
(last sample point to this sample point - in this version even without
divisions by zero) and assign a label based on the rate of change. These
labels are translated through a predefined table into colors:
Dark green is +/- 5ft/min (stable)
Light green is descents up to 30ft/min and ascents up to 15ft/min
Yellow is descents up to 60ft/min and ascents up to 30ft/min
Orange is descents up to 100ft/min and ascents up to 60ft/min
Red is outside of those ranges - you are most likely in danger
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Even though we go down to an 8pt font the info_frame changes size when the
air info is added. I don't like this but want to see how Linus would like
this resolved before going overboard.
Minor tweaks to the formating (we don't need two decimals when printing
the liters of air consumed).
This patch does NOT remove the plot of the air information in the profile
graph. I think we want to remove that once we like the text where it is,
but I wanted to do one thing at a time.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
If the temperature is in a very narrow range the existing code visually
exaggerated the fluctuations. This tries to dampen that effect a bit.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Change the duration max rounding as noted by Dirk, and move the air
consumption down further towards the bottom right corner. In
particular, I make the text positions not scale with the window size,
purely by the size of the text.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
- the time stamp where we printed the last temp was wrong
- we really shouldn't check mK for being identical - especially on dive
computers that store a lot of samples
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Ok, this is pretty much it now. Instead of having various random checks
for "is the time of the sample past the end of the dive" hacks, we not
plot all graphs from the cleaned-up plot_info structure instead of the
raw samples.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Further movement to using the sanitized and cleaned-up plot info rather
than the raw data.
The raw dive data contains samples from the end of the dive that we
don't want to drop, but that we also don't want to actually use for
plotting the dive. So the eventual end goal here is to not ever use the
raw dive samples directly for plotting, but use the diveplot data that
we have analyzed for min/max (properly ignoring final entries) etc.
There's still some data that we take from the samples when plotting, but
it's getting rarer.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Do the min/max calculations only *after* we have removed the extra
surface events at the end.
The Uemis data in particular has a lot of surface events after the dive,
and we don't really want to take them into account since we won't be
plotting them anyway.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
.. I'll want to move pressure limit calculations into the 'plot_info',
so that we can do several passes of analysis and change dive limits etc
without having to actually modify the dive data itself (or add new
fields to 'struct dive' just for plotting).
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Currently we print the temperature every five minutes. Especially with
dive computers that keep rather frequent temperature samples that means
that we have one more interesting data point that we don't label: the
surface temperature at the end of the dive.
This patch adds some logic to try to print the last temperature sample
that was recorded before the dive ended - unless that same value has
already been printed (to avoid silly duplications on dive computers with
less frequent sampling)
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Just like we end depth and tank pressure plots once we are on the surface
(this is relevant for dive computers like the uemis Zurich that keep
recording samples after the end of the dive)
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
I'll get there. Shrink it down a bit, start adding notes and location,
and maybe put three per page. That might work.
.. or maybe I should just take a look at how others have done this.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Ok, this is the ugliest f*&$ing printout I have ever seen in my life,
but think of it as a "the concept of printing works" commit, and you'll
be able to hold your lunch down and not gouge out your eyeballs with a
spoon. Maybe.
I'm just doing the cairo display as-is for the printout, which is a
seriously bad idea. I need to not try to do colors etc, and instead of
having white lines on a black background I just need to make thelines be
black on white paper.
But that would involve actually changing the current "plot()" routine,
which is against the point of the exercise right now. This really is
just a demonstration of how to add printing capabilities.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
It doesn't really make much of a difference, but it can be visible
especially with lots of tight samples. Miter joins really look horrible
for acute angles.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Oooh, pretty.
Or not. The temperature graph is usually ugly as hell, but Dirk has the
cool dive computer with lots and lots of temperature readings. Which
makes the graph a pretty graph, rather than a butt-ugly staircase like
mine.
Next time: get a dive computer with an OLED screen, and that can draw
pretty temperature graphs.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
.. without the actual text, because I'm a "random plots that cannot
actually be interpreted" kind of guy.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
I'm trying to make sure that we can shrink the main window and still get
a useful experience. Sometimes you have small bad netbooks when diving..
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Instead of relying on our ad-hoc minmax finder, just use the local
minima/maxima information directly.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This way we can always find the actual min/max entry that generated the
local minima/maxima. Which is useful for visualization.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Dirk likes purple. I mean - Dirk REALLY likes purple.
And what's better than "purple"? You got it: "funky purple".
So this shows the one- two- and three-minute min/max information in some
seriously funky purple fringing. It's not really necessarily meant to
be serious, but it's a quick hack to visualize the data until we figure
out what to *really* do with it.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This turns the depth profile into a generic "plot_info" and calculates
minima, maxima and averages over 1-, 2- and 3-minute intervals for each
point. It also creates a smoothed version.
We currently don't actually show the results, but that's the next step..
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
.. unless they are so shallow that they are basically at the surface.
These show up automatically in out min/max logic, so just go ahead and
show them.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
No change in semantics, I'm just contemplating doing some text renderign
from within the "minmax" function itself.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Ok, so it's really a 'double', but for now we're only using integer font
sizes, so let's see if we ever want to do anything but that.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Add new valign enum to text_render_options_t and update all callers to
plot_text
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
[ Fixed spelling, updated to newer base - Linus ]
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
I still think there should be some way to partition the space
automatically, but the algorithm that worked best was the simple
tail-recursive one.
Which might as well be expressed as a loop.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
We don't actually use the 'dive' structure any more, since we now always
have the sample pointers directly.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The way cairo does scaling is really really inconvenient, and one of the
things in cairo that is fundamentally mis-designed.
Cairo scaling always affects both coordinates and object sizes, and the
two can apparently never be split apart. Which is very much not what we
want: we want just coordinate scaling.
So we cannot use 'cairo_scale()' to scale our canvas, because that
screws up lines and text size too. And no, you cannot "fix" that by
de-scaling the line size etc - because line size is one-dimensional, so
you can't undo the (different) scaling in X/Y.
Sad. I realize that often you do want to scale object size with
coordinate transformation, but quite often you *don't* want to.
Yeah, we could do random context save/restore in odd places etc, but
that's just a sign of the bad design of cairo scaling.
Work around it by introducing our own graphics context with scaling,
which does it right. I don't like this, but it seems to be better than
the alternatives.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This is a bit more natural, and makes it much easier to do scale
independence. In particular, I want to make it possible to grow and
shrink the graph, and this should make it particularly simple to react
by giving more or fewer minmax points.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Use start/end sample pointers to make a recursive algorithm possible.
Also, clean up the end condition - we don't want to return an
uninteresting minmax result just because we ran out of samples.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This shows the depth properly in meter or feet depending on unit
selection.
It also changes the horizontal depth rulers to be at 10m/30ft intervals
rather than the previous 15ft. With the textual depth markers, the
horizontal lines aren't as important any more.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Ok, it's an odd place to start, but this now shows the pressure curve
details and the air usage in the proper units.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Soon we'll show things in psi or bar depending on user choice. Let's
not get confused about units before we do.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The UEMIS Zurich SDA keeps recording samples for quite a while after the
dive ended. These provide no additional information, but confuse our
drawing algorithm as they can cause us to draw both the depth and tank
pressure plots beyond the right edge of our canvas.
Stop drawing if sample->time.seconds is larger than dive->duration.seconds.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
* dirk:
Print starting and ending pressures
Fix up conflicts in profile.c due to different ways to set the text
formatting. Dirk's 'text_format_options' thing is prettier than mine.
Use it.
This is very simplistic as far as placement of the text goes.
It makes the plot_text function somewhat more generic.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Show "absolute volume" used, and SAC/m (surface-equivalent per minute).
I'm not going to guarantee the calculations. And I show the result in
cubic feet. Sue me.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Text notes need to be last, so that they don't get stepped on by the
other graph elements.
Also, separate the depth text plot out into a function of its own.
Tidier that way.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Use a 10-minute window *or* when the depth has reversed sufficiently to
make the max we've found interesting.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Add some actual numbers to the depth plot too. Do it by finding the
deepest points (within a five-minute rolling window), and show the
depths of those points.
Sure, we could have just labeled the depth markers, but this seems
nicer. But what do I know - I'm not exactly famous for my GUI design.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Only draw the pressure line to the final data point
(duration / end.mbar) if we haven't already drawn samples
past that point (as the UEMIS records pressure data for a
number of additional samples after the actual dive has ended)
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>
[ Changed to use 'last actual drawn sample time that had pressure
data' instead of 'last sample time' - Linus ]
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Doing per-dive cylinder start/end pressures is insane, when we can have
up to eight cylinders. The cylinder start/end pressure cannot be per
dive, it needs to be per cylinder.
This makes the save format cleaner too, we have all the cylinder data in
just one place.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Instead of just tracking gasmix, track the size and workng pressure of
the cylinder too.
And use "cylinder" instead of "tank" throughout.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Now the dive profile plot *really* needs some units. The pressure is
just a random line otherwise.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The time minimum was in seconds, not minutes, and we really do want to
show at least to 90ft to make shallow dives look shallow rather than
scaled to some full depth.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
.. and repaint the profile when the selection changes.
Now, if it just wasn't so ugly, it might even be useful. Except it
obviously needs to also show all the other dive information. And allow
the user to fill in details. And save the end results.
So no, it's not useful.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>