Thread: [Matplotlib-users] Crash in CVS

Hi there,

I've just started playing with matplot and am VERY impressed.
However, I thought I'd mention two items in case no one else has seen 
them.

1) There are a lot of Deprecation warnings on loading.
2) I can generate an X Window Error via a gdk crash. Here's the tail of 
the trace:
[------
  File "/usr/lib/python2.3/site-packages/matplotlib/text.py", line 310, 
in _set_font
    self._layout  = self._drawingArea.create_pango_layout(self._text)
TypeError: GtkWidget.create_pango_layout() argument 1 must be string, 
not int
The program 'stats.py' received an X Window System error.
This probably reflects a bug in the program.
The error was 'BadMatch (invalid parameter attributes)'.
  (Details: serial 4152 error_code 8 request_code 73 minor_code 0)
  (Note to programmers: normally, X errors are reported asynchronously;
   that is, you will receive the error a while after causing it.
   To debug your program, run it with the --sync command line
   option to change this behavior. You can then get a meaningful
   backtrace from your debugger if you break on the gdk_x_error() 
function.)

-----]

And here's the code. If you run it as is, it works fine (though the plot 
is ugly.) However, if you uncomment the "set(gca(),...", it gives the 
error.

I've started using matplot in a script to auto-generate a report.

-Charles
-- 
Charles R. Twardy               Monash University, School of CSSE
ctwardy at alumni indiana edu   +61(3) 9905 5823 (w)  5146 (fax)
                             ~^~
  "eloquence ought to be banish'd out of all civil Societies as a
   thing fatal to Peace and good Manners..."          ~Sprat 1667

>>>>> "Charles" == Charles Twardy <ct...@ma...> writes:

    Charles> Hi there, I've just started playing with matplot and am
    Charles> VERY impressed.  

Thanks.

    Charles>However, I thought I'd mention two items
    Charles> in case no one else has seen them.

    Charles> 1) There are a lot of Deprecation warnings on loading.

Thanks for the alert.  Are you using python2.3?  I've heard that 2.3
generates deprecation warnings and I plan to get them cleaned up in
the near future and do a bug-fix release, probably 0.21.

    Charles> 2) I can generate an X Window Error via a gdk
    Charles> crash. Here's the tail of the trace: [------ File
    Charles> "/usr/lib/python2.3/site-packages/matplotlib/text.py",
    Charles> line 310, in _set_font self._layout =
    Charles> self._drawingArea.create_pango_layout(self._text)
    Charles> TypeError: GtkWidget.create_pango_layout() argument 1
    Charles> must be string, not int The program 'stats.py' received

Looks like you passed an integer to a text label command in the
set(gca..., command that is causing your troubles.  If you are trying
to set a ticklabel, note that they must be strings, not ints.  Eg,

  set(gca(), 'xticklabels', ['%d' % val for val in arange(8)])

not

  set(gca(), 'xticklabels', arange(8))


Is this the source of your troubles (The example code you meant to
send didn't come through)?  Check it out.  If not, send me the script
that generates the error and I'll take a look.  I haven't seen this
bug before (and I use the library a lot <wink>).

John Hunter

Yes, I've just started using python2.3, which is almost certainly the 
source of the deprecation warnings.

And yes, I had passed an integer to the text label command. The crash 
seems a bit harsh, but at least I can avoid that now.

I generated a graph with 8 subplots (histograms). I wasn't able to use 
the title command on the subplots, because the title appeared on top of 
the plot above it. And I haven't yet figured out how to give the plot an 
overall title. I'm happy for hints if you have the time. Otherwise I'm 
fairly happy with what I've got (titles inside each subplot, via the 
text command, and no overall title).
 
Thanks! This is great!
-Charles
-- 
Charles R. Twardy               Monash University, School of CSSE
ctwardy at alumni indiana edu   +61(3) 9905 5823 (w)  5146 (fax)
                             ~^~
  "eloquence ought to be banish'd out of all civil Societies as a
   thing fatal to Peace and good Manners..."          ~Sprat 1667

>>>>> "Charles" == Charles Twardy <ct...@ma...> writes:

    Charles> Yes, I've just started using python2.3, which is almost
    Charles> certainly the source of the deprecation warnings.

    Charles> And yes, I had passed an integer to the text label
    Charles> command. The crash seems a bit harsh, but at least I can
    Charles> avoid that now.

I've added an automatic string conversion in the set ticklabels
functionality, so that any object that can be converted to a string
with %s will work as an argument to the ticklabels command.

    Charles> I generated a graph with 8 subplots (histograms). I
    Charles> wasn't able to use the title command on the subplots,
    Charles> because the title appeared on top of the plot above
    Charles> it. And I haven't yet figured out how to give the plot an
    Charles> overall title. I'm happy for hints if you have the
    Charles> time. Otherwise I'm fairly happy with what I've got
    Charles> (titles inside each subplot, via the text command, and no
    Charles> overall title).

I've needed this functionality too.  I just followed matlab's design
which is to add the title to the current axes.  In CVS, I've added a
new figure text function to the class API for Figure.  The figure text
is in relative 0-1 coords (0,0 is lower left; 1,1 is upper right).
You can add a title with

  gcf().text(0.5, 0.95, 'Figure Title', fontsize=12)
 
You can also add text anywhere else you want in figure coordinates
with this command (axis text is in data coordinates).  In the examples
dir in CVS there is a file figtext.py that illustrates this.  Perhaps
I'll add a new function figtitle to the matlab functional interface
since this comes up fairly often.

    Charles> Thanks! This is great!

Your welcome!

John Hunter

Hi John,

The latest CVS changes seem to help! The new deprecations are gone, and 
my own title works on center justify.

But if I try an absurdly long title:
gcf().text(0.5, 0.95,'Distance Histograms by Category is a really long title that should flow off', font, fontsize=12)

Then I get an X Window Error and my script crashes. Most ungraceful.

For fun, I've attached the .pngs I've been working on. Lost-person 
behavior: distance travelled and outcome of the search, broken down by 
major categories. (Australian data)

-Charles

-- 
Charles R. Twardy               Monash University, School of CSSE
ctwardy at alumni indiana edu   +61(3) 9905 5823 (w)  5146 (fax)
                             ~^~
  "eloquence ought to be banish'd out of all civil Societies as a
   thing fatal to Peace and good Manners..."          ~Sprat 1667

>>>>> "Charles" == Charles Twardy <ct...@ma...> writes:

    Charles> Hi John, The latest CVS changes seem to help! The new
    Charles> deprecations are gone, and my own title works on center
    Charles> justify.

    Charles> But if I try an absurdly long title: gcf().text(0.5,
    Charles> 0.95,'Distance Histograms by Category is a really long
    Charles> title that should flow off', font, fontsize=12)

    Charles> Then I get an X Window Error and my script crashes. Most
    Charles> ungraceful.

Yes, most.  Are you sure you are using the right version?  Does line
1702 of figure.py read

        def clip_gc(gc):
            gc.set_clip_rectangle( (0, 0, self.width, self.height) )
        for t in self._text:
            t.clip_gc = clip_gc

I also tested with a ridiculously long figure title and did not get
the X windows crash after making these changes .  CVS should have
this, but as you've noted, sometimes the mirrors are a little behind.
If you have this and are still getting the crashes, let me know and
I'll sort it out.

    Charles> For fun, I've attached the .pngs I've been working
    Charles> on. Lost-person behavior: distance travelled and outcome
    Charles> of the search, broken down by major
    Charles> categories. (Australian data)

That's great.  Unfortunately I can't look at them right now because
I'm logged in remotely from a win32 machine but will check them out
Monday.  If you'd like to send a script, I can add one of them to the
screenshots section of the home page.

JDH

It seems I'm running the new version:
  * line 1702 of figure.py reads as you asked
  * I just tried another CVS get and got nothing
  * I just did another run of setup.py and nothing happened

I still get the crash, and can reproduce with a simple 3-line script,
attached.

}Monday.  If you'd like to send a script, I can add one of them to the
}screenshots section of the home page.
My script is an ugly beast that parses an ugly dataset. But I'll see about
getting a demo version.
	-C

--
Charles R. Twardy, Res.Fellow,  Monash University, School of CSSE
ctwardy at alumni indiana edu   +61(3) 9905 5823 (w)  5146 (fax)

>>>>> "Charles" == Charles R Twardy <ct...@ma...> writes:

    Charles> It seems I'm running the new version: * line 1702 of
    Charles> figure.py reads as you asked * I just tried another CVS
    Charles> get and got nothing * I just did another run of setup.py
    Charles> and nothing happened

    Charles> I still get the crash, and can reproduce with a simple
    Charles> 3-line script, attached.

    Charles> }Monday.  If you'd like to send a script, I can add one
    Charles> of them to the }screenshots section of the home page.  My
    Charles> script is an ugly beast that parses an ugly dataset. But
    Charles> I'll see about getting a demo version.  -C

Thanks for the update.  I'm knee deep in getting a postscript backend
working so I'll take a look at this in the next few days.  This is
proving to be very helpful because I've abstracted all the drawing
operations away from gtk which means it will be relatively easy to
port the lib to new output drivers or GUI toolkits.

As for the bug, I can replicate the bug on my system with your example
script, so it shouldn't be too hard to find and fix.  I'll let you
know when I get an updated CVS.

JDH