Re: [matplotlib-devel] Re: Matplotlib: wxWindows backend

SourceForge Headquarters 225 Broadway Suite 1600 San Diego, CA 92101 +1 (858) 422-6466

Hi John

On Thursday 06 November 2003 3:57 pm, John Hunter wrote:
>     Jeremy> My Debian machine is Python 2.3, and it seems to work
>     Jeremy> there. As I mentioned, the code seems (on Linux) to need
>     Jeremy> to run in a PyCrust shell at the moment - I need to fix
>     Jeremy> this, but I suspect that it means getting bogged down in
>     Jeremy> details of how the event loop works, which I'd rather put
>     Jeremy> off until the main backend is stable.
>
> I think this is not too hard.  I created a
>
>   wxapp = wxPySimpleApp()
>
> at the top of backend_wx (to solve the bitmap problem).  Then in
> ShowOn.realize_windows I called wxapp.MainLoop().  With these changes
> I can call
>
>   > python simple_plot.py -dWX

Thanks for fixing this. To be honest, my main focus is using matplotlib inside 
an applications, which is why I hadn't looked into this, beyond ensuring that 
it works in a PyCrust (which is my usual interactive shell, to be honest)

> Thanks for the CVS info.  You'll need to get a sourceforge account if
> you don't already have one, and send me your user name.  Then I'll add
> you as a developer so you can have write access to the repository.

My Sourceforge user name is jodonoghue

> I'll do some reading up on CVS.  What happens if we both check out a
> copy of the wx backend and make changes to different functions in the
> same file, and then both check the code back in?  Will CVS
> automagically update the separate parts of the file, or will the
> second person get an error saying "You can't check this in because the
> CVS file has changed since you checked out", or what?

CVS, like most SCM tools, doesn't help you quite this much. All it really does 
is to keep a diff every time a new file is committed, so it is quite possible 
that if you make a change to a file, and then I check in a new change without 
an update, HEAD would 'loose' your update (although it would be easy using a 
CS diff to determine the 'lost' code.

Normally CVS is configured so that files are checked out 'unlocked' - i.e. 
more than one developer can check out a file (so a check out is 'advisory' 
only).

Here's how I work in CVS:

I have a local mirror of CVS tree, which I regularly update to head. From this 
I copy files to my matplotlib development directory. When I wish to check in 
an update, I do an update on my CVS copy, and diff the new file with the CVS 
copy. There are two possibilities:

1) Any changes to HEAD are on lines I haven't changed. In this case
   (which is the most usual) I just have to do run merge between the
   files and then commit (having regression tested, of course).
2) We have made incompatible changes (e.g. both fixed the same bug
   in slightly different ways, or added new code in the same place). In this
   case I have to do a manual merge on the files, taking parts from both
   and working until regression tests pass - then commit.

One way to avoid this kind of issue is to (informally, CVS doesn't let you 
enforce this) say that only one person will normally work on a given file - 
in practice, this is how we operate at work (but then we have a 10,000 file 
source tree, so everyone would go crazy if we didn't). If someone has a fix 
for a file, they send it to the file maintainer, and ask him/her to 
incorporate the patch. This is basically how Linus maintains Linux (there are 
only a very small number of people who have commit rights to the Linux CVS 
tree - maybe 10 or 20 of the thousands of developers.

> I made a couple of small changes to your code (BTW, I don't know if
> matplotlib-devel is working because none of my emails have made it
> through).  The images dir is now in the matplotlib root (where
> setup.py and the fonts dir reside).  The distutils installer puts them
> in the matplotlib shared dir for data files, and I changed
> _load_bitmap to look for them there.  This maintains consistency with
> how I deal with other data files, eg, font files.

All of your changes seem eminently sensible. You have done an excellent job of 
architecting matplotlib, and I'd rather let you keep things designed as you 
prefer them. We can (and should) discuss any architectural changes, but I'll 
implement whatever you finally decide.

In fact, it makes sense to say that backend_gtk should be the test bed for any 
architectural changes - when you're reasonably satisfied that things work, 
I'll implement equivalent changes in backend_wx. This will always put 
backend_wx slightly behind backend_gtk, but we can always ensure that they're 
back in sync for major releases.

By the way, based on what I've found so far, the main areas I've found which 
caused me some confusion in implementing the backend are:

- Think there should be a 'default' set of fonts with names common on each 
platform (I've implemented this)
- Should also be a set of common names for dotted lines - there are only four 
types, so no problem, I should think.
- Somehow AxisText seems to be disconnected from the way the graphics are 
done. I managed to confuse myself quite a lot here. It seems to me that 
renderer is ultimately responsible fro rendering both text and graphics (on 
the same co-ordinate system)
- I was also confused by the how scaling and co-ordinates are mapped to the 
device - this is probably a matter of documentation rather than design 
change.

Regards

Jeremy