Do you use python setup.py build_ext --inplace ?

"Nicolas Grilly" <nicolas.grilly@...475...> writes:

When you are working on matplotlib, after a checkout from the SVN
repository, which method do you use to compile extensions without
re-building and re-installing everything each time?

I use "python setupegg.py develop", which I think does the same thing
as build_ext --inplace. I have made a symbolic link for the mpl-data
subdirectory. (Perhaps some setuptools wizard can tell us how to fix
setupegg.py so that mpl-data gets installed in-place?)

···

--
Jouni K. Sepp�nen
http://www.iki.fi/jks

Jouni K. Sepp�nen wrote:

"Nicolas Grilly" <nicolas.grilly@...475...> writes:

When you are working on matplotlib, after a checkout from the SVN
repository, which method do you use to compile extensions without
re-building and re-installing everything each time?
    
I use "python setupegg.py develop", which I think does the same thing
as build_ext --inplace. I have made a symbolic link for the mpl-data
subdirectory. (Perhaps some setuptools wizard can tell us how to fix
setupegg.py so that mpl-data gets installed in-place?)

IMO, the way to fix it is to move all the data into
lib/matplotlib/mpl-data to start with. If we did that, all the shuffling
that setup.py does would be unnecessary and using setuptools' develop
mode would work out-of-the-box. I'd be happy to move these files in the
svn repository if people agree this is a good idea.

-Andrew

I'm +1 on that. For some reason, I thought we had already decided to
do that when we moved to support setuptool in the first place.

JDH

···

On 2/12/07, Andrew Straw <strawman@...36...> wrote:

IMO, the way to fix it is to move all the data into
lib/matplotlib/mpl-data to start with. If we did that, all the shuffling
that setup.py does would be unnecessary and using setuptools' develop
mode would work out-of-the-box. I'd be happy to move these files in the
svn repository if people agree this is a good idea.

You're changing positions on me now, John! :wink:

I asked about moving everything into the module, but you didn't want
to mix the fonts/images/etc, so I made the install merge the data into
a folder. I would love to be able to run "setup develop" as well.

- Charlie

···

On 2/12/07, John Hunter <jdh2358@...149...> wrote:

On 2/12/07, Andrew Straw <strawman@...36...> wrote:
> IMO, the way to fix it is to move all the data into
> lib/matplotlib/mpl-data to start with. If we did that, all the shuffling
> that setup.py does would be unnecessary and using setuptools' develop
> mode would work out-of-the-box. I'd be happy to move these files in the
> svn repository if people agree this is a good idea.

I'm +1 on that. For some reason, I thought we had already decided to
do that when we moved to support setuptool in the first place.

You're changing positions on me now, John! :wink:

It's better that way -- that way I can be right both times :slight_smile: Or is
it wrong both times? hmmm...

I asked about moving everything into the module, but you didn't want
to mix the fonts/images/etc, so I made the install merge the data into
a folder. I would love to be able to run "setup develop" as well.

OK, I see, so the problem was we had to flatten everything and mix all
these datatypes together? eggs can't handle the existing directory
structure in the mpl-data? It would clearly be better if it could,
because dumping all the data files together isn't terribly elegant, so
I think it would be useful for some egghead to verify this and see if
there isn't a way to have the best of all worlds. I haven't used eggs
so can't say anything intelligent, except that I'm happy to defer to
those who know more about it and let's get it setup that that work
properly with mpl-data for eggs and and regular old-fashioned mpl
installs. If this means that we do have to flatten the data directory
structure, so be it. That's why god gave us 'ls *.EXT '

JDH

···

On 2/12/07, Charlie Moad <cwmoad@...149...> wrote:

Me too, I'm +1 on that. This way, in development mode, we can run
matplotlib just by adding directory "lib" to the PYTHONPATH, even if
using standard distutils without setuptools.

NG

···

On 2/12/07, Andrew Straw <strawman@...36...> wrote:

IMO, the way to fix it is to move all the data into
lib/matplotlib/mpl-data to start with. If we did that, all the shuffling
that setup.py does would be unnecessary and using setuptools' develop
mode would work out-of-the-box. I'd be happy to move these files in the
svn repository if people agree this is a good idea.

John Hunter wrote:

You're changing positions on me now, John! :wink:

It's better that way -- that way I can be right both times :slight_smile: Or is
it wrong both times? hmmm...

I asked about moving everything into the module, but you didn't want
to mix the fonts/images/etc, so I made the install merge the data into
a folder. I would love to be able to run "setup develop" as well.

OK, I see, so the problem was we had to flatten everything and mix all
these datatypes together? eggs can't handle the existing directory
structure in the mpl-data? It would clearly be better if it could,
because dumping all the data files together isn't terribly elegant, so
I think it would be useful for some egghead to verify this and see if
there isn't a way to have the best of all worlds. I haven't used eggs
so can't say anything intelligent, except that I'm happy to defer to
those who know more about it and let's get it setup that that work
properly with mpl-data for eggs and and regular old-fashioned mpl
installs. If this means that we do have to flatten the data directory
structure, so be it. That's why god gave us 'ls *.EXT '

Well, another option is to maintain a sub-directory structure when creating mpl-data in the source package. The downside is that this will introduce code changes all over the code -- for example, code that previously asked for "os.path.join( mpl.get_data_dir(), 'blah.glade' )" would have to be converted to "os.path.join( mpl.get_data_dir(), 'gui', 'blah.glade' )".

So, the low energy, low probability-of-breakage approach is indeed to pile everything into mpl-data and make judicious use of "ls *.ext". This is what we install in site-packages anyway, so it can't be that bad, right? :slight_smile:

-Andrew

···

On 2/12/07, Charlie Moad <cwmoad@...149...> wrote:

I think this should work even with distutils alone, without setuptools.

In order to not flatten everything, one solution is to simply create
subfolders in mpl-data.

We could have those folders:
/lib/matplotlib/mpl-data/fonts/afm
/lib/matplotlib/mpl-data/fonts/ttf
/lib/matplotlib/mpl-data/images

-- Nicolas

···

On 2/12/07, John Hunter <jdh2358@...149...> wrote:

OK, I see, so the problem was we had to flatten everything and mix all
these datatypes together? eggs can't handle the existing directory
structure in the mpl-data? It would clearly be better if it could,
because dumping all the data files together isn't terribly elegant, so
I think it would be useful for some egghead to verify this and see if
there isn't a way to have the best of all worlds. I haven't used eggs
so can't say anything intelligent, except that I'm happy to defer to
those who know more about it and let's get it setup that that work
properly with mpl-data for eggs and and regular old-fashioned mpl
installs. If this means that we do have to flatten the data directory
structure, so be it. That's why god gave us 'ls *.EXT '

Well, another option is to maintain a sub-directory structure when
creating mpl-data in the source package. The downside is that this will
introduce code changes all over the code -- for example, code that
previously asked for "os.path.join( mpl.get_data_dir(), 'blah.glade' )"
would have to be converted to "os.path.join( mpl.get_data_dir(), 'gui',
'blah.glade' )".

I think we should maintain a clean tree and make the code changes. I
bet it would only take one person a diligent hour or two to fix all
the occurrences as mentioned above. Maybe we could split up the work?
I would be willing to update the "backends" folder.

- Charlie

Charlie Moad wrote:

Well, another option is to maintain a sub-directory structure when
creating mpl-data in the source package. The downside is that this will
introduce code changes all over the code -- for example, code that
previously asked for "os.path.join( mpl.get_data_dir(), 'blah.glade' )"
would have to be converted to "os.path.join( mpl.get_data_dir(), 'gui',
'blah.glade' )".

I think we should maintain a clean tree and make the code changes. I
bet it would only take one person a diligent hour or two to fix all
the occurrences as mentioned above. Maybe we could split up the work?
I would be willing to update the "backends" folder.

Although my understanding of setup* is minimal, I agree; I think that keeping some organization in the data will be helpful. It looks like get_data_path() is not called in many places, so if that is essentially what has to be fixed, it should not be very difficult. It might be facilitated by making _get_data_path() accept *args, a set of subdirectories to be appended.

Eric

Although my understanding of setup* is minimal, I agree; I think that
keeping some organization in the data will be helpful. It looks like
get_data_path() is not called in many places, so if that is essentially
what has to be fixed, it should not be very difficult. It might be
facilitated by making _get_data_path() accept *args, a set of
subdirectories to be appended.

I think the way to go is what Andrew typed:

os.path.join( mpl.get_data_dir(), 'gui', 'blah.glade' )

Charlie Moad wrote:

Although my understanding of setup* is minimal, I agree; I think that
keeping some organization in the data will be helpful. It looks like
get_data_path() is not called in many places, so if that is essentially
what has to be fixed, it should not be very difficult. It might be
facilitated by making _get_data_path() accept *args, a set of
subdirectories to be appended.
    
I think the way to go is what Andrew typed:

os.path.join( mpl.get_data_dir(), 'gui', 'blah.glade' )
  

OK, I just implemented this nested idea and it appears to work.

I tried to commit it to SVN, but SF's servers were giving me some talk about permissions... Sigh.

So, I sent the patch file to John, who'll hopefully commit it if he approves. To paraphrase Robert Kern, "Working code ends all arguments". So, since this was the nested way, and the biggest reason against it was that it require code change, well, I changed the code, and perhaps the arguments are ended?

Because I did change a fair amount of data_path -using code, it'd be good to exercise all your favorite font and image finding sections of code. The bits I tried worked for me, of course, but no guarantees...

Thanks for tackling this.

···

On 2/12/07, Andrew Straw <strawman@...36...> wrote:

Charlie Moad wrote:
>> Although my understanding of setup* is minimal, I agree; I think that
>> keeping some organization in the data will be helpful. It looks like
>> get_data_path() is not called in many places, so if that is essentially
>> what has to be fixed, it should not be very difficult. It might be
>> facilitated by making _get_data_path() accept *args, a set of
>> subdirectories to be appended.
>>
>
> I think the way to go is what Andrew typed:
>
> os.path.join( mpl.get_data_dir(), 'gui', 'blah.glade' )
>
OK, I just implemented this nested idea and it appears to work.

I tried to commit it to SVN, but SF's servers were giving me some talk
about permissions... Sigh.

So, I sent the patch file to John, who'll hopefully commit it if he
approves. To paraphrase Robert Kern, "Working code ends all arguments".
So, since this was the nested way, and the biggest reason against it was
that it require code change, well, I changed the code, and perhaps the
arguments are ended?

Because I did change a fair amount of data_path -using code, it'd be
good to exercise all your favorite font and image finding sections of
code. The bits I tried worked for me, of course, but no guarantees...

-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier.
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Matplotlib-devel mailing list
Matplotlib-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/matplotlib-devel

OK, with a sledgehammer and a prayer, I got it in. I got the same
error you did, and then had to physically remove mpl-data and it's
contents from svn and commit, restore everything in the new order and
commit again. It compiled, installed and ran, passing the tests I
threw at it.

I only ran about half of backend driver before I grew impatient, so
others may want to do some additional testing. I copied your note
from CHANGELOG to API_CHANGES.

I suggest flushing your old mpl-data in site-packages/matplotlib as
well as your ~/.matplotlib/ttffont.cache file before testing.

Thanks for settling the argument!

JDH

···

On 2/12/07, Andrew Straw <strawman@...36...> wrote:

So, I sent the patch file to John, who'll hopefully commit it if he
approves. To paraphrase Robert Kern, "Working code ends all arguments".
So, since this was the nested way, and the biggest reason against it was
that it require code change, well, I changed the code, and perhaps the
arguments are ended?

Because I did change a fair amount of data_path -using code, it'd be
good to exercise all your favorite font and image finding sections of
code. The bits I tried worked for me, of course, but no guarantees...

Great, thanks for checking that in.

It looks like images/*.png didn't make it in.

John Hunter wrote:

···

On 2/12/07, Andrew Straw <strawman@...36...> wrote:

So, I sent the patch file to John, who'll hopefully commit it if he
approves. To paraphrase Robert Kern, "Working code ends all arguments".
So, since this was the nested way, and the biggest reason against it was
that it require code change, well, I changed the code, and perhaps the
arguments are ended?

Because I did change a fair amount of data_path -using code, it'd be
good to exercise all your favorite font and image finding sections of
code. The bits I tried worked for me, of course, but no guarantees...
    
OK, with a sledgehammer and a prayer, I got it in. I got the same
error you did, and then had to physically remove mpl-data and it's
contents from svn and commit, restore everything in the new order and
commit again. It compiled, installed and ran, passing the tests I
threw at it.

I only ran about half of backend driver before I grew impatient, so
others may want to do some additional testing. I copied your note
from CHANGELOG to API_CHANGES.

I suggest flushing your old mpl-data in site-packages/matplotlib as
well as your ~/.matplotlib/ttffont.cache file before testing.

Thanks for settling the argument!

JDH

-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier.
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Matplotlib-devel mailing list
Matplotlib-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/matplotlib-devel

Andrew Straw wrote:

Great, thanks for checking that in.

It looks like images/*.png didn't make it in.
  

And, grr, I can't put them in, either:

$ svn commit -m "added .png files that didn't make it into new mpl-data location" images
Adding (bin) images/back.png
svn: Commit failed (details follow):
svn: COPY of back.png: 403 Forbidden (https://svn.sourceforge.net)

···

John Hunter wrote:
  

On 2/12/07, Andrew Straw <strawman@...36...> wrote:

So, I sent the patch file to John, who'll hopefully commit it if he
approves. To paraphrase Robert Kern, "Working code ends all arguments".
So, since this was the nested way, and the biggest reason against it was
that it require code change, well, I changed the code, and perhaps the
arguments are ended?

Because I did change a fair amount of data_path -using code, it'd be
good to exercise all your favorite font and image finding sections of
code. The bits I tried worked for me, of course, but no guarantees...
    

OK, with a sledgehammer and a prayer, I got it in. I got the same
error you did, and then had to physically remove mpl-data and it's
contents from svn and commit, restore everything in the new order and
commit again. It compiled, installed and ran, passing the tests I
threw at it.

I only ran about half of backend driver before I grew impatient, so
others may want to do some additional testing. I copied your note
from CHANGELOG to API_CHANGES.

I suggest flushing your old mpl-data in site-packages/matplotlib as
well as your ~/.matplotlib/ttffont.cache file before testing.

Thanks for settling the argument!

JDH

-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier.
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Matplotlib-devel mailing list
Matplotlib-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/matplotlib-devel
  
-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier.
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Matplotlib-devel mailing list
Matplotlib-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/matplotlib-devel

OK, I committed these. I'm not sure why they didn't go in the first
time since they were in my mpl-data/images dir when I added it. Give
it a test drive.

···

On 2/12/07, Andrew Straw <strawman@...36...> wrote:

Great, thanks for checking that in.

It looks like images/*.png didn't make it in.

I've checked out this (with the PNG files !) and it works perfectly. Thanks!

Nicolas

···

On 2/13/07, John Hunter <jdh2358@...149...> wrote:

On 2/12/07, Andrew Straw <strawman@...36...> wrote:
> Great, thanks for checking that in.
>
> It looks like images/*.png didn't make it in.

OK, I committed these. I'm not sure why they didn't go in the first
time since they were in my mpl-data/images dir when I added it. Give
it a test drive.