As the release is out, we can now think about how to allow for Cython in our build process.
The current build system is rather schizophrenic (numpy distutils and makefiles), so ideas which way to go are welcome. The process should be simple :o)
r.
Hi Robert,
Some of my deadlines are passing, so I hope to be able to turn more attention to SfePy again. Sorry for my disappearance lately. :| Things are still busy, but hopefully I have more control over it. :)
Regarding the build process, as you know, CMAKE seems to be making progress and supports cross-platform builds. Plus we would have the advantage of the examples hermes guys have developed.
A lighter alternative may be waf, but I haven't followed the progress there lately.
Anyway, just wanted to start some discussion on this issue.
Greetings, Logan
2010/8/6 Robert Cimrman <cimr...@ntc.zcu.cz>:
As the release is out, we can now think about how to allow for Cython in our build process.
The current build system is rather schizophrenic (numpy distutils and makefiles), so ideas which way to go are welcome. The process should be simple :o)
r.
-- You received this message because you are subscribed to the Google Groups "sfepy-devel" group. To post to this group, send email to sfepy...@googlegroups.com. To unsubscribe from this group, send email to sfepy-devel...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/sfepy-devel?hl=en.
Hi Logan,
I am glad you are back!
On Tue, 14 Sep 2010, Logan Sorenson wrote:
Hi Robert,
Some of my deadlines are passing, so I hope to be able to turn more attention to SfePy again. Sorry for my disappearance lately. :| Things are still busy, but hopefully I have more control over it. :)
Good!
Regarding the build process, as you know, CMAKE seems to be making progress and supports cross-platform builds. Plus we would have the advantage of the examples hermes guys have developed.
You mean examples how cmake tests the environment?
I am ok with cmake. But considering (maybe wrongly), that future sfepy is going to be mostly pure python + some cython, I wonder whether we should stick to python build means like numpy distutils. I have catched some "gossip" that there are ways how to make them work with cython.
A lighter alternative may be waf, but I haven't followed the progress there lately.
I was not aware of this, and it's Python, good.
There is also the new package of David Cournapeau called bento [1], [2] which might be very interesting - after all he is the guy that had to solve numpy/scipy building process (on windows) with all the blas/lapack version, compilers, etc.
Anyway, just wanted to start some discussion on this issue.
Great!
So, which tool shall we use? I am slightly in favor of sticking to numpy distutils, but bento seems interesting (it's license looks BSD-like, so it might be even possible to include it in our sources).
Best regards, r.
Greetings, Logan
2010/8/6 Robert Cimrman <cimr...@ntc.zcu.cz>:
As the release is out, we can now think about how to allow for Cython in our build process.
The current build system is rather schizophrenic (numpy distutils and makefiles), so ideas which way to go are welcome. The process should be simple :o)
r.
-- You received this message because you are subscribed to the Google Groups "sfepy-devel" group. To post to this group, send email to sfepy...@googlegroups.com. To unsubscribe from this group, send email to sfepy-devel...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/sfepy-devel?hl=en.
-- You received this message because you are subscribed to the Google Groups "sfepy-devel" group. To post to this group, send email to sfepy...@googlegroups.com. To unsubscribe from this group, send email to sfepy-devel...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/sfepy-devel?hl=en.
Hi Robert,
2010/9/15 Robert Cimrman <cimr...@ntc.zcu.cz>:
Hi Logan,
I am glad you are back!
Glad to be back!
On Tue, 14 Sep 2010, Logan Sorenson wrote:
Hi Robert,
Some of my deadlines are passing, so I hope to be able to turn more attention to SfePy again. Sorry for my disappearance lately. :| Things are still busy, but hopefully I have more control over it. :)
Good!
Regarding the build process, as you know, CMAKE seems to be making progress and supports cross-platform builds. Plus we would have the advantage of the examples hermes guys have developed.
You mean examples how cmake tests the environment?
Yes, and also how to search subdirectories, build executables, link libraries, etc. I played a bit with the cmake files in hermes to get my build to work; it's very flexible and not too difficult to figure out, but also seems a bit heavy and not pythonic.
I am ok with cmake. But considering (maybe wrongly), that future sfepy is going to be mostly pure python + some cython, I wonder whether we should stick to python build means like numpy distutils. I have catched some "gossip" that there are ways how to make them work with cython.
My preference is a python-based build system as well. :) Interesting to hear that cython support is planned for numpy.distutils.
A lighter alternative may be waf, but I haven't followed the progress there lately.
I was not aware of this, and it's Python, good.
There is also the new package of David Cournapeau called bento [1], [2] which might be very interesting - after all he is the guy that had to solve numpy/scipy building process (on windows) with all the blas/lapack version, compilers, etc.
Ah, interesting, wasn't aware of bento. Definitely, being able to build numpy/scipy on multiple platforms is a vote of confidence. A quick google search for waf cython support turned up a few interesting discussions on the numpy-discussion list [3,4]. Also, waf seems to have at least an initial cython support [5].
Anyway, just wanted to start some discussion on this issue.
Great!
So, which tool shall we use? I am slightly in favor of sticking to numpy distutils, but bento seems interesting (it's license looks BSD-like, so it might be even possible to include it in our sources).
+1 for sticking to numpy distutils since it is working now and the developers seem to be aware of the need for cython support. We can always add in bento+yaku or waf if we need more powerful tools later on. I think waf is also BSD licensed (according to their webpage) and is a self-contained script so we can distribute it w/ SfePy.
Best, Logan
Best regards, r.
Greetings, Logan
2010/8/6 Robert Cimrman <cimr...@ntc.zcu.cz>:
As the release is out, we can now think about how to allow for Cython in our build process.
The current build system is rather schizophrenic (numpy distutils and makefiles), so ideas which way to go are welcome. The process should be simple :o)
r.
[3] http://www.mail-archive.com/numpy-discussion@scipy.org/msg23048.html [4] http://www.mail-archive.com/numpy-discussion@scipy.org/msg27260.html [5] http://code.google.com/p/waf/issues/detail?id=668
Hi Logan,
On Wed, 15 Sep 2010, Logan Sorenson wrote:
Hi Robert,
2010/9/15 Robert Cimrman <cimr...@ntc.zcu.cz>:
Hi Logan,
I am glad you are back!
Glad to be back!
On Tue, 14 Sep 2010, Logan Sorenson wrote:
Hi Robert,
Some of my deadlines are passing, so I hope to be able to turn more attention to SfePy again. Sorry for my disappearance lately. :| Things are still busy, but hopefully I have more control over it. :)
Good!
Regarding the build process, as you know, CMAKE seems to be making progress and supports cross-platform builds. Plus we would have the advantage of the examples hermes guys have developed.
You mean examples how cmake tests the environment?
Yes, and also how to search subdirectories, build executables, link libraries, etc. I played a bit with the cmake files in hermes to get my build to work; it's very flexible and not too difficult to figure out, but also seems a bit heavy and not pythonic.
I see. Well, hermes is primarily C++, so it needs a more heavy-weight build system than we need, IMHO.
I am ok with cmake. But considering (maybe wrongly), that future sfepy is going to be mostly pure python + some cython, I wonder whether we should stick to python build means like numpy distutils. I have catched some "gossip" that there are ways how to make them work with cython.
My preference is a python-based build system as well. :) Interesting to hear that cython support is planned for numpy.distutils.
There was a pyrex support from the very beginning (IMHO), but Cython needed some hacks. It would be great if you find some time to check the current status. :)
A lighter alternative may be waf, but I haven't followed the progress there lately.
I was not aware of this, and it's Python, good.
There is also the new package of David Cournapeau called bento [1], [2] which might be very interesting - after all he is the guy that had to solve numpy/scipy building process (on windows) with all the blas/lapack version, compilers, etc.
Ah, interesting, wasn't aware of bento. Definitely, being able to build numpy/scipy on multiple platforms is a vote of confidence. A quick google search for waf cython support turned up a few interesting discussions on the numpy-discussion list [3,4]. Also, waf seems to have at least an initial cython support [5].
Anyway, just wanted to start some discussion on this issue.
Great!
So, which tool shall we use? I am slightly in favor of sticking to numpy distutils, but bento seems interesting (it's license looks BSD-like, so it might be even possible to include it in our sources).
+1 for sticking to numpy distutils since it is working now and the developers seem to be aware of the need for cython support. We can always add in bento+yaku or waf if we need more powerful tools later on. I think waf is also BSD licensed (according to their webpage) and is a self-contained script so we can distribute it w/ SfePy.
ok
So if you could try (e.g. on an artificial test project) to build some test cython module with numpy distutils, that would be a good start.
Let's discuss this in a new thread (e.g. build system) - the current subject is not very good.
Thanks & cheers! r.
2010/8/6 Robert Cimrman <cimr...@ntc.zcu.cz>:
As the release is out, we can now think about how to allow for Cython in our build process.
The current build system is rather schizophrenic (numpy distutils and makefiles), so ideas which way to go are welcome. The process should be simple :o)
r.
[3] http://www.mail-archive.com/numpy-discussion@scipy.org/msg23048.html [4] http://www.mail-archive.com/numpy-discussion@scipy.org/msg27260.html [5] http://code.google.com/p/waf/issues/detail?id=668
-- You received this message because you are subscribed to the Google Groups "sfepy-devel" group. To post to this group, send email to sfepy...@googlegroups.com. To unsubscribe from this group, send email to sfepy-devel...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/sfepy-devel?hl=en.
2010/9/15 Robert Cimrman <cimr...@ntc.zcu.cz>:
Hi Logan,
On Wed, 15 Sep 2010, Logan Sorenson wrote:
Hi Robert,
2010/9/15 Robert Cimrman <cimr...@ntc.zcu.cz>:
Hi Logan,
I am glad you are back!
Glad to be back!
On Tue, 14 Sep 2010, Logan Sorenson wrote:
Hi Robert,
Some of my deadlines are passing, so I hope to be able to turn more attention to SfePy again. Sorry for my disappearance lately. :| Things are still busy, but hopefully I have more control over it. :)
Good!
Regarding the build process, as you know, CMAKE seems to be making progress and supports cross-platform builds. Plus we would have the advantage of the examples hermes guys have developed.
You mean examples how cmake tests the environment?
Yes, and also how to search subdirectories, build executables, link libraries, etc. I played a bit with the cmake files in hermes to get my build to work; it's very flexible and not too difficult to figure out, but also seems a bit heavy and not pythonic.
I see. Well, hermes is primarily C++, so it needs a more heavy-weight build system than we need, IMHO.
I am ok with cmake. But considering (maybe wrongly), that future sfepy is going to be mostly pure python + some cython, I wonder whether we should stick to python build means like numpy distutils. I have catched some "gossip" that there are ways how to make them work with cython.
My preference is a python-based build system as well. :) Interesting to hear that cython support is planned for numpy.distutils.
There was a pyrex support from the very beginning (IMHO), but Cython needed some hacks. It would be great if you find some time to check the current status. :)
A lighter alternative may be waf, but I haven't followed the progress there lately.
I was not aware of this, and it's Python, good.
There is also the new package of David Cournapeau called bento [1], [2] which might be very interesting - after all he is the guy that had to solve numpy/scipy building process (on windows) with all the blas/lapack version, compilers, etc.
Ah, interesting, wasn't aware of bento. Definitely, being able to build numpy/scipy on multiple platforms is a vote of confidence. A quick google search for waf cython support turned up a few interesting discussions on the numpy-discussion list [3,4]. Also, waf seems to have at least an initial cython support [5].
Anyway, just wanted to start some discussion on this issue.
Great!
So, which tool shall we use? I am slightly in favor of sticking to numpy distutils, but bento seems interesting (it's license looks BSD-like, so it might be even possible to include it in our sources).
+1 for sticking to numpy distutils since it is working now and the developers seem to be aware of the need for cython support. We can always add in bento+yaku or waf if we need more powerful tools later on. I think waf is also BSD licensed (according to their webpage) and is a self-contained script so we can distribute it w/ SfePy.
ok
So if you could try (e.g. on an artificial test project) to build some test cython module with numpy distutils, that would be a good start.
Let's discuss this in a new thread (e.g. build system) - the current subject is not very good.
Sounds good, I will start looking into this!
Best, Logan
Thanks & cheers! r.
2010/8/6 Robert Cimrman <cimr...@ntc.zcu.cz>:
As the release is out, we can now think about how to allow for Cython in our build process.
The current build system is rather schizophrenic (numpy distutils and makefiles), so ideas which way to go are welcome. The process should be simple :o)
r.
[3] http://www.mail-archive.com/numpy-discussion@scipy.org/msg23048.html [4] http://www.mail-archive.com/numpy-discussion@scipy.org/msg27260.html [5] http://code.google.com/p/waf/issues/detail?id=668
-- You received this message because you are subscribed to the Google Groups "sfepy-devel" group. To post to this group, send email to sfepy...@googlegroups.com. To unsubscribe from this group, send email to sfepy-devel...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/sfepy-devel?hl=en.
-- You received this message because you are subscribed to the Google Groups "sfepy-devel" group. To post to this group, send email to sfepy...@googlegroups.com. To unsubscribe from this group, send email to sfepy-devel...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/sfepy-devel?hl=en.
participants (2)
-
Logan Sorenson -
Robert Cimrman