- Added support for Sphinx and its dependences, docutils and jinja2: #17
Conversation
bjones1
commented
Aug 10, 2012
bjones1
commented
- Created hooks for sphinx and docutils to copy their data files
- Patched archive.py to set file to a module-named subdir of the executable
- Modified hookutils.py to include routines used by sphinx, docutils hooks
- Added tests to verify these hooks, with accompanying support files
- Created hooks for sphinx and docutils to copy their data files - Patched archive.py to set __file__ to a module-named subdir of the executable - Modified hookutils.py to include routines used by sphinx, docutils hooks - Added tests to verify these hooks, with accompanying support files
- Code cleanup
|
Add a series of commits to current branch is ok. |
|
please use 4 spaces for indentation and not 2. |
| # iu.py to expose _os_sep would probably be better. | ||
| mod_pathed = nm.replace(".","/"); | ||
| mod.__file__ = iu._os_path_join(iu._os_path_dirname(sys.executable),mod_pathed); | ||
| mod.__file__ += "/"; |
There was a problem hiding this comment.
Why does file have to end with slash?
There was a problem hiding this comment.
This does need clarification. I'm adding the following to the source:
# In order to support typical use cases, end the "file" name with
# a slash, so that os.path.dirname(__file__) reports the correct
# directory.
|
Done. On Tue, Aug 14, 2012 at 5:06 AM, Martin Zibricky
Bryan A. Jones, Ph.D. Our Master, Jesus Christ, is on his way. He'll show up right on
|
- Changed from 2 spaces to 4 spaces for indents
…directory. In particular, Sphinx requires that conf.py have exactly that name - test_sphinx-conf.py doesn't work.
- Removed unnecessary entries - Removed comment-out lines
| # Set __file__ attribute of a module relative to sys.executable so | ||
| # that data files can be found. This really shouldn't use "/" as the | ||
| # path separator always, but it works on Windows/Unix for now. Patching | ||
| # iu.py to expose _os_sep would probably be better. |
There was a problem hiding this comment.
Good point! I've taken a different approach, preserving most of the existing code, so this becomes a moot point.
…turned '', instead of the original string.
- Skips directories with no __init__.py - Skips all subdirectories under directories with no __init__.py
Added a set of supporting files to test collect_submodule
/absolute/path/frozen_executable/path/to/module/module_name.pyc.
|
Seems ok for me so far. I'll merge it tomorrow. |
| for dirpath, dirnames, files in os.walk(mod_dir): | ||
| for f in files: | ||
| fpath = os.path.join(dirpath, f) | ||
| if '.py' not in f and not os.path.isdir(fpath): |
There was a problem hiding this comment.
We should ignore more suffixes and not only '.py'. We should ignore .py .pyc .pyo .dll .pyd .so and .dylib files. Maybe .dll files could be handled as data files since some modules on windows distribute dlls with binary installers. Could you please add note to docstring that this function will not work for zipped python eggs?
There was a problem hiding this comment.
I agree -- I'll add this in. Plus, I want to do at least a few unit tests for this routine as well.
There was a problem hiding this comment.
I think the .dlls are already collect by pyinstaller's dependency checker; if not, I'd make this a hook-specific change, since the .dll would need to go in the executable directory, not in the data files directory. I have not experience with Macs -- is this also true for .dylib files as well?
There was a problem hiding this comment.
On windows they might not be collected by dependency checker when the .dll is loaded by ctypes, is not available on PATH and manifest is not included. This is seen more often. On windows dll files should be handled as data files. It is not true for mac. Windows only.
So let's summarize it: ignore extensions .py, .pyc, .pyo, pyd, .so, .dylib
There was a problem hiding this comment.
OK. But, for .dlls, I would think they need to be copied to the executable directory, unless the path is changed by archive.py to the data files directory prior to executing the module.
There was a problem hiding this comment.
If a dll is loaded by ctypes from a specific location then the location could be based on file. Let's keep them in the data files directory module.
There was a problem hiding this comment.
Sounds good -- done.
- Renamed and rewrote remove_suffix function, since os.path.splitext does this - Added PY_EXECUTABLE_EXTENSIONS, a tuple of files which will be collected - Moved to set() instead of list() to eliminate duplicate modules, added because they have differing extensions
- Rewrote tests for remove_extension - Added tests of collect_submodules for varying extensions - Added files with these extensions
…extension Cleaned up the code by removing an unnecessary loop
- Files added to support collect_data_files - Clean up
|
I thing all comments are reflected. Let's try to merge it. |
- Added support for Sphinx and its dependences, docutils and jinja2:
|
Still a few bugs, would you wait?
|
|
create a new request. It is ok since current code is not much broken. Only compatibility with python 2.4 is broken. See |
|
Sounds good -- will do! On Wed, Aug 15, 2012 at 2:32 PM, Martin Zibricky
Bryan A. Jones, Ph.D. Our Master, Jesus Christ, is on his way. He'll show up right on
|
