Having an $EDITOR with arguments raises an exception

[davidparsson]

editor.edit('file') fails when the defined $EDITOR contains spaces:

$ touch /tmp/file
$ export EDITOR='vim -v'
$ python -c 'import editor; editor.edit("/tmp/file")' 
Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "editor.py", line 103, in edit
    proc = subprocess.Popen(args, close_fds=True, stdout=stdout)
  File "/System/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/subprocess.py", line 710, in __init__
    errread, errwrite)
  File "/System/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/subprocess.py", line 1335, in _execute_child
    raise child_exception
OSError: [Errno 2] No such file or directory

After removing the space, it works:

$ export EDITOR='vim'
$ python -c 'import editor; editor.edit("/tmp/file")'

Originally (and improperly) reported here as an Alembic issue.

[eugene-eeo]

@davidparsson I forgot about the old behaviour, but this is most likely ibecause to avoid arbitrary code injection, the library uses subprocess.Popen and passes a list according to the following spec:

program, *args = [a1,a2,...]

the OS then resolves the program executable, e.g. looking in $PATH, etc. and tries to execute the program with the name program along with args as arguments (the strings you find in sys.argv). In this case, vim -v is not a valid program as it tries to execute vim<SPACE>-v, which is not found on your system.

If you're using the latest codebase, as in the master branch, it should not be an issue.

EDIT: I was mistaken as I briefly skimmed through the code. Sorry!

[davidparsson]

@eugene-eeo, are you saying that this functionality is intended and not likely to change? I'm seeing this on the master branch as well.

[eugene-eeo]

@davidparsson I can't read minds, but probably yes. Relevant code: https://github.com/fmoo/python-editor/blob/master/editor.py#L83

[davidparsson]

All right, if so then that's a shame. Since having a space in an editor is supported by most Unix commands, I think it would make sense to support that here as well.

[eugene-eeo]

I think it can be implemented quite easily using the shlex module. Give me a moment and I'll hack up a version and see if it suits your needs.

@davidparsson does this look good? https://gist.github.com/eugene-eeo/f0baa1e79b5efbe240c2d5b3d9a899e8

[fmoo]

I think I'm a fan of making this use case work. Does this break editors that have a whitespace in their executable name?

Not sure that's a use case we need to care about, but I guess we can add some heuristics if we need to support it.

My main concern with the PR that I initially looked at was that it changed a couple of other things that seemed unrelated, but I haven't looked too closely yet.

On Nov 11, 2016, at 7:27 AM, Eeo Jun [email protected] wrote:

I think it can be implemented quite easily using the shlex module. Give me a moment and I'll hack up a version and see if it suits your needs.

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub, or mute the thread.

[eugene-eeo]

shlex handles editor\ binaries\ with\ spaces perfectly.

>>> import shlex
>>> shlex.split('abc\ def')
['abc def']

[davidparsson]

@eugene-eeo, nearly, but not quite. The args passed to Popen for that code are in my case ['subl --wait', '--wait', 'file.txt'] which raises the same exception.

[eugene-eeo]

@davidparsson it should work now :)

[davidparsson]

@eugene-eeo, indeed it does. Thanks!

Do you think that this will be a part of a future release?

[eugene-eeo]

@davidparsson that will be up to @fmoo. However I don't see any drawbacks from this and it makes the library more Unix-friendly so my best guess would be that it would be part of a future release.

[fmoo]

I published a 1.0.2, but setuptools is throwing a hissy fit and not letting me upload via cli.

So I tried uploading via the UI, but pip doesn't seem to see it. Any ideas?

On Nov 16, 2016, at 3:03 PM, Eeo Jun [email protected] wrote:

@davidparsson that will be up to @fmoo. However I don't see any drawbacks from this and it makes the library more Unix-friendly so my best guess would be that it would be part of a future release.

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub, or mute the thread.

[eugene-eeo]

Can you share the warning that setuptools displayed? Maybe you've previously published the same version. BTW, my PR that you merged does not include this fix. Let me know if I should include the support for spaces as well, and I'll drop it in and write some tests.

[fmoo]

It was an authentication error. Which is weird because I can use the credentials to login through the web UI.

On Nov 17, 2016, at 4:42 AM, Eeo Jun [email protected] wrote:

Can you share the warning that setuptools displayed? Maybe you've previously published the same version.

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub, or mute the thread.

[eugene-eeo]

Try checking for extraneous spaces at the end and such.

[davidparsson]

@fmoo, please note that a fix for this has not been merged, since it was not included in #12, as stated above.

[fmoo]

I guess we could try to resolve the binary before execution. If it fails, we could fall back to a split/rejoin approach. Thoughts?

[davidparsson]

@fmoo, sorry, I don't think I can be of any help there.

@eugene-eeo, would you mind opening a pull request with the fix for this issue as well?

[eugene-eeo]

@fmoo please don't do that because it's IMO really scary magic that you need to maintain in the future.

@davidparsson sure, give me a moment.

[davidparsson]

I'm still having this issue. Is there anything I can do to help get this fixed and released?

[eugene-eeo]

@davidparsson I think the bugfix is released in the latest version. If not you can try bugging @fmoo to release the current code to PyPI.

[fmoo]

The latest build on pypi is 1.0.3, which is the most recent commit in the repo. Spaces should no longer raise an exception, but it will treat spaces like part of the editor filename (e.g., looking for a file named `vim -e` instead of passing `-e` to vim).

[davidparsson]

I'm still able to reproduce the problem in 1.0.3, and I don't see the fix with shlex for this in editor.py on master.

The gist in #11 (comment) was never committed/merged.

I've opened #15 with those changes.

[fmoo]

Also wow I totally missed the updates to this thread last year. I'll take a look after work today.