`search_src` and friends shouldn't look in hidden files

[ddrake]

I'm currently editing a file in the Sage library, and emacs created a hidden file .#plot.py that, until I save plot.py, is a broken symlink. When I do search_def, I get:

IOError: [Errno 2] No such file or directory: '/scratch/sage-4.7.2/devel/sage/sage/plot/.#plot.py'

I think search_src and friends should just ignore hidden files, since no genuine source file in the Sage library is hidden!

[jhpalmieri]

I've been annoyed about this too. Something like this should do it:

sage/misc/sagedoc.py

@@ -695 +695 @@
     # done with preparation; ready to start search
     for dirpath, dirs, files in os.walk(os.path.join(base_path, module)):
         for f in files:
+            # if f matches some pattern because it's a temporary file, skip it:
+            if f.startswith('.#'):
+                continue
             if re.search("\.(" + "|".join(exts) + ")$", f):
                 filename = os.path.join(dirpath, f)
                 if re.search(path_re, filename):

But maybe a regular expression match for temporary files would be better, perhaps combined with the line "if re.search(...)". Are there any patterns to ignore besides ".#..."?

[ddrake]

Replying to jhpalmieri:

Are there any patterns to ignore besides ".#..."?

I would ignore all hidden files. Do we only look through .py and .pyx files? Because emacs also makes a #plot.py# file which we should ignore. I know vim makes .foo.swp files, and I don't know anything about any other editor.

[kini]

I notice that ls -B (from GNU coreutils anyway) ignores files that end with "~". Not that ls has anything to do with search_src() in particular, but maybe we should consider ignoring those too since such a universal tool considers them ignore-worthy.

[jhpalmieri]

As it is, the search functions only look at files ending in .html (for search_doc) or ['.py', '.pyx', '.pxd'] (for search_src and search_def). So we don't need to worry about suffixes, just prefixes.

[kini]

Ah, good point. Never mind then. This also means that looking for ^#(.*)#$ is unnecessary (emacs seems to produce these too, at least on my machine).

[ddrake]

A one-line patch! Please review.

(John, I modified your example. Add yourself to the author field if you like.)

I tested this by creating a dangling symlink; I suppose I can make a doctest to test that, but that doesn't seem all that necessary. Should I do so anyway?

[jhpalmieri]

How could you doctest this?

[ddrake]

Replying to jhpalmieri:

How could you doctest this?

I'm not sure...create a hidden file with an unlikely string in it, do search_src for that string? Check it returns nothing?

OTOH, I feel pretty confident about the code "not f.startswith('.')". So it seems reasonable to leave this one as it is.

Thanks for the review!