Help string support for fish

[volkov]

Help string support for fish.

Work in progress.

dfs and _ARGCOMPLETE_DFS should probably be renamed to something more meaningful.

[codecov-io]

Codecov Report

Merging #264 into master will increase coverage by 0.23%.
The diff coverage is 94.44%.

@@            Coverage Diff             @@
##           master     #264      +/-   ##
==========================================
+ Coverage   82.72%   82.95%   +0.23%     
==========================================
  Files           7        7              
  Lines         741      751      +10     
==========================================
+ Hits          613      623      +10     
  Misses        128      128              

Impacted Files	Coverage Δ
argcomplete/shell_integration.py	100.00% <ø> (ø)
argcomplete/__init__.py	91.55% <94.44%> (+0.23%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 651e9fe...67e617f. Read the comment docs.

[kislyuk]

@volkov thanks for your work on this, and sorry about the delay in my review.

This looks good overall. Can you add a description of this feature, and a rendering of what it looks like in the shell, in the project README under the "fish support" section?

Also, it's certainly possible that completions will contain spaces (I'm sure you realize that completions can be produced for values, not just option names). In those cases, can you describe and paste renderings of what happens? How does this code interact with completions that contain both quotes and spaces? Since the only completions that contain spaces are value completions, and those are unlikely to have help, perhaps it's better to skip them instead of splitting them the way you are doing?

[volkov]

@kislyuk thanks for your review

Can you add a description of this feature, and a rendering of what it looks like in the shell, in the project README under the "fish support" section?

Yes, before finishing work on it.

Also, it's certainly possible that completions will contain spaces (I'm sure you realize that completions can be produced for values, not just option names). In those cases, can you describe and paste renderings of what happens? How does this code interact with completions that contain both quotes and spaces? Since the only completions that contain spaces are value completions, and those are unlikely to have help, perhaps it's better to skip them instead of splitting them the way you are doing?

I can't find how to set help string for value completions. Tried to use dict as value for choices, but it does not help. There is no clue in documentation also. Do you know how to set it?

I have added space to subparser name:

subparsers.add_parser('basic with space', help="basic help").add_argument('arg', choices=['foo', 'bar', 'baz'])

In this it drops help message:

Same with quotes:

I found another issue - help message could contain _ARGCOMPLETE_DFS or _ARGCOMPLETE_IFS - in this case additional completions are added.

[volkov]

So I fixed problem with ifs in descriptions and added some documentation.

But I still don't know good solution space handling.

If I don't split I miss description for options with aliases, e.g help here:

If I split (this is currently realized in PR) I miss descriptions for subcommands with spaces

[evanunderscore]

From my reading, it looks like the original is trying to display all of the aliases together, e.g. -h --help -- show this help and exit, whereas you're trying to duplicate the help message for each parameter, e.g. -h (show this help message and exit) and --help (show this help message and exit).

Since the previous API hides the internal dict behind a function call get_display_completions, nothing should stop you from modifying what's stored in self._display_completions and then adding whatever logic you need in that function to restore it to what it was previously.

For example (not tested):

    def _get_option_completions(self, parser, cword_prefix):
        self._display_completions.update(
            # replace " ".join with tuple
            [[tuple(ensure_str(x) for x in action.option_strings
                       if ensure_str(x).startswith(cword_prefix)), action.help]
             for action in parser._actions
             if action.option_strings])

def get_display_completions():
    # " ".join the tuple to preserve previous API
    return {" ".join(k): v for k, v in self._display_completions}

        if dfs:
            display_completions = {key_part: value.replace(ifs, " ") if value else ""
                                   # use self._display_completions directly
                                   for key, value in self._display_completions.items()
                                   # no need to split key since it is already a tuple
                                   for key_part in key}

The actual change will be a little more involved since you'll either have to find all the other places in the code adding string keys and turn them into one element tuples, or account for both strings and tuples when you use the dict.

[volkov]

@evanunderscore that's good idea, I had changed pull request accordingly.

[volkov]

And now I'm investigating problem with my code around argcomplete.CompletionFinder._get_subparser_completions

[codecov-commenter]

Codecov Report

Merging #264 into master will increase coverage by 0.58%.
The diff coverage is 93.75%.

@@            Coverage Diff             @@
##           master     #264      +/-   ##
==========================================
+ Coverage   82.32%   82.91%   +0.58%     
==========================================
  Files           7        7              
  Lines         747      749       +2     
==========================================
+ Hits          615      621       +6     
+ Misses        132      128       -4     

Impacted Files	Coverage Δ
argcomplete/shell_integration.py	100.00% <ø> (ø)
argcomplete/__init__.py	91.50% <93.75%> (+1.14%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update d504584...bb14c9e. Read the comment docs.

[evanunderscore]

This looks great! Thanks for your persistence with this @volkov. I'm happy with the alias changes as illustrated by the changes to the test code, but since this is technically a breaking change I'll leave it to @kislyuk to sign off on it.

[volkov]

@kislyuk have a look please

[kislyuk]

LGTM - thanks for sticking with this PR through the long development cycle @volkov. I'll roll a release shortly.