all_completions returns nothing (simple fix suggested)

[memeplex]

The Completer.all_completions method, which emacs python.el and elpy mode use, is returning nothing defined in the current scope. For example:

In [87]: xxx = 2

In [91]: get_ipython().Completer.all_completions('x')
Out[91]: ['%xdel', '%xmode']

In [92]: get_ipython().Completer.all_completions('xx')
Out[92]: []

Internally it's using jedi but discarding jedi completions. Now, all_completions calls the deprecated complete instead of completions. A simple change fixes both the usage of a deprecated method and a bug:

In [127]: list(map(lambda c: c.text, get_ipython().Completer.completions('x', len('x'))))
Out[127]: ['xxx', '%xdel', '%xmode']

Why not redefining all_completions like that? At least the following is working for me inside an emacs inferior python session:

get_ipython().Completer.all_completions = lambda t: map(lambda c: c.text, get_ipython().Completer.completions(t, len(t)))

[memeplex]

I realized the map is slightly more complicated because of a mismatch between what is returned and what emacs expects:

lambda c: t + c.text if t.endswith(".") else c.text

This way argp -> argparse but argparse. -> argparse.ArgumentError.

[memeplex]

Sorry, this one seems to be the right fix: lambda c: t[:t.rfind(".") + 1] + c.text

• argp -> argparse
• argparse. -> argparse.ArgumentError
• argparse.Arg -> argparse.ArgumentError

It's probably better to first compute t[:t.rfind(".") + 1] once and then do the mapping, I wrote it this way instead because it's a dirty hack I'm adding to exec_lines for now.

[memeplex]

@Carreau would you consider adding this to the next milestone? It's easy to fix and it's really critical for emacs users. Thanks!

[Carreau]

Sure, sending a PR would help. I'm way behind on working on IPython.

[memeplex]

@Carreau check the PR above, please.

[memeplex]

Can we close this? The fix has already been merged.