Make rlcompleter avoid duplicate global names

[vadmium]

BPO	25663
Nosy	@vadmium, @serhiy-storchaka
Files	global-dupes.patch

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = <Date 2015-11-24.02:07:50.922>
created_at = <Date 2015-11-19.02:14:20.907>
labels = ['type-bug', 'library']
title = 'Make rlcompleter avoid duplicate global names'
updated_at = <Date 2015-11-24.02:07:50.921>
user = 'https://github.com/vadmium'

bugs.python.org fields:

activity = <Date 2015-11-24.02:07:50.921>
actor = 'martin.panter'
assignee = 'none'
closed = True
closed_date = <Date 2015-11-24.02:07:50.922>
closer = 'martin.panter'
components = ['Library (Lib)']
creation = <Date 2015-11-19.02:14:20.907>
creator = 'martin.panter'
dependencies = []
files = ['41071']
hgrepos = []
issue_num = 25663
keywords = ['patch']
message_count = 5.0
messages = ['254873', '254882', '255238', '255239', '255243']
nosy_count = 3.0
nosy_names = ['python-dev', 'martin.panter', 'serhiy.storchaka']
pr_nums = []
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue25663'
versions = ['Python 2.7', 'Python 3.4', 'Python 3.5', 'Python 3.6']

[vadmium]

When playing with the Editline alternative to Readline, I noticed that “global” name completions can include duplicates:

>>> No
None                 NotADirectoryError(  NotImplementedError(
None                 NotImplemented      
>>> None
None None

It completed my line to “None”, but if you hit Tab again, it lists two duplicate options both identical to what I already have. The reason is that “None” is both a reserved keyword, and a member of the builtins module.

My patch avoids adding extra completions if a name has already been added. It also prioritizes the global namespace over builtins, so that say if you alias “int” to a non-callable, it is no longer listed with an opening bracket “(” suffix. Earlier behaviour:

>>> int = 81
>>> in
in     input( int    int(  

Now:

>>> in
in     input( int

[serhiy-storchaka]

Nice work.

An alternative approach is to make "matches" a dict. And instead of

    if match not in seen:
        seen.add(word)
        matches.append(match)

use just

matches[word] = match

I don't know what approach is better.

Added other minor comments on Rietveld.

[python-dev]

New changeset 4799615f4f26 by Martin Panter in branch '3.4':
Issue bpo-25663: Make rlcompleter avoid duplicate global names
https://hg.python.org/cpython/rev/4799615f4f26

New changeset 4ed70c568baf by Martin Panter in branch '3.5':
Issue bpo-25663: Merge rlcompleter fix from 3.4 into 3.5
https://hg.python.org/cpython/rev/4ed70c568baf

New changeset 96fb9daf64a5 by Martin Panter in branch 'default':
Issue bpo-25663: Merge rlcompleter fix from 3.5
https://hg.python.org/cpython/rev/96fb9daf64a5

New changeset f742c31491e3 by Martin Panter in branch 'default':
Issue bpo-25663: Update rlcompleter test for new 3.6 behaviour
https://hg.python.org/cpython/rev/f742c31491e3

[python-dev]

New changeset d3a9b055206c by Martin Panter in branch '2.7':
Issue bpo-25663: Make rlcompleter avoid duplicate global names
https://hg.python.org/cpython/rev/d3a9b055206c

[vadmium]

I chose to keep my strategy of using the set. Using a dictionary to hold the matches could mess up the order, although it looks like Readline sorts the matches before displaying them.