PySmell seems like a good starting point.
I think it should be possible, PySmell's idehelper.py
does a majority of the complex stuff, it should just be a case of giving it the current line, offering up the completions (the bit I am not sure about) and then replacing the line with the selected one.
>>> import idehelper
>>> # The path is where my PYSMELLTAGS file is located:
>>> PYSMELLDICT = idehelper.findPYSMELLDICT("/Users/dbr/Desktop/pysmell/")
>>> options = idehelper.detectCompletionType("", "" 1, 2, "", PYSMELLDICT)
>>> completions = idehelper.findCompletions("proc", PYSMELLDICT, options)
>>> print completions
[{'dup': '1', 'menu': 'pysmell.pysmell', 'kind': 'f', 'word': 'process', 'abbr': 'process(argList, excluded, output, verbose=False)'}]
It'll never be perfect, but it would be extremely useful (even if just for completing the stdlib modules, which should never change, so you wont have to constantly regenerate the PYSMELLTAGS file whenever you add a function)
Progressing! I have the utter-basics of completion in place - barely works, but it's close..
I ran python pysmells.py /System/Library/Frameworks/Python.framework/Versions/2.5/lib/python2.5/*.py -O /Library/Python/2.5/site-packages/pysmell/PYSMELLTAGS
Place the following in a TextMate bundle script, set "input: entire document", "output: insert as text", "activation: key equivalent: alt+esc", "scope selector: source.python"
#!/usr/bin/env python
import os
import sys
from pysmell import idehelper
CUR_WORD = os.environ.get("TM_CURRENT_WORD")
cur_file = os.environ.get("TM_FILEPATH")
orig_source = sys.stdin.read()
line_no = int(os.environ.get("TM_LINE_NUMBER"))
cur_col = int(os.environ.get("TM_LINE_INDEX"))
# PYSMELLS is currently in site-packages/pysmell/
PYSMELLDICT = idehelper.findPYSMELLDICT("/Library/Python/2.5/site-packages/pysmell/blah")
options = idehelper.detectCompletionType(cur_file, orig_source, line_no, cur_col, "", PYSMELLDICT)
completions = idehelper.findCompletions(CUR_WORD, PYSMELLDICT, options)
if len(completions) > 0:
new_word = completions[0]['word']
new_word = new_word.replace(CUR_WORD, "", 1) # remove what user has already typed
print new_word
Then I made a new python document, typed "import urll" and hit alt+escape, and it completed it to "import urllib"!
As I said, it's entirely a work-in-progress, so don't use it yet..
Last update:
orestis has integrated this into the PySmell project's code! Any further fiddling will happen on github