Jorgen Schaefer
9ed0dc4861
Always load source files in binary mode.
...
Source files can be in any coding system, provided Python can
read a coding: line at the beginning of the file. So source files
should be loaded in binary format and decoded according to that
line, not assumed to be in the default coding system.
Fixes #398 .
2014-04-24 11:10:59 +02:00
Dave Halter
67f49da15c
Increase maximum recursion depth. Makes sense for very complicated code (completions within Jedi), since Jedi itself uses so many recursions.
2014-04-20 14:55:05 +02:00
Dave Halter
81cc50e8b5
well yeah give NamePart a get_code method after all, makes sense, because all the other parser representation objects have it.
2014-04-20 14:32:48 +02:00
Dave Halter
50c4b7bfd9
memoize problems with defaults in combination with raised exceptions
2014-04-20 14:14:30 +02:00
Dave Halter
27f01ca1f6
use Param.stars to avoid certain syntax error params to cause an exception
2014-04-20 13:38:10 +02:00
Dave Halter
ab154d46d5
fixed a very nasty problem with exponential growth in precedence for loops
2014-04-20 12:42:53 +02:00
Dave Halter
4f3a15e235
solve random stdlib issue
2014-04-20 01:25:02 +02:00
Dave Halter
404baf5020
use Param.stars more generally
2014-04-20 00:12:52 +02:00
Dave Halter
2b091076c1
fix the named param issue in the autocompletion
2014-04-19 22:57:41 +02:00
Dave Halter
10e2dc2f6c
remove unused imports from helpers
2014-04-19 16:28:31 +02:00
Dave Halter
0b5ed6a5fe
array_for_pos -> call_signature_array_for_pos
2014-04-19 16:27:41 +02:00
Dave Halter
1f7fd6f439
search_call_signature refactoring
2014-04-19 16:26:15 +02:00
Dave Halter
bb6874bc7c
fix for problems with incomplete one liner string literals, after a start of an incomplete string literal the whole line should be seen as an error token
2014-04-19 13:56:29 +02:00
Dave Halter
33fa2b0cb4
call_signatures should work on more nested edge cases
2014-04-18 17:17:10 +02:00
Dave Halter
efc24a9ecc
forgot __ne__ function for python 2 compatibility in Definition
2014-04-18 14:53:43 +02:00
Dave Halter
0301606d18
Now remove Usage completely.
2014-04-18 14:45:03 +02:00
Dave Halter
f0e7b5583d
update parser pickling version again, because we've changed a lot in the last few commits
2014-04-18 14:41:29 +02:00
Dave Halter
547ec56bd3
remove NamePart.string, can be done by casting it to unicode
2014-04-18 14:40:28 +02:00
Dave Halter
240b0c9581
NamePart doesn't have an __eq__ method anymore
2014-04-18 14:36:10 +02:00
Dave Halter
c2bdda339b
again Definition/Usage merging
2014-04-18 01:51:09 +02:00
Dave Halter
b643325889
type of NamePart in definition should work.
2014-04-18 01:31:07 +02:00
Dave Halter
5f4c4de229
Usage is now a subclass of Definition, #395
2014-04-17 14:39:22 +02:00
Dave Halter
a92c9dd81b
added a docstring to _Help
2014-04-17 14:21:20 +02:00
Dave Halter
084a4a5a43
Help -> _Help for now
2014-04-17 14:14:30 +02:00
Dave Halter
0582979db5
documentation to docstring, #395
2014-04-17 14:13:47 +02:00
Dave Halter
2384556861
use Help instead of Documentation. see #392 for a discussion about that.
2014-04-16 11:18:22 +02:00
Dave Halter
ddc1cb0e6a
same problem with KeywordStatement
2014-04-16 10:28:12 +02:00
Dave Halter
15fdecdb61
for flows triggered an exception if goto_assignments was used on the keyword. found with the help of sith.py
2014-04-16 09:57:01 +02:00
Dave Halter
a341791fda
found list indexing issue with sith and fixed it.
2014-04-16 01:31:49 +02:00
Dave Halter
b81eb9f8b3
NameFinder.filter_name is simpler now.
2014-04-14 13:45:31 +02:00
Dave Halter
6a40c9b671
remove get_set_vars completely from existance
2014-04-14 12:40:59 +02:00
Dave Halter
237af765b7
start to get rid of the get_set_vars/get_defined_names distinction
2014-04-14 12:28:24 +02:00
Dave Halter
4c53a64ca0
increase speed again for compiled objects by not using sorted + lambda on its list of defined names
2014-04-14 10:12:46 +02:00
Dave Halter
ff810d9ece
cache compiled.CompiledObject.defined_names attribute, improves test suite performance by 20%, numpy by more than 50%. Awesome!
2014-04-14 02:20:22 +02:00
Dave Halter
3fced34544
the last commit featured an incomplete caching for compiled objects. The current one should improve this.
2014-04-13 21:55:29 +02:00
Dave Halter
4bc55be103
significant speedup due to compiled caching
2014-04-13 16:31:38 +02:00
Dave Halter
edeebd0bb9
Name initialization should always take the same param types as input
2014-04-13 15:00:49 +02:00
Dave Halter
058e123879
fix issues with a copy.copy usage for Name
2014-04-13 14:55:07 +02:00
Dave Halter
3d9d0bfd03
Name.get_code caching to make lookups faster
2014-04-12 15:28:19 +02:00
Dave Halter
27645af6d2
make a slow inspect.getdoc() call lazy, which improves performance for numpy completions by 30%
2014-04-11 11:53:11 +02:00
Dave Halter
a6bfb1b3ad
fix issues with os.path completions
2014-04-11 10:33:32 +02:00
Dave Halter
840a806246
generalize import optimizations and make them behave more like sys.modules
2014-04-10 15:21:23 +02:00
Dave Halter
4bc89d638c
start caching some parts of the imports to significantly speed up numpy completions.
2014-04-10 14:55:49 +02:00
Dave Halter
14bf618af0
put some list casts into a different place.
2014-04-10 12:24:30 +02:00
Dave Halter
6de46fe373
realizing that #241 is more of a feature than a bug, I disabled the corresponding test. This will be changed once the whole implementation of isinstance becomes more flexible and also allows checks on subclasses, so that no information is lost.
2014-04-09 16:27:42 +02:00
Dave Halter
d8d6b20a17
fix line split issues in cache
2014-04-09 13:16:28 +02:00
Dave Halter
46277eb9c9
call_signatures caching should be much more precise, now. fixes #390
2014-04-09 12:27:23 +02:00
Dave Halter
d2dc39e0c2
also allow unicode, bytes, bytearray, dict -> fixes #297
2014-04-07 16:16:31 +02:00
Dave Halter
193e04ae8e
get rid of side effects in the interpreter if trying to use the index on an iterable (using __getitem__)
2014-04-07 16:11:23 +02:00
Dave Halter
ea62ad6a50
backwards tokenizer can now handle 10e-5 and so on literals
2014-04-07 15:51:24 +02:00
