6e59ed5f0e
Reformatted code through PyCharm.
2014-04-18 10:53:51 +02:00
e3ece754bd
All the strings for new contains have to be bytes
...
Unicode strings are not supported yet. See #121 .
2014-04-07 07:48:35 +02:00
27a44cd664
Added a simple test checking simple strings
2014-03-03 08:12:11 +01:00
9598a4c0f9
Some code beautification for new contains tests
2014-03-03 07:56:06 +01:00
04d42baade
Merging contains branch
2014-02-27 09:47:58 +01:00
dce372cf96
working string "contains" operator
2014-02-18 14:49:03 +01:00
031e8e85ac
Added support shift-left and shift-right operators
...
Removed numexpr/version.pyc
2014-02-13 20:49:19 -05:00
3cf3136532
implemented substring search function: contains(haystack, needle)
2014-02-12 20:49:55 +01:00
a4d8da442b
ENH: Adding function conj
2014-01-22 13:28:24 +01:00
7f1e97b6b9
Avoid running multithreading tests in Sparc architecture. Closes #124 .
2014-01-22 12:30:27 +01:00
d76708cb19
Import only the necessary function from np.testing. Fixes #144 .
2013-09-08 19:54:06 +02:00
8cd9c76d6a
Improved the algorithm to decide the initial number of threads to be used. Closes #110 .
2013-08-29 10:39:46 +02:00
cab9f7cbae
TestResult in numexpr.test(). Closes #111
2013-08-29 09:37:48 +02:00
6d917127c9
Fixing the exception when doing modulus 0 on ints. Closes #107 .
2013-08-26 19:34:28 +02:00
0638f4365e
Restored operators wrongly removed in previous commit
2013-08-26 19:19:32 +02:00
d7e44b738e
Test suite is now fully compatible with nosetests
2013-08-26 18:43:16 +02:00
56c8e1ff3e
Not test subprocess on Hurd/kFreeBSD
2013-04-26 13:36:21 +02:00
3fac3b6cae
Reactivating fix for #24 .
2013-04-26 13:16:26 +02:00
3486c4ad2f
Devised a way for distinguishing between int/long constants in python 3.3. Now every functionality must be the same between both versions of python (bar untested features).
2013-04-13 14:33:42 +02:00
99554fb298
Disable test_zero_div in python 3
2012-10-16 21:54:15 +02:00
d878123905
Fixed disassemble function to work with python 3
2012-10-16 21:44:02 +02:00
9bb32f03f0
Fix handling of int32 on python 3
2012-10-16 21:36:31 +02:00
b78a7ed0d5
Replace deprecated assert_ with assertTrue in unittests
2012-04-22 19:23:08 +02:00
7388296131
Some more bytes vs str fixes in python code
2012-04-22 19:21:00 +02:00
ccf15d693a
Fix py3k incompatibilities in python code
2012-04-21 22:13:05 +02:00
32b37ed8dc
Replace tabs with spaces
2012-04-15 20:40:11 +02:00
b83e68bdb3
Switch from PyString to PyBytes API
2012-04-15 13:18:11 +02:00
db92b02ed9
Fix issue 69, wrong result when out is also an input
...
This doesn't deal with the more general overlapping case.
2012-03-01 15:29:29 -08:00
cf887a68c8
Reduction operations better documented now.
2012-01-08 10:53:47 +01:00
eeff5bba9f
The multiprocess module is not shipped on Python 2.5 by default.
2011-12-31 14:55:27 +01:00
f776d7a770
Added licence boilerplates with proper copyright information. Solves #62 .
2011-11-27 12:05:29 +01:00
2e8798245f
Disabled test for ticket #24 .
2011-11-06 11:41:49 +01:00
d1210be5da
Added sumpport for operations with arrays with dimensions that can be zero. Fixes #64 .
2011-11-05 18:26:35 +01:00
a25b27eaba
Removed a context manager for the sake of Python 2.4 compatibility
2011-11-01 11:40:29 +01:00
93d0c0f9d5
Suppressed undesired RuntimeWarnings (overflow for float32 or div by zero).
2011-11-01 11:14:20 +01:00
b66b539578
raise explicit exception when one is trying to use Python's boolean operators (closes issue #24 )
2011-09-27 15:16:44 +02:00
dab2deacf7
made numexpr behave differently depending on whether or not
...
__future__.division is in the caller globals. It can also be manually
toggled through the truediv keyword argument in both evaluate and NumExpr.
Note that precompile no longer takes **kwargs but a sanitized context and
the key used to cache expressions (names and expr themselves) uses the context,
not the kwargs so that evaluate(xyz) hits the same spot than
evaluate(xyz, optimization='aggressive')
2011-09-27 14:43:03 +02:00
8a1b9a027f
- fixed the test suite: most of the TestExpressions tests were not run !!!
...
in fact, the first test was run over and over again, making it look like
all tests were executed and were successful. Moral of the story: do *not*
define a closure within a for loop.
- added a few tests for pow
- cosmetic changes
2011-09-19 15:12:05 +02:00
a6430e07aa
Merge the newiter branch into default
2011-09-16 15:28:49 -07:00
e453ce084c
Update to work with NumPy trunk, remove original looping method
2011-02-02 08:11:00 +00:00
72222d9d63
Merged in r257:274 from 1.4 branch (releasing 1.4.2).
2011-01-25 13:00:48 +00:00
ee0c156d4d
Removed some overflow warnings in reduction operations.
2011-01-14 13:56:29 +00:00
d7d8b73c7e
Initial import of changes in https://github.com/m-paradox/numexpr
2011-01-14 13:01:55 +00:00
091ce7c4c3
Dividing int arrays by zero gives a 0 as result now. Closes #37 .
2010-11-16 16:41:39 +00:00
f7798e25da
Merged in r251:253 from 1.4 branch (Windows fixes for 1.4.1).
2010-10-20 16:48:08 +00:00
e744ae24f4
Name disambiguation for tests.
2010-09-30 10:25:40 +00:00
623a33c6d6
Workaround for a problem when running with threads in subprocesses.
...
It seems that threads needs to be initialized whenever a subprocess is
created. Addresses #33 .
2010-09-27 09:41:25 +00:00
4d4ccbd8c9
Added tests from newly added where_bbbb and cast_ib. Thanks to Gaëtan de Menten. Fixes #28 .
2010-09-10 10:46:43 +00:00
8a93329681
Added a new test for decreasing the number of threads.
2010-07-31 10:26:01 +00:00
488c7a9205
Fixed the bug exposed in bug.py.
...
The problem was that the internal buffers in expressions need to be refreshed when the
number of threads changes. I delete now the cache for expression each time the number of
threads is changed.
The bug.py has been migrated to a test case in the test suite.
2010-07-30 22:24:33 +00:00
Francesc Alted