[Python-modules-commits] [bottleneck] annotated tag master/1.2.0 created (now 2cf25d1)
Ghislain Vaillant
ghisvail-guest at moszumanska.debian.org
Sat Nov 4 10:10:27 UTC 2017
This is an automated email from the git hooks/post-receive script.
ghisvail-guest pushed a change to annotated tag master/1.2.0
in repository bottleneck.
at 2cf25d1 (tag)
tagging a92326c3956de3ba5b529b0a209ebb95e3a1e7f0 (commit)
tagged by Pietro Battiston
on Thu Oct 20 09:51:22 2016 +0200
- Log -----------------------------------------------------------------
Tagged master/1.2.0
Alessandro Amici (1):
Minimal change in metadata that fixes the top_level.txt file
Ben Root (1):
Quick-n-dirty fixes to "chicken-n-egg" build problem
Christoph Gohlke (3):
TST: fix thousands of deprecation warnings during tests with numpy 1.8 on Python 3
Define inline and NAN for msvc9 and msvc10
Declare variables at the beginning of a scope block
Dougal Sutherland (12):
switch file() to open() in setup.py
fix print and raise statements to be py3-compatible
fix some relative imports
make range(...) + [None] work in python3
None isn't comparable to ints in python 3
python3 compatability for range() in slow funcs
slow rankdata/nanrankdata should use np.asarray
integer division in benchmark code
this relative import isn't python 2.5 friendly
allow specifying python command in makefile
some templating code compatabilities
update release notes, readme, setup.py to note Python 3 support
Jaime Fernandez (1):
ENH: Use Welford's method for move_std
Jenn Olsen (6):
move median accepts nans, and min_count
catch OOM error
fix corner case, and whitespace cleanup
no longer need special case
fix test
fixed test
Jens Hedegaard Nielsen (1):
Additional fix related to #49. The gnuc99 extension should also be enabled when building from the c source to fix the build on Mountain Lion.
John Benediktsson (1):
BUG change to use gnu89 on all platforms.
Keith Goodman (843):
Initial commit of Nanny prototype.
Added partial support for nansum(int32); doc tweaks.
Put structure in place to add nanmax.
nanmax() now passes unit tests. Still need to take care of nanmax([]).
Unit tests for more corner cases; fix bugs exposed.
Added a performance benchmark suite.
nanmax now raises a ValueError with size 0 input to match numpy.
Faster (less overhead) plus user can now get underlying cythion function.
Added nanmin(). Bug fix for nanmax_3d_int32_axisNone().
Added nanmean().
Added nanstd().
Added nanvar().
Renamed package to dsna: descriptive statistics of numpy arrays.
Removed the nan prefix from all function names.
Update readme for change in project direction.
Added directory structure for move and group functions.
Added sphinx docs to project.
Doc strings, doc strings, doc strings.
The bits needed to make a release.
Added the first moving window function: move_sum.
First group function: group_mean.
Renamed package from DSNA to Bottleneck.
Doc tweaks.
Added move and group extension modules to setup.py.
Add URLs to doc.
Arg. Pypi pages are case senstive and won't let me change case.
Added back nan prefix to function names.
Add Sturla Molden's code for median. URLs below:
Bug fix in original select code; Added negative axis support.
Added median().
Update benchmark code and results.
Remove sum(). Save for a later release.
Replace move_sum() by move_nanmean().
Silence "Warning: invalid value encountered in (blank)".
Add bounds and wraparound checks to group_nanmean().
Older versions of scipy.stats.nanstd might not handle axis < 0.
Add Development section to sphinx doc.
Old versions of scipy.stats.nanstd choke on axis=None.
Increase tolerance for unit test differences between bn.nanstd.
Bump requirements to NumPy 1.5.1+, SciPy 0.8.0+.
Set 0.1 release date to today!
Bump version to 0.2.0dev; add a sandbox with an example.
Faster nanmean; *.pyx files missing; Support older numpy/scipy:
Tiny bit less overhead after removing unused arr.size.
Remove Cython generated C files---repo was getting too big.
Doc update after removal of C source files from repo.
Preparing to improve overhead: add small array to benchmark.
Less overhead in all functions when output is not scalar!
Rename header to func_header to match move and group headers.
First commit of unfinished template code.
More work on unfinished template prototype.
Template of nanmin works!
Rewrite. This time do not template ndim, just dtype and axis.
Added a templated version of median(). Passes unit tests.
Templated move_nanmean(). Passes unit tests.
Clean template code by adding doc strings and renaming.
Merge branch 'template'
Templated nanmax, nanmean, nanvar, nanstd.
Last function, group_nanmean(), templated!
Added support for float32.
Slow non-Cython functions for unaccelerated ndim/dtypes.
Fallback to non-Cython functions for unsupported ndim/dtype.
Rewrote benchmark module.
Removed SciPy dependency.
Added move_nanmean() to benchmark suite.
Added a new function: nanmedian()
Added a new function: nanargmax().
Added new function: nanargmin().
Merge branch 'nanarg'
Shuffle np.arange input array for benchmark. Plus tweaks:
Embed function signatures in docstrings and other doc tweaks.
Update packages list in setup.py.
Set v0.2.0 release date to today!
Bump version number to 0.2.0.
Typo in setup.py.
Bump version to 0.3.0dev; Fix doc typos.
Use integer specific inner loop for move_nanmean.
Added new function: move_min().
Added new function: move_max().
Added new function: move_nanmax().
Added new function: move_nanmin().
Merge branch 'move_minmax'
Bug fix: int input called non-cython version of move_nanmean.
'stdlib' is deprecated, use 'libc.stdlib'.
Ascending minima algorithm copyright correction.
Can now specify dtype and axis in benchmark.
Reuse functions (details below):
Test that no slow functions creep in where they don't belong.
Merge branch 'reuse'
Make benchmark output more compact.
Bug: benchmark with interger dtype.
Added new function: nansum().
Added a new function: move_mean().
Improve behavior of move_min and move_max wrt NaNs:
Added two new functions: move_sum(), move_nansum().
Cut unit test time in half.
Added two new functions: move_std(), move_nanstd().
Doc work.
Made all high-level functions faster. Less overhead.
Less overhead in selector functions. Plus:
Yet another speed up of the selector functions.
Moving window functions now default to axis=-1 instead of 0.
Doc work and bump development status from pre-alpha to alpha.
Set v0.3 release date to today!
Bump version to 0.4.0dev.
All functions faster; less overhead in low-level functions.
Add infrastructure to create 32 and 64 bit OS source releases.
Remove group_nanmean() function.
Added two optional inputs to benchmark: shapes and nans.
Added new function: rankdata()
Make new rankdata function support 32 bit OS.
New function: nanrankdata()
Six functions faster for axis=None:
Bug fixes and unit tests for size zero input arrays.
Make bottleneck work with Cython 0.14.1.
Doc work.
Step release date to today! Bump version to 0.4.0.
Bump development status to beta and version to 0.5.0dev.
Bug fix release 0.4.1. Closes #9.
Bug fix release 0.4.2. Closes #10.
New unit tests to demonstrate median and nanmedian bugs.
Bug: median, nanmedian modified input array. Closes #11.
nanmedian wrong when odd number of elements with all but last a NaN.
Merge branch 'medianbugs'
Lazy import of SciPy speeds Bottleneck import by 3x.
Add scipy.stats.rankdata to bottleneck to avoid dependency:
Set release date to today for 0.4.3!
Release note typo. Year is 2011 not 20011.
Support python 2.5 by importing `with` statement. Closes #14.
Add new function ss(), sum of squares.
Delete ss.pyx and add to .gitignore.
Remove ss.pyx for real this time.
Add partsort() function.
Add argpartsort() function.
Add partsort() and argpartsort() to benchmark.
Merge branch 'partsort'
Fix docstrings: selector functions do not have default axis.
nanmax, nanmin now follow Numpy 1.6 when input is all NaN.
nanmedian wrong for particular ordering of NaN and non-NaN elements.
Run unit tests from Makefile in 4 parallel processes.
Make single source distribution.
Update release notes.
Update release notes.
Update docs.
First commit of move_median(). One unit test failure.
Add .gitignore so that move_median.c gets committed.
Faster move_median for 2d and 3d input.
Edit move_median docstring.
move_median() now passes all unit tests!
Add David's move_median.c code and comment cleanup.
Bump version to 0.5.0beta.
0.5.0beta2: Christoph's bug fixes for windows compilers.
32/64 bit clean up.
Bump version to 0.5.0rc1.
argpartsort, nanargmin, nanargmax returned wrong dtype on win64.
Bump version to 0.5.0rc2.
rankdata and nanrankdata crashed on win64.
Set 0.5.0 release date to today!
Bottleneck 0.6.0dev is open for development.
Merge branch 'buildfixes'
Update pyx and c files when `make sdist`.
Confusing error message in partsort and argpartsort.
readme: bottleneck takes a while to build on old computers.
Update copyright date.
BUG Update MANIFEST.in for new location of Makefile.
Upgrade numpydoc from 0.3.1 to 0.4 to support Sphinx 1.0.1.
ENH Add a replace() function. Closes #38.
ENH Tweak unit tests from previous commit.
DOC Docstring typo. Closes #39.
BUG Wrong output for very large input arrays.
DOC recommend numpy 1.6.1 instead of 1.6.0.
DOC Update release file.
ENH Nearest neighbor of 1d arr0 in 2d arr.
ENH Improve output for all NaNs; Add performance notes.
Merge branch 'nn'
DOC Add See also for bn.nn <==> bn.ss
ENH Add bn.anynan(arr, axis)
ENH Add bn.allnan(arr, axis)
DOC Tweak docstring.
DOC Update
ENH Generate C file with Cython 0.16 instead of 0.14.1
DOC Remove scipy mailing list from readme.
DOC Update readme for numpy 1.6.2, scipy 0.10.1
RLS Bump version to 0.6rc1
BUG bn.slow.replace()
BUG bn.nn() did not take list input.
DOC update readme
DOC remove built-in in <built-in function median...>
RLS Set release date to today; bump to 0.6.0
Open development for 0.7.0
DOC ddof not described in docstring.
BUG `make test` fails on some computers.
BUG nanstd([1.0], ddof=1) and nanvar crash.
REF Tiny speed up
TST Make new tests independent of numpy version.
ENH Add support for tox
BUG Some unit tests depended on scipy.
BUG Docstring and error message out of date
TST tox against multiple python/numpy versions
DOC tweak
DOC Add example; closes #64
TST Add coverage report to makefile
ENH bn.rankdata() is now twice as fast!
REF scipy now has a cython version of rankdata
DOC update benchmark results
TST Bump numpy from 1.7.0 to 1.7.1 in tox tests
ENH Add C files to repo.
DOC updates
BLD Update "make sdist" now that C files included.
REF No need for separate 32 and 64 bit C files.
DOC update release notes
DOC tweak
Merge branch '3264'
Change copyright holder
DOC bug in example
DOC bug in example
Update copyright holder
Add copyright date
TST Change tox test from py32 to py33
STY pep8
STY pep8
WIP toward getting flake8 to pass; Merge branch 'pep8'
STY more pep8 work
DOC update readme for numpy 1.7.1
ENH C files are now generated with Cython 0.19.1
RLS 0.7 release date set to today
RLS bump version to 0.7.0
RLS open 0.8.0dev for development
ENH all tests now pass with numpy 1.8.0rc1
WIP use new numpy nan functions...
DOC compiles with clang; closes #76
ENH nanargmin and nanargmax now match numpy 1.8.0
DOC tweak
REF remove local copies of scipy's nanstd, nanmean
REF remove local copies of scipy's nanstd, nanmean
DOC update benchmark results
REF tox now only tests numpy 1.8
DOC typo
Merge remote-tracking branch 'shoyer/fix-nanargmax' into np18
Update C file
Merge branch 'np18'
DOC update copyright dates
RLS set version to 0.8.0
RLS set release date to today
RLS open 0.9 for development
DOC update release file
DOC tweak readme
ENH update C files
Merge branch 'shoyer-move-ndim'
Correct benchmark output strings
correct outdated code comment
nanmedian template bug causes slowdown; closes #91
WIP rewrite nansum
WIP forgot to add rewrite.pyx to commit
WIP nansum
Remove redundant code
fused types
style tweaks
move iterator creation to main nansum function
rename tweaks
speed up
add missing support for some dtypes
code comments
tweak
revert nanmean.pyx
handle 0d arrays
clean up (hopefully cleaner)
reorg of nansum
WIP remove all functions and templates
move pyx code
WIP integrate new nansum function
alternative reduce code
a third reduce alternative
minor refactor
big reduction in overhead in reduce3
delete reduce design alternatives 1 and 2
rename nansum3 to nansum
remove sandbox
Merge branch 'rewrite_rollout' into rewrite
add TODO list
STY rename variables
STY moved stuff around
handle 0d input arrays
unit tests are running
unit tests now pass after this bug fix
turn on more unit tests
benchmark hacked up to run for nansum
remove `mode` from benchmark suite
import functions in benchmark suite
add optional integer input to reducer
divide into template and auto_pyx directories
WIP hacking a compact templating system
giving up on compact templating for now
WIP template dtypes
WIP working on getting make all to run
`make all` now runs
refactor template
Moved dtypes to function templates
templating seems to work
Merge branch 'rewrite_template' into rewrite
update TODO
change function pointers
refactor done
rename variables
Merge branch 'rewrite_refactor' into rewrite
Added nanmean
handle cython decorators in template
BUG handle axis >= ndim
simplify bn.slow.nansum
handle unaccelerated dtypes
remove slow.nan{sum, mean}; import from numpy
update TODO
add bn.nanstd()
include smaller array in benchmark
update benchmark docstring
add bn.nanvar()
add bn.nanmin()
add bn.nanmax()
refactor benchmark
refactor some imports
update todo list
WIP trying to get move_nanmean to work
bn.move_nanmean builds
bug in bn.move_nanmean
unit test pass for bn.move_nanmean
add bn.move_nanmean to benchmark
Merge branch 'rewrite_move' into rewrite
add bn.move_mean()
`make clean` now remove pyc files
Silence numpy warnings during unit tests
update readme
tweak readme
update readme
`make sdist` and `tox` now work
update release file
update release notes
As suggested by Christoph Deil update release notes
rename func to reduce
silence warnings
Merge branch 'rewrite_reduce' into rewrite
add function signatures to docstrings
update todo
WIP begin working on nonreduce_axis
WIP nonreduce_axis
WIP more hacking
WIP partsort
WIP partsort compiles
BUG negative input axis
BUG infinite loop
BUG hook up unit tests and get them to pass
add partsort benchmark
split slow.reduce into reduce and nonreduce_axis
update readme
Merge branch 'rewrite_nonreduce' into rewrite
update release notes
update todo list
tweak readme
add unit test to demonstrate bug in partsort
partsort bug fix
update readme
Merge branch 'rewrite_partsort_bug' into rewrite
update todo list
add bn.argpartsort()
readme typos
update todo list
WIP nonreducer and bn.replace()
bn.replace compiles
bn.replace unit tests now run but fail
bn.replace unit tests now pass
clean up
add bn.replace() to benchmark
Merge branch 'rewrite_reduce' into rewrite
remove unused imports in nonreduce.pyx
split slow.reduce into reduce and nonreduce
add to release notes
bn.nn() is now completely removed
turn on 4 more unit tests
add bn.move_sum() and bn.move_nansum()
less verbose benchmark
tweak benchmark output format
add bn.move_std() and bn.move_nanstd()
copy move_median c code from old bottleneck
move_median compiles but dumps core
move_median no longer core dumps
bn.move_median unit tests pass
add bn.move_median() to benchmark
forgot to include move.c updates
Merge branch 'rewrite_move_median' into rewrite
turn on two more unit tests
bug reported by @Midnighter
add move_max and move_nanmax
add bn.move_min() and bn.move_nanmin()
remove unused move.pyx imports
WIP start converting move_sum to using nmin
unified bn.move_sum
simplify move_sum
simplify move_sum
more extensive moving function unit tests
clean up slow.move_sum
slow.move_sum speed up
check for ndim > 0 in slow.move_sum
clean up fast.move_sum
unify bn.move_mean
unify move_std
unify move_min
unify move_max
bug fix in function signatures
add unit tests
get benchmark working
rename nmin to minc for min count
As suggested by @shoyer rename minc to min_count
change default min_count from -1 to None
Merge branch 'rewrite_move_nmin' into rewrite
update todo list
added bn.rankdata and bn.nanrankdata
shorten and update readme
remove unused imports
add bn.ss(), sum of squares
update slow.reduce docstrings
update todo list
add median and nanmedian
add nanargmin and nanargmax
add unit tests for 0d input; fix bugs
add bn.anynan()
refactor bn.anynan()
add bn.allnan()
update copyright year
Rename unit test modules; move around some tests
import docstrings from master branch
edit docstrings of reduce functions
reduce functions limited to arr.ndim < 12
update function reference in sphinx manual
add nanvar to benchmark
remove unused imports
remove unused imports
make rankdata benchmark more fair
use np.partition in bn.bench and bn.slow
moving window docstring
update moving window function docstrings
update copyright
rewrite and expand release notes
edit release file
remove todo file
tweak release file
tweak unit tests
Make sure unit tests do not display numpy warnings
Merge branch 'rewrite'
update copyright
fix sphinx build warnings
edit release notes
set release date (bn 1.0.0) to today!
open 1.1.0 for development
Merge branch 'welford' of https://github.com/jaimefrio/bottleneck into jaimefrio-welford
int array can't have NaN; upgrade to Cython 0.23.4
update move_std doc string
add unit test
update docs
Merge branch 'jaimefrio-welford'
upgrade to numpy 1.10.4
Merge branch 'movemedian' of git://github.com/jennolsen84/bottleneck into jennolsen84-movemedian
add back merge-conflict move.c
int arrays can't contain NaN
uncomment commented out functions
test move_median with NaNs
update move_median docstring
update docs
Merge branch 'jennolsen84-movemedian'
np 1.10.4 upgrade added numpy warning to bn.bench
add byte swap unit tests; lots of failures
unit test failures fixed
more unit tests and corresponding fixes
another unit test and failure fix
update docs
Merge branch 'byte_swap'
new function: bn.move_var
copy and paste errors in docstring
Happy New Year! Update copyright date
update c file after updating docstring
upgrade to numpy 1.11.0
check that PyArray_FillWithScalar completes
Added new function bn.move_argmax
Merge branch 'move_argmax'
add new function bn.move_argmin
Thank those who helped
add helper functions for running with threads
release GIL for reduce functions
Class (Threads) to help run multi threaded
Coding style tweaks to Thread class
release GIL in move functions
Release GIL in remaining functions
Add release note
Merge branch 'nogil'
prototype of push function...
bn.push can now handle 0d input arrays
Add bn.push to benchmark suite
Complete work on bn.push
Merge branch 'push'
upgrade to cython 0.24
automate updating of benchmark results in readme
Added unit tests for scalar input
WIP prototype of move_rank; failing unit test
all unit tests pass
Docstring and benchmark
bn.move_rank docstring tweaks
Merge branch 'move_rank'
Can now specify which functions to benchmark
median now returns NaN for slice that contains NaN
Made unit tests easier to maintain
bottleneck is now flake 8 clean; long may it live
Bug fix: bn.slow.move used min_count incorrectly
Simplify moving window unit tests
Clean up unit tests of reduce functions
Improve moving window unit tests...
refactor move_median.c
rename size in move_median.c to window to...
Now no lines over 80 characters in move_median.c
bn.move_median is 15% faster
Split move_median.c into no-NaN and NaN functions
more refactoring of move_median.c
pull out NaN handling code from for non-NaN...
move_median.c refactor and doc work
more refactoring of move_median.c; unit test
move_median.c: inline helper functions
speed up move_median.c by changing variable type
remove duplicate unit test
more refactoring of move_median.c
speed up for int input in move_median.c
code style tweaks in move_median.c
remove unused input parameter in move_median.c
rename variables in move_median.c
code stype tweaks in move_median.c
mm_dump (move_median.c) can now print binary heaps
rename variables in move_median.c
move_median.c: remove variable; edit comments
bug fix in move_median.c
move work on move_median rewrite from...
integrate new move_median code into bottleneck
some code cleaning of move_nanmedian.c
reuse code in move_nanmedian.c
bug fixes in move_nanmedian.c
more unit testing of move_nanmedian.c
bug fixes in move_nanmedian.c
bug fixes in move_median.c
move_median.c passes all unit tests!
clean code in move_nanmedian.c
change region back to an int in move_nanmedian.c
allocate less memory in move_nanmedian.c
Complete rewrite of bn.move_median for nan input...
update benchmarks for move_median rewrite
update release file
complete rewrite of bn.move_median
Be more selective in RELEASE.rst
Be even more selective in RELEASE.rst
some code cleaning of move_nanedian.c
move_nanmedian.c: cleaner and faster
remove unused lines in move.pyx
combine move_median.c and move_nanmedian.c
rename move_nanmedian.c to move_median.c
pull out move_median.h and move_median_debug.c...
clean up and speed up of move_median.c
move_median.c: cleaner and faster
bug fixes in move_median.c
bug fixes in move_median.c
include headers in sdist
py3 compatability
possible optimization of move_median.c
fix 64-bit OS unit test failures
no travis email on failures
silence numpy warnings on py3 during unit tests
specify numpy version in .travis.yml
trying to unbork .travis.yml
add back csrc directory to sdist
Merge branch 'toobaz-single_setup_py'
update readme, release, makefile
add `make doc` to makefile
flake8
tweak readme
force '-O2' compiler flag
No speed penalty for median, nanmedian...
Add travis build status to readme but not...
gitignore doc/source/intro.rst
template to create reduce2.pyx
first pass at nansum reduce all...
bn.nansum2 now passes all unit tests
tweaks to reduce2 template
nansum2 refactor
Add nansum3
optimize nansum3 for 1d input
code clean up of nansum3
code clean up of nansum3
add nanmean3
added nanstd3
added nanvar3
added nanmin3 and nanmax3
added ss3
added median3
added nanmedian3
added nanargmin3 and nanargmax3
added anynan3 and allnan3
remove reduce2
update benchmarks without using reduce3
switch to using reduce3
update benchmark
Merge branch 'reduce_all'
make more use of numpy c api
clean up setup.py
multi-thread tools not yet ready for release
change NAN from float to ai_t in move_median.c
Merge branch 'cgohlke-patch-1'
set bottleneck 1.1.0 release date to today!
open bottleneck 1.2.0 for development
port a reduced featured nansum to pure C...
bug fix of nansum in reduce2
remove reduce2 build warning
remove variable not needed in reduce2
do not release GIL for 1d,float64 in nansum
custom arg parsing for nansum in reduce2.c
reuse arg parsing code in reduce2.c
handle more variety of inputs (nansum reduce2.c)
clean up of reduce2.c
pull out arg parsing code to separate function
clean up of reduce2.c
add support for more dtypes in reduce2.c
add reduction over a single axis in reduce2.c
add support for 0d input arrays in reduce2.c
add support for calls to bn.slow in reduce2.c
nansum2 runs on py2.7 with 1 test failure
nansum2 passes all unit tests on py2.7
add py3 support to reduce2.c
fix py3 compat bugs in reduce2.c
remove most function prototypes in reduce2.c
move slow() from reduce2.c to slow.c
include slow.h in sdist
unit test argument parsing
release GIL in reduce2.c
move new c rewrite code to src/
create bottleneck.h for reuse of code
added templating of dtype in reduce2.c
add nanmean to reduce2.c
code clean up in reduce2.c
add templating of docstrings
make templating more efficient
add benchmark of c rewrite
optimize dtype templating
optimize slow.c
handle runtime errors in slow.c
add function signatures to docstrings in reduce2.c
add C version of nanmin
forgot to release GIL
coding style tweaks
add C version of nanmax
rename variable
lower function call overhead...
code cleaning
refactor bench2 benchmark
declarations at top of blocks for windows support
use NPY_NAN, NPY_INFINITY, NPY_INLINE for compat
macros for setting raise text
copy numpy's nan, infinity, inline macros
add C version of move_sum
add move_sum to bench2 benchmark
add more unit tests of move_sum
refactor move_sum for speed
set keyword defaults in one place
refactor move_sum
add c version of move_mean
use macros for common code
cating not needed
typo in sphinx doc
add C version of move_std
add C version of move_var
DECREF with GIL
more unit tests
add to bench2 benchmark
I hope this makes the code easier to read
move2.c no longer uses numpy iterators
refactor moving window iterators for...
doc of moving window interators
remove numpy iterators from the most common...
remove RETURN macro
removal of last use of numpy iterators
lots of macros
more efficient NEXT iterator macros
code tweaks
more macros
faster iterators (NEXT macros)
simplify reduction function; slow down
reduce number of lines of code
performance bug
code tweaks
optimizations for int input arrays
faster move_mean for float input
faster move_std, move_var but higher overhead
speed ups
code tweaks
macros
simplify move.c
bug fixes
undo accidental commit
simpler arg parsing and prepare for nanstd
make all2 only builds C not Cython
add nanstd
underscore prefix for internal iterator variables
add C version of nanvar
add C version of anynan
add C version of allnan
add C version of ss
add C version of nanargmin
special case contiguous arrays
redesign of nanargmin
add C version of nanargmax
add C version of median
bug in unit tests; sorted input array does not...
refactored median
add C version of nanmedian
underscore prefix for internal iterator variables
add C version of move_min
rewrite of move_min
add C version of move_max
add C version of move_argmin
add C version of move_argmax
add C version of move_median
trying to get bn to build on osx without -std89
replace // comments with /* */
bug fixes for C version of move_median
hide details in macro
add C version of move_rank
some optimization of move_rank
coding style of templating code
use more macros
macros! macros! macros!
more templating
tweak templating
more unit testing
add messy C version of partsort
fix unit tests for partsort and argpartsort
py3 compat fix
use macros
optimize median, nanmedian
more optimizations of median, nanmedian
clean median, nanmedian macros
speed up of median, nanmedian
fix memory leak in median
faster nanmedian
faster median and reuse code
faster partsort; reuse code
add C version of argpartsort
update copy of scipy's rankdata function
simplify bn.slow.rankdata and bn.slow.nanrankdata
reuse code in move_rank
begin redesign of bottleneck iterators
move new iterator to separate file
tweak design of new iterator
minor code tweak
expand use of new iterator
move iterator creation to reducer
hide details of iterator in macros
add iterator for reduction along all axes
clean up iterator code
speed up iterators by removing malloc
iterator tweaks
accelerate reduction of 0d arrays
code tweaks
use new iterator in move.c
partial clean up of iterators in move.c
redesign how iterators are used in reduce.c
clean up iterator macros
faster breaking out of nested loop
windows compat: declare at top of blocks (C89)
simplify signature of low-level move functions
use new iterators
add some comments for iterators
Merge branch 'iterators' into c_rewrite
reduce build output; closes #134
add code comment
move slow() function to bottleneck.h
add C version of rankdata
simplify function signatures
can now use multiline dtype templating info
add C version of nanrankdata
bug hunting is painful
add C version of push
optimize common use case
add more unit tests of nonreduce_axis functions
MSVC (C89) compat. See issue #142
doc
move move_median code to src directory
add C version of replace
bug fix; closes #145
macro mania
macro tweak
WIP bench_detailed
bn.bench_detailed runs
refactor bench_detailed
bench_detailed can now optionally include NaNs
add docstring for bench_detailed
Merge branch 'bench_detailed' into c_rewrite
first pass at removing cython; tests pass
second pass at removing cython
third pass at removing cython
remove cython from travis
cython completely removed
Merge branch 'no_cython' into c_rewrite
update MANIFEST.in
don't include tools directory in sdist
update travis icon
update docs
Bottleneck is now written in C.
remove ci directory
BUG replace arr with a; closes #131
rename (arg)partsort to (arg)partition
refactor and add more input arrays: bench_detailed
add new feature to bench_detailed
attempt to remove compiler warnings...
build bottleneck before creating sphinx doc
flake8 test more bottleneck files
try installing bottleneck before numpy
setup requires numpy
install bottleneck before numpy
can't import numpy before installing it
remove one travis test
update release file
Merge branch 'no_numpy'; closes #141; closes #140
remove unused variable
use positional args when benchmarking functions
appveyor integration
fix appveyor badge
change default arrays in benchmark
bigger arrays, fewer NaNs in benchmark
remove a few lines not needed in setup.py
code style tweaks
docstring tweak. See #152.
use osx 10.10
new transpose unit tests give 4 failures
bug fix; closes #154
Merge branch 'transpose_bug'
upgrade to numpy 1.11.2; closes #149
workaround attempt for xcode 7
do not pin xcode
workaround for travis issue
syntax error
rewrite of unit tests plus more inf tests
add more test arrays
Lev Givon (6):
Rebase.
Use setuptools, include ez_setup, enable installation of numpy by pip if it isn't installed.
Merge branch 'master' of https://github.com/lebedov/bottleneck
Fix extension package name.
Add PSF and ZPL licenses (for ez_setup.py) to bottleneck/LICENSE.
Update RELEASE.rst.
Midnighter (12):
add continuous integration using travis
remove lsb_release since not available for osx
ignore unset variables
make activating the conda environment less verbose
actually need to install the bottleneck
include dependency `nose`
correctly handle multiple dependencies in script
ensure that the dependencies string is split
apply correct split without quotes
make cosmetic modifications
add flake8 check and source distribution
reduce number of test cases
Moritz E. Beber (15):
add AppVeyor support to bottleneck
rename manager to wrapper
add testing after install
fix quotes
move commands out of global env variable
make flake8 formatting fixes
play with conda force 32bit
set force 32bit as environment variable
comment full section
check for OS architecture
print python location and int size
correct quotes
get more system information
adjust miniconda use for platform
clean up and remove debugging output
Pietro Battiston (15):
Delete obsolete files
Moved template files to bottleneck/template/
Updated paths for generated .pyx files
All generation now happens in setup.py
Also ignore generated C files
Move csrc to bottleneck/, delete src/
Do not build Cython extensions on build_sphinx
doc/Makefile is not needed
Updated MANIFEST.in
Require cython to run tests
Pass np.get_include() to cythonize too
Try to satisfy Travis with include_dirs...
"-std=gnu89" for "bottleneck.move"
Put generating code back in template/template.py
Reintroduce (updated) Makefile
Stephan Hoyer (12):
BUG: Fix nanargmax/nanargmin for 1D integer arguments
Generalized move_func_strides to arbitrary dimensions
WIP: fast functions for N-dimensions
General template machinery to insert for statements
Add option to skip ndim=1 for axis is not None
Switch all moving window functions to N-dimensions
All functions are now N-dimensional
Fix typo in template.py
Switch to use 'ndims' instead of 'skip_1d' in templates
Turn back on use of nogil when appropriate
Use nogil more aggressively (including all moving window functions)
Update README and RELEASE notes with NDIM_MAX
jaimefrio (1):
Broken link in comment
jennolsen84 (1):
updated docs
jmcloughlin (1):
BUG move_std and move_nanstd neg sqrt bugs fixed
kwgoodman (12):
Merge pull request #44 from dougalsutherland/py3k
Merge pull request #51 from jmcloughlin/dev
Merge pull request #59 from mrjbq7/macosx
Merge pull request #73 from cgohlke/patch-1
Merge pull request #81 from shoyer/better_move_func_strides
Merge pull request #79 from lebedov/master
Merge pull request #98 from jaimefrio/broken_link
Merge pull request #108 from alexamici/master
Merge pull request #113 from Midnighter/travis-ci
Merge pull request #120 from Midnighter/source-dist
Merge pull request #147 from Midnighter/appveyor
Merge pull request #148 from Midnighter/build_arch
-----------------------------------------------------------------------
This annotated tag includes the following new commits:
new 28e711b add flake8 check and source distribution
new 3f83f8d reduce number of test cases
new 1b41e0d Merge pull request #120 from Midnighter/source-dist
new b1871d1 flake8
new bba85d5 tweak readme
new b3e6182 force '-O2' compiler flag
new d0828ef No speed penalty for median, nanmedian...
new 398ede3 Add travis build status to readme but not...
... 288 lines suppressed ...
--
Alioth's /usr/local/bin/git-commit-notice on /srv/git.debian.org/git/python-modules/packages/bottleneck.git
More information about the Python-modules-commits
mailing list