Commits on Dec 27, 2010
  1. Typo in

    committed Dec 27, 2010
  2. Bump version number to 0.2.0.

    committed Dec 27, 2010
  3. Shuffle np.arange input array for benchmark. Plus tweaks:

    Instead of all NaNs, use 1/3 NaNs. Remove nanmin (nanmax
    is enough). Rename geta function to getarray.
    committed Dec 27, 2010
Commits on Dec 24, 2010
  1. Merge branch 'nanarg'

    committed Dec 24, 2010
Commits on Dec 17, 2010
Commits on Dec 16, 2010
  1. Removed SciPy dependency.

    committed Dec 16, 2010
  2. Rewrote benchmark module.

    committed Dec 16, 2010
Commits on Dec 14, 2010
  1. Fallback to non-Cython functions for unsupported ndim/dtype.

    Instead if raising an error for ndim/dtype combinations for
    which there are no Bottleneck cython function, call the
    non-cython version of the function.
    committed Dec 14, 2010
  2. Slow non-Cython functions for unaccelerated ndim/dtypes.

    Instead of raising an error on unaccelerated combinations
    of ndim/dtype, I plan to pass the work off to slower,
    non-cython versions of the functions. The functions are
    in this commit. They have not yet been hooked up to
    receive the unaccelerated ndim/dtypes combinations.
    committed Dec 13, 2010
Commits on Dec 13, 2010
  1. Added support for float32.

    committed Dec 13, 2010
Commits on Dec 11, 2010
Commits on Dec 10, 2010
  1. Merge branch 'template'

    committed Dec 10, 2010
Commits on Dec 9, 2010
Commits on Dec 8, 2010
  1. Rewrite. This time do not template ndim, just dtype and axis.

    Previous version was nice since it automated the for loops
    for arbitrary ndim. But it only worked for a particular
    format of for loops. Not all functions use that format.
    So in this rewrite, I only template dtype and axis. Loops must
    be hard coded.
    committed Dec 8, 2010
  2. Template of nanmin works!

    committed Dec 8, 2010
Commits on Dec 7, 2010
Commits on Dec 6, 2010
  1. Remove Cython generated C files---repo was getting too big.

    Every time I modify a cython pyx file, the corresponding
    change to the auto-generated C file is huge. The repo is
    growing in size quickly. So I am removing the C files.
    I will include them in each release. Closes #4.
    committed Dec 6, 2010
Commits on Dec 2, 2010
  1. Tiny bit less overhead after removing unused arr.size.

    The function selectors for nanmean, nanstd, and nanvar
    all had the unused line:
    cdef int a = a.size
    committed Dec 2, 2010
  2. Faster nanmean; *.pyx files missing; Support older numpy/scipy:

    Faster nanmean() for 2d, 3d arrays with NaNs when axis is not None.
    move, group pyx and files missing from source distribution.
    Added support in bn.benchit() for numpy 1.4.1/scipy 0.7.2
    Closes #2. Closes #3
    committed Dec 1, 2010
Commits on Dec 1, 2010