Deprecate PyArrayObject* direct field access #116

Merged
merged 25 commits into from Jul 26, 2011
Commits on Jul 19, 2011
  1. ENH: core: Deprecating direct access to the PyArrayObject fields

    mwiebe committed with Mark Wiebe Jul 12, 2011
    NOTE: WIP, code doesn't build
  2. ENH: core: Work in progress making things work without direct field a…

    mwiebe committed with Mark Wiebe Jul 13, 2011
    …ccess
  3. ENH: core: More progress refactoring code to not use PyArrayObject fi…

    mwiebe committed with Mark Wiebe Jul 13, 2011
    …elds directly
  4. ENH: core: Some fixes, change some tests to not use yield

    mwiebe committed with Mark Wiebe Jul 15, 2011
  5. ENH: core: More cleanup towards deprecating direct arrayobject field …

    mwiebe committed with Mark Wiebe Jul 16, 2011
    …access
  6. WRN: nditer: remove some warnings

    Mark Wiebe committed Jul 18, 2011
  7. ENH: core: Got the tests running after the ArrayObject field access d…

    Mark Wiebe committed Jul 18, 2011
    …eprecation
  8. BUG: dtype: comma-list dtype formats didn't accept M8[] parameterized…

    Mark Wiebe committed Jul 18, 2011
    … dtypes
  9. BUG: core: PyArray_GetArrayParamsFromObject was treating __array_inte…

    Mark Wiebe committed Jul 19, 2011
    …rface__ incorrectly
    
    I'm not sure why this came up after the arrayobject field access
    changes, since it looks like the bug was in there before already.
  10. TST: rec: DType in join_by test was inconsistent

    Mark Wiebe committed Jul 19, 2011
  11. BUG: core: Needed to initialize 'obj' to NULL for Py_XDECREF_ERR in f…

    Mark Wiebe committed Jul 19, 2011
    …ail section
  12. BUG: mmap: Make the memmap subclass rely on the Python mmap's destruc…

    Mark Wiebe committed Jul 19, 2011
    …tor being correct
    
    There was some fishy code flushing and closing the _mmap property,
    which looked as if it was paranoid in not trusting mmap.mmap to
    behave properly. This was done in a way which assumed the .base
    attribute isn't collapsed, something which I've changed in the refactoring.
    
    The easiest way to fix this is to trust mmap.mmap - if this is incorrect,
    the specific platform on which this fails should have been already
    documented in the comments!
  13. ENH: core: Add access macros back in conditionally, for backwards com…

    Mark Wiebe committed Jul 19, 2011
    …patibilities
    
    Sticking with inline functions when NPY_NO_DEPRECATED_API is defined.
Commits on Jul 20, 2011
Commits on Jul 22, 2011
  1. ENH: dtype: Make handling of struct dtype align= flag more rigorous (…

    Mark Wiebe committed Jul 22, 2011
    …also fixes ticket #1912)
    
    This adds an 'aligned'= property to the format dict, which gets put
    in the str() representation when necessary to preserve it.
  2. ENH: core: Rename PyArray_SetBase to PyArray_SetBaseObject to be more…

    Mark Wiebe committed Jul 22, 2011
    … clear
  3. DOC: core: Document the PyArray_SetBaseObject function

    Mark Wiebe committed Jul 22, 2011
Commits on Jul 26, 2011
  1. STY: Remove trailing whitespace

    Mark Wiebe committed Jul 26, 2011