Wed, 28 Jul 2010 12:17:27 +0900, David Cournapeau wrote: [clip]
I put a new repostory (same location)
Compared this against git-svn produced repository. There are a number of commits missing from the early history, apparently because numpy trunk was moved around in SVN.
Corresponding log messages from a diff between `git log --oneline|sed -e 's/^........//'` in the two branches:
... Alter shape setting. Moving newcore to trunk -Added f2c.h to depends. -Fix array scalars __array_struct__ interface -Added Masked Array tests. -Minor memory leaks -Fix array_structinterface bug -Added Paul's enhanced masked arrays. -Fix up bad commit. -Changed Transpose C-API. Cleaned up memory leaks. -Beginnings of faster scalar math. -Fixed put and putmask functions to not work for nonarrays. -Switched order of put and putmask. -Fix int stop -> intp stop in parse_subindex -Fixed up slice parsing to support larger integers on 64-bit platforms. -Created a better Copy function that can copy FORTRAN order too. -Fixed another leak. -Fixed-up array interface. -Comment changes. -Remove PyArrayMapIter_Type from user space. -Updated FillWithScalar. -removed MapIter commands from C-API -More valgrind-found bugs. -valgrind-found fixes and adjust indexing for object arrays. -11/4/05 -- CJH -- Moved variable declaration to make it consistent with -Fixed some type issues. -Fixed back the test to cover all of scipy. -Added flexible support for Choose. -Found and fixed Put Object bug. -Disable problem test. -Small fixes for Intel Compiler class - tested and builds correctly -Raise error for iteration over scalar. -Fixed problems with non-existent attributes in array interface. -Added Intel C Compiler class. -Comment changes, and change read to write in PyArray_Dump -Fixed bug in getfield, and some consistency issues. -Added ability to create arrays of user-defined type using the typeobjec -Altered flexible types so that NotImplemented is raised in case another -Updated PyArray_Zero and PyArray_One. -Add missing variable declaration -Fixed some array interface issues. -Comment cleanup. -CJH -- 11/2/2005 -- Modified the nbytes dictionary to return values in -added support for __array_struct__ -Added nbytes attribute and nbytes dictionary -Fixed Concatenate to use new ArrayType interface so flexible arrays are -Update Blitz++ included with weave to 0.9 (which was released 2005-10-1 -Update Blitz++ included with weave to 0.9 (which was released 2005-10-1 -Reapplied patch to interpret sequences more cautiously. -Reverted the PySequence_Check patch. DOK matrices should be fixed inst -Changed array_fromobject() to treat failure of the Array_FromSequence() - r723@Blasphemy: kern | 2005-11-01 01:07:20 -0800 scipy_distutils -> -Added ISALIGNED macro and documentation update. -Fixed to allow reflected operators instead of OBJECT addition. - r707@803636f3: kern | 2005-10-31 13:19:55 -0800 Fix put() typo -Added more attribute access to flags object. -Allow flags dictionary attribute access -Added bool8 type. -Fixed small typo -Fixed incorrect usage of PyArray_CanCastSafely. -Save and restore sys.displayhook around running tests, so that doctests -Don't use deconvolve for polydiv, as that requires scipy.signal to be i -Don't try to handle ImportError in deconvolve when scipy.signal doesn't -Cosmetics with NO_SCIPY_IMPORT. -Fixed nx -> NX -Change to using PY_ARRAY_TYPES_PREFIX instead of PYARRAY_TYPES_PREFIX, -Make it easier for extension writers. If PYARRAY_TYPES_PREFIX is define -some clean up for scimath.py, use type objects instead of typechars -small whitespace -Convert polynomial to use type objects instead of old Numeric typechars -TLC for polynomial.py: cleanup, poly1d class is now new-type, and unitt -Give ufunclike some TLC. Cleanup, add docstrings and unit tests. The un -Whitespace cleanup and make classes new-type -Altered vectorize to allow default arguments in function. -More cosmetic changes. -Added PyUFunc_On_Om to UFunc C-API -Some refactoring - poly1d is a new-style class now - no import * now - -Replaced PyArray_FromDimsAndData with PyArray_SimpleNewFromData. -Cosmetic changes mainly. Remove unused itemsize from scanfunc arguments -Moved f2py2e to f2py. -change PyArray_Scalar to take a void * instead of char *, and some whit -Change PyArray_Scalar to take a void * instead of char * -Whitespace cleanup -Add __all__ statements to base.{oldnumeric,numeric,numerictypes} This s -Fixed SimpleNewFromData to use correct flags. -Fixed complex to bool issues. -Introduced PyArray_SimpleNewFromData as a replacement to PyArray_FromDi -Removed unused requirement that compare deal with misbehaved arrays. -Small cleanups in arrayobject.h -Altered default flags in PyArray_BufferConverter and fixed signbit and -Fixed problem with correlate. -Fix 0-d array indexing. -Fix for rank-0 arrays. Now a[0] = 1 throws an IndexError, just like re -More docstring fixes for mean() and sum(). -Trivial changes to a.mean() docstring -Improved docstrings: * Fuller explanation of default casting behaviou -Fixed get_lib_source_files bug. -Made ScalarType tuple so that it could be 2nd argument to isinstance. -Fixed assert_approx_equal(0,0) failure bug. -Fixed stepping issue for length-1 arrays. -Fixed problem with temporary allocation in PyArray_Transpose. -Bug fixes. -Fixed tests on opteron -Fixed intent(in,out) array shape when copy is made. -Setting __scipy_version__. -Finished impl. of new array_from_pyobj. Updates and clean up. Removed a -Fixes to array_ass_item and array_item so that other uses of these besi -Cosmetic fixes. -More weave fixes. -Added more types to weave. -Fixed string_from_pyobj to handle multi-length string conversion. -Fix segfault with Character arrays. -Fix toscalar so it only works when a.size == 1 -Reimplemented array_from_pyobj using scipy.base CAPI. -Unittests for new array_from_pyobj in f2py. -Fixed default typecode on copy. -Fixed docstring (line-break problem) -Bug fix and new test cases for average() -Cleaned up exceptions: * improved consistency in exception strings; re - r583@Blasphemy: kern | 2005-10-23 17:04:42 -0700 Try to mask key wit -More fixes to remove allowing numbers for types. -Removed allowing numbers for typecodes. -EJS: disabled the stricter type checking on boolean values for now, unt -EJS: Patch to use proper True and False values rather than 1 and 0. Th -Fixed bug in CDOUBLE_setitem -added common_type -Remove useless size change. -Fixed memory alignment issue in frompyfunc -Added a new test for isinf. -Fixed up array interface -Make sure errorobj is NULL. -Applied Ted Horst's patches. -Tried different ones algorithm. Wasn't faster. -Added .fill method. -Set memory to zero when using SimpleNew -Changed lookup to a single list. -Silly error in _ScalarAsCtype. -Transpose fixes --- flags needed fixing before calling copy_ND_array. -Changed default rtype for Bool and low-bit integers -Replaced PyArray_FromDimsAndData with PyArray_New. 2nd try. -Replaced PyArray_FromDimsAndData with PyArray_New. -Fix flags and don't set mem to 0 in FromDimsAndData -Changed default on matrix to copy=1 -Allow array scalars in f2py conversions. -Clean up, updated initforcomb for 64 bit platform. -Impl. finfo as a new-style singleton class. Added singleton tests. -Dont define ushort,uint ulong when _BSD_SOURCE. -Fixed intp formats. -Fixed INTP_FMT when SIZEOF_PY_INTPTR_T == SIZEOF_LONG. -Fixed intp formats. -Fixed typedef, part of int->intp changes. -Fixed missing type. -Using PyArray_ITEMSIZE instead of ->descr->elsize -Changed int to intp for array dimensions and strides. -Changes to fix int -> intp -Removed character conversion from convertcode. -Added NO_IMPORT to list of workable defines. -Clean up -- removed Numeric and numarray specific code. -If env.variable NO_SCIPY_IMPORT is defined then importing scipy package -PyArray_API management copied from Numeric. -scipy include_dirs are set by scipy.distutils, no need to do that in f2 -Fixed segfaults from f2py generated modules: fortranobject needed impor -Fixed bug: items in data statement may be separated by commas. -Generalized newshape for adding and removing ones. -Fixed so that setting shape works even for discontiguous arrays as long -Additional support for gnu compiler on 64-bit platform. -Fixed return_integer tests. - r409@Blasphemy: kern | 2005-10-14 20:26:04 -0700 Comment out unused - r391@Blasphemy: kern | 2005-10-14 19:54:35 -0700 Fix RandomState.ran -removed % from FMT string -Impl. hooks to import scipy packages with info.py files. -Fix remaining compile warnings (except for one) in mtrand by using a po -Support for setuptools, if found. You can use python setup.py bdist_egg -Few more fixes for compiler warnings for mtrand. I'm still getting 'war -Cleaned up a few things in arrayobject.h -Initial hooks for package imports. -Doc updates and initial hooks for scipy imports. -Fixed the installation path for __svn_version__.py. -Updated f2py/weave setup.py files. -Clean up. -Fixed more add_data_dir/add_data_files bugs. -Fixed __array_data__ interface. -Removed mapiter from Python exposure. -Fixes made necessary by change to asarray. -Fixed subclass PyArray_New error segfault. Updated f2py for newscipy ( -Fixed can cast safely. -Fixed so that array always returns an array object. -Fixed typo. -Regenerated mtrand.c with a version of Pyrex which emits code that comp -Fixed typo. -Enabled scipy.distutils tests. -Dropped install_init_py file and generating __core_config__.py in order -Dropped install_init_py file and generating __core_config__.py in order -Made svn version available to main setup.py script. -Added initial scipy.basic.tests site. -Renamed test_limits to test_getlimits. -Renamed test_basic to test_umath. -Shut up scipy.test(..) on missing tests. Use scipy.test(verbosity=2) to -Fixed getting base svn version number - avoiding importing base while i -Fixed getting base svn version number - avoiding importing base while i -Moved importing scipy modules from __init__.py to install__init__py. -Updated thank yous.. -Added more doc-strings. -change floor precision of small-bit integers -Added mingw32ccompiler and lib2def to support building with Python 2.3 -Fixed remainder to behaved like Python. -Fix buglet in set_numeric_ops -Added tuple argument support to add_data_dir. -Moved back to 0.4.2 -Moved doc to subdir of scipy -Updated version after beta tag -Fixed memset to 0 in dotblas. -Fixed memset to 0 in dotblas. - r328@Blasphemy: kern | 2005-10-11 19:50:41 -0700 matrix.py fixes and - r327@Blasphemy: kern | 2005-10-11 19:49:52 -0700 Missing self. -Fixed getting things from scipy.linalg -cookedm doing test commit -Added new user. -Fixed flags problem on array_interface conversion. -Removed comments. Fixed can_cast so that 64-bit integer can be cast sa -Fixed long double unary func problem. -10/11/05 -- CJH -- Commented out diagnostic print statement. -Added ndenumerate. Some print statements for debugging on 64-bit -test commit -test change -Fixed typo -Print statements for 64-bit debug. -Fixed scalar stuff again. -Infinite recursion for long doubles. -Removed deleted mtrand routines from scipy/corelib/setup.py and fixed r - r300@Blasphemy: kern | 2005-10-09 09:54:11 -0700 Remove dependency o -More fixes for 64-bit systems. -Fixed loop->bufcnt for 64-bit fixes. -Fixed off-by-one at end problem with fromfile text files. -Added comments -Fixed backwards compatibility with Numeric on 64-bit machines. -Fixes to get core version working. -Fixed typo and moved __svn_version__ for core -More information on flags... -Added a blank README file for rpm -Improve the array docstring a bit. -Fix use of PyArray_New in PyArray_FromDims -Some printing fixes. -More 64 bit fixes. -Fixed some typecode initilization warnings. Made core libraries access -Minor changes -Moving DISTUTILS.txt to newcore/doc. -Moved extra modules to basic -Minor formatting changes. -Fixed formatting. -Fixed up one change in PyArray_New -Added some simple CAPI documentation. -Fixed more intp issues. - r206@Blasphemy: kern | 2005-10-07 03:12:29 -0700 Okay, for real this -Finished delting random. -Deleted random package. - r201@Blasphemy: kern | 2005-10-07 02:46:04 -0700 Don't need this any -More 64-bit fixes. -More checks for SIZEOF_LONGLONG==SIZEOF_LONG - r188@Blasphemy: kern | 2005-10-07 02:37:24 -0700 Move scipy.stats to -Fixed import problem. -Added Inf to base -Some more fixes -Fixed missing intp's -Get rid of rfftfreq and add random. -Trial run of random... -Some f2py fixes -Fixed up add_library for a bit. -Fixed up fftpack a bit. Don't want to conflict with scipy's use of rff -Added helper.py from scipy. - r165@Blasphemy: kern | 2005-10-06 22:02:09 -0700 More distributions! - r164@Blasphemy: kern | 2005-10-06 22:01:39 -0700 Typo -Fixed vectorize segfault... -Fixed issues with threading... -Added optimization for array_subscript and eleminated threaded support -Added more COMPATIBILITY info. -Added compatibility information -Fixed tests. -Fixed some BOOL-related bugs. BOOLs are now cast to their correct type -Fixed remaining problems with casting buffer sizes.. - r107@Blasphemy: kern | 2005-10-05 04:21:41 -0700 Activate apply_alon - r104@Blasphemy: kern | 2005-10-05 04:20:16 -0700 Fix apply_along_axi - r88@Blasphemy: kern | 2005-10-04 07:03:42 -0700 Some fixes to code a -Added feature that output arrays less than bufsize make temporaries... -Fixed some more bugs -Moved inclusion of files from MANIFEST.in to proper setup.py files. -Fixed object put bug.. - r85@Blasphemy: kern | 2005-10-04 03:56:27 -0700 Ahem - r84@Blasphemy: kern | 2005-10-04 03:55:47 -0700 methodize take() and - r82@Blasphemy: kern | 2005-10-04 03:25:34 -0700 defined(darwin) -> d -Added extra_info kw to add_extension method. Cleaned corelib setup.py. -Added cblas.h. -Improved logic so sequence can act like tuple for advanced indexing as -Made sure that arbitrary sequences (smaller than MAX_DIMS) with slice o -Fixed Broadcasting problem introduced by previous fixes... -Fixed segfaults in buffered casting when castsize was different than bu Testing new user Testing new user -Several fixes. -Fix problem with uninitialized memory in _dotblas. Changed old C-API t -Fixed bug in umath absolute. -Fixing some linalg bugs... -Improved logic. -One minor change to sdist. -Fixed setup.py to add extensions when sdist is run -Fixes by Eric Firing.. -git-svn-id: http://svn.scipy.org/svn/numpy/branches/newcore@1147 94b884 -Matrix indexing weirdness -Simpler names in fftpack -Removed sign and round as methods -- more like ufuncs -Add error checking for object ufuncs. -Improve overflow detection. -Altered error detection to use raiseexcept when appropriate. Also, cle -Fixed some casting rule issues. -Impl. atanhf and using it for detecting HAVE_INVERSE_HYPERBOLIC_FLOAT. -Added two to TEST_COMMIT -Testing fperez commit -Define atanhf,asinhf,acoshf when missing (like in mingw32). -Fixed config_cmd.check_func call bugs that used to return always True. Fix checking Python C API return values in fastumathmodule.c:_scipy_get -Exposing scipy.test(..), scipy.base.test(..) functions. -Fixed and improved scipy.distutils. -Removed Numeric dependence. -Making sure it committed -Removed Numeric dependence. -Fixing header. -Removed bad define. -Fixed compatibility issues. -Pearu can write. -Final changes.. -Removed obsolete test-site. -Added f2py2e as scipy.f2py, improved scipy.distutils, fixed setup.py fi -Changed names from axxxxx to xxxx_ for types that shadow builtin types. -Added test_commit file -More thanks... -Added Stephen Walton's patch for absoft.py -Fixed richcmp inheritance. - r3539@Blasphemy: kern | 2005-09-29 03:12:18 -0700 Bit of a merge con - r3532@Blasphemy: kern | 2005-09-29 03:08:25 -0700 More test fixes - r3531@Blasphemy: kern | 2005-09-29 02:05:01 -0700 Restoring tests -Added a can_cast Python function. -Eliminated random_lite -Simplified scalar type checking. - r3529@Blasphemy: kern | 2005-09-29 00:30:52 -0700 Remove mconf_{BE,L -git-svn-id: http://svn.scipy.org/svn/numpy/branches/newcore@1109 94b884 - r3519@Blasphemy: kern | 2005-09-28 22:56:10 -0700 mtrand pickle supp - r3509@Blasphemy: kern | 2005-09-27 22:50:18 -0700 mtrand fixes -No pre-computing limits in getlimits.py -improving a.flags -Updated flags attribute to special dictionary. -Updated flags attribute to special dictionary. -a.flags returns fancy array-connected dictionary. -Fixed mixed-type array scalar computation issues. -Don't shadow builtins anymore. -Fixed bugs. -Changes.. -Altered string and unicode array scalar types so they inherit from Pyth -Eliminated random_lite library - r3493@803638d6: kern | 2005-09-27 19:52:10 -0700 get_scipy_include() - r3492@803638d6: kern | 2005-09-27 19:03:06 -0700 Update scipy.stats -git-svn-id: http://svn.scipy.org/svn/numpy/branches/newcore@1093 94b884 - r3476@803638d6: kern | 2005-09-26 23:37:30 -0700 Added proper __deal -git-svn-id: http://svn.scipy.org/svn/numpy/branches/newcore@1091 94b884 - r3471@803638d6: kern | 2005-09-26 17:50:29 -0700 Ensured that mtrand - r3470@803638d6: kern | 2005-09-26 17:50:04 -0700 Don't install heade -git-svn-id: http://svn.scipy.org/svn/numpy/branches/newcore@1088 94b884 - r3463@803638d6: kern | 2005-09-26 14:31:52 -0700 Fix for blank math -git-svn-id: http://svn.scipy.org/svn/numpy/branches/newcore@1086 94b884 - r3165@803638d6: kern | 2005-09-26 01:31:46 -0700 Use optimized LAPAC - r3164@803638d6: kern | 2005-09-26 01:27:54 -0700 Added scipy.lib.mtr - r3163@803638d6: kern | 2005-09-26 01:26:04 -0700 Fix compile flags f - r3162@803638d6: kern | 2005-09-26 01:25:20 -0700 Comment out faulty -Fix Accelerate.framework flags -Deleted sub-package manifests -Getting ready to release beta. -Added f2py2e -- not tested. -Added cvs version 1.46.23.2019 of f2py2e to svn -Added thanks file -Don't register multiple copies. -Added simple type registration system. -Fixed some remaining problems. -Removed redundant testing. -Changed OWN_DATA to OWNDATA -Changed new typedef bool to Bool throughout. -Defines for backward compat C-API -Added a cross index function -Return Long integer on size only if necessary. -Return Long integer on size only if necessary. -Allow use of generic type objects for instancing arrays -- use standard -Fixed nasty shape setting bug. -Moved weave to be a sub-package of scipy (so as not to conflict with an -Moved weave -Added weave. Hopefully fixed segfault on Python2.4 and windows -Adding weave and test -Getting weave setup -Fixed up for release. -Moving back to simple Numeric interface. -git-svn-id: http://svn.scipy.org/svn/numpy/branches/newcore@1055 94b884 -git-svn-id: http://svn.scipy.org/svn/numpy/branches/newcore@1054 94b884 -git-svn-id: http://svn.scipy.org/svn/numpy/branches/newcore@1053 94b884 -git-svn-id: http://svn.scipy.org/svn/numpy/branches/newcore@1052 94b884 -git-svn-id: http://svn.scipy.org/svn/numpy/branches/newcore@1051 94b884 -git-svn-id: http://svn.scipy.org/svn/numpy/branches/newcore@1050 94b884 -git-svn-id: http://svn.scipy.org/svn/numpy/branches/newcore@1049 94b884 Testing a new user created with Virtualmin Testing a new user created with Virtualmin Testing a new user created with Virtualmin @@ -5958,60 +5521,6 @@ Testing a new user created with Virtualmin -Altered so that ndarray inherits from ndbigarray which does not export -Fixed fortran ravelling issues. -Added more missing files -Added missing setup.py -Minor comment additions. -Fixed ufunc error handling. -Added ufunc attributes. -added hashability of scalars, and fixed printing of large integers. -Added missing __init__ files -More fixes. -Remove stray print. -problematic MIN removed. -Fixed up limits -Found problem with reshape. -Making more consistent. -Fixing some array scalar inconsistencies. -Adding core scipy functions -Converting masked arrays -Fixed more bugs. -Fixed PyArray_Copy when swapping needed. Removed PyArray_CopyArray (ma -Improving convert code. -Fixed _isnan -Improved isinf if platform doesn't provide. -Even more simplifications to isinf macro -Reworked isinf isnan macros a bit. -Added isnan, isinf, isfinite, and signbit -moved fft back to fftpack -Checked that lapack_lite sources build if needed. -Seems to be completish -Added lapack (if libs already present) -Added fft. -Many fixes. -Added _dotblas install -Added ldexp, frexp, and modf to ufuncs -Added ma hook. -Finished new _dotblas.c -Added PyArray_ArrayType -Fixing added Numeric -Moving Numeric to newcore -Added EnsureArray to C-API and used it to clean up some method calls -SVN version builds -Adding scipy.base -Adding scipy.base -Adding scipy.base to SVN -Moving things under scipy -Moving things to live under scipy -Added new distutils -Fixing for new distutils -Trying to delete CodeGenerators -Adding base -Added base -renamed directories -Creating a branch for the new scipy_core Updated for weave support for numarray. Testing oliphant account Testing oliphant account @@ -6950,4 +6459,3 @@ added build_py module split scipy_test off into its own module. factored out the scipy specific distutils code into a scipy_distutils p -New repository initialized by cvs2svn.