| 92c0a1c | Jeff Whitaker | 22 April 2014, 19:07:30 UTC | prepare for 1.0.9 release tag | 22 April 2014, 19:07:30 UTC |
| 010e747 | Jeff Whitaker | 21 April 2014, 15:35:50 UTC | Merge pull request #244 from jswhit/master update docs | 21 April 2014, 15:35:50 UTC |
| 1887f19 | Jeff Whitaker | 21 April 2014, 15:34:45 UTC | update docs | 21 April 2014, 15:34:45 UTC |
| 9814fe2 | Jeff Whitaker | 20 April 2014, 12:37:00 UTC | Merge pull request #243 from jswhit/master speed up the creation of new Group instances (issue 239) | 20 April 2014, 12:37:00 UTC |
| 008fe07 | Jeff Whitaker | 18 April 2014, 16:24:16 UTC | remove another unused variable | 18 April 2014, 16:24:16 UTC |
| 3bdf609 | Jeff Whitaker | 18 April 2014, 16:23:16 UTC | remove some unused variables | 18 April 2014, 16:23:16 UTC |
| 2bb51e9 | Jeff Whitaker | 18 April 2014, 13:50:54 UTC | fix bug in previous commit | 18 April 2014, 13:50:54 UTC |
| f983142 | Jeff Whitaker | 18 April 2014, 13:45:38 UTC | speed up the creation of new Groups (issue 239). | 18 April 2014, 13:45:38 UTC |
| 877f5f4 | Jeff Whitaker | 18 April 2014, 12:30:37 UTC | Merge pull request #240 from jswhit/master fix for issue 236. | 18 April 2014, 12:30:37 UTC |
| 00bf22e | Jeff Whitaker | 17 April 2014, 19:29:53 UTC | Merge pull request #241 from shoyer/better-string-dtype-itemsize Switch from using datatype.str[2:] to datatype.itemsize | 17 April 2014, 19:29:53 UTC |
| 58209aa | Stephan Hoyer | 17 April 2014, 17:54:38 UTC | Switch from using datatype.str[2:] to datatype.itemsize This couples us less heavily to the numpy dtype string representation. Also, it might be possible to have numpy strings which don't specify a byte order? I'm not sure, but this is cleaner code in any case. | 17 April 2014, 17:54:38 UTC |
| b019167 | Jeff Whitaker | 17 April 2014, 17:25:43 UTC | fix for issue 236. | 17 April 2014, 17:25:43 UTC |
| 4311c06 | Jeff Whitaker | 17 April 2014, 17:23:49 UTC | Merge pull request #238 from jswhit/master addresses issue 237 | 17 April 2014, 17:23:49 UTC |
| a48408d | Jeff Whitaker | 17 April 2014, 15:00:05 UTC | clean up previous commit | 17 April 2014, 15:00:05 UTC |
| 52c6575 | Jeff Whitaker | 17 April 2014, 12:34:04 UTC | fix for pull request 224 - makes sure that dtype='U1' is converted to vlen string in createVariable (not character). | 17 April 2014, 12:34:04 UTC |
| c2874b7 | Jeff Whitaker | 05 April 2014, 19:43:41 UTC | python 2.5 is required | 05 April 2014, 19:43:41 UTC |
| 602d28d | Jeff Whitaker | 30 March 2014, 01:48:12 UTC | Merge pull request #234 from shoyer/reduce-getitem-memory Replace zip with itertools.izip under Python 2 | 30 March 2014, 01:48:12 UTC |
| 34fb2df | Stephan Hoyer | 26 March 2014, 23:07:16 UTC | Replace zip with itertools.izip under Python 2 This reduces the amount of memory required for complicated calls to Variable.__getitem__ under array indexers, which iterates over all indexers with "zip". netCDF4-python's handling of array indexers is still very complicated/slow when there are multiple array indexers involved, but this at least improves the situation. | 26 March 2014, 23:12:24 UTC |
| 91867bf | Jeff Whitaker | 10 March 2014, 18:06:14 UTC | Merge pull request #224 from shoyer/numpy-string-dtype Allow creating and setting variables with numpy string arrays | 10 March 2014, 18:06:14 UTC |
| b9665db | Stephan Hoyer | 10 March 2014, 16:54:56 UTC | Fixed typo in Changelog | 10 March 2014, 17:09:15 UTC |
| ef7452b | Stephan Hoyer | 09 March 2014, 22:38:40 UTC | Added documentation for handling numpy arrays with string datatypes | 10 March 2014, 17:09:15 UTC |
| deb5688 | Stephan Hoyer | 09 March 2014, 21:54:52 UTC | Added error message for using dtype=str when format!='NETCDF4' | 10 March 2014, 17:08:16 UTC |
| d5ced70 | Stephan Hoyer | 06 March 2014, 07:01:25 UTC | Allow creating and setting variables with numpy string arrays This patch automatically converts numpy string datatypes of length > 1 into `str` when used as arguments to `createVariable`, and allows for setting arrays with numpy string datatypes (they are converted internally into arrays with dtype=object, but the user doesn't need to know that detail). The reason for this patch is for consistency: I expected that if `array.dtype` is not a valid argument argument to `createVariable`, then it isn't possible to create (or fill) a Variable with the values of the `numpy.ndarray` `array`. Strings are the exception to this rule. This patch eliminates this quirk in a manner that I believe will be backwards compatible, since only strings of length 1 (character arrays) are a supported primitive datatype. If this looks good to you, I will add this feature to the Change Log and the documentation in my next commit. | 10 March 2014, 17:08:16 UTC |
| f7a6e21 | Jeff Whitaker | 06 March 2014, 22:56:20 UTC | Merge pull request #231 from jswhit/master change error message when Dataset has been garbage collected. | 06 March 2014, 22:56:20 UTC |
| d9fdafd | Jeff Whitaker | 06 March 2014, 22:54:55 UTC | change error message when Dataset has been garbage collected. | 06 March 2014, 22:54:55 UTC |
| 9c28cee | Jeff Whitaker | 06 March 2014, 21:22:38 UTC | Merge pull request #230 from jswhit/master fix python 3 print | 06 March 2014, 21:22:38 UTC |
| 2af22a5 | Jeff Whitaker | 06 March 2014, 21:22:06 UTC | fix python 3 print | 06 March 2014, 21:22:06 UTC |
| 8888b0a | Jeff Whitaker | 06 March 2014, 21:20:40 UTC | fix python 3 print | 06 March 2014, 21:20:40 UTC |
| bdb7654 | Jeff Whitaker | 06 March 2014, 21:07:43 UTC | Merge pull request #229 from jswhit/master add item for issue 225 | 06 March 2014, 21:07:43 UTC |
| 8657dde | Jeff Whitaker | 06 March 2014, 21:06:57 UTC | add item for issue 225 | 06 March 2014, 21:06:57 UTC |
| c129ed4 | Jeff Whitaker | 06 March 2014, 21:03:27 UTC | Merge pull request #228 from jswhit/master silence warnings, remove cruft from run_all.py, print skipped tests | 06 March 2014, 21:03:27 UTC |
| f852314 | Jeff Whitaker | 06 March 2014, 20:58:33 UTC | silence warnings, remove cruft from run_all.py, explain what tests are skipped. | 06 March 2014, 20:58:33 UTC |
| 4cdfda6 | Jeff Whitaker | 06 March 2014, 20:09:53 UTC | Merge pull request #227 from jswhit/master fix .travis.yml so it runs test/run_all.py | 06 March 2014, 20:09:53 UTC |
| 7eb8ff9 | Jeff Whitaker | 06 March 2014, 20:08:33 UTC | turn off email notifications | 06 March 2014, 20:08:33 UTC |
| 1d2118f | Jeff Whitaker | 06 March 2014, 20:00:29 UTC | update | 06 March 2014, 20:00:29 UTC |
| e78a0b3 | Jeff Whitaker | 06 March 2014, 19:57:19 UTC | update | 06 March 2014, 19:57:19 UTC |
| 2914d37 | Jeff Whitaker | 06 March 2014, 19:36:08 UTC | remove test script | 06 March 2014, 19:36:08 UTC |
| 7100c0f | Jeff Whitaker | 06 March 2014, 19:34:02 UTC | update | 06 March 2014, 19:34:02 UTC |
| 0e3fe54 | Jeff Whitaker | 06 March 2014, 17:44:16 UTC | make shell script for travis to run | 06 March 2014, 17:44:16 UTC |
| c25d1a9 | Jeff Whitaker | 06 March 2014, 17:18:02 UTC | update to trigger travis | 06 March 2014, 17:18:02 UTC |
| ac96813 | Jeff Whitaker | 06 March 2014, 16:16:10 UTC | have travis run test/run_all.py, send email | 06 March 2014, 16:16:10 UTC |
| 2e691fa | Jeff Whitaker | 06 March 2014, 15:19:59 UTC | Merge pull request #226 from shoyer/add-travis Add .travis.yml for Travis-CI | 06 March 2014, 15:19:59 UTC |
| 7f9e2fc | Stephan Hoyer | 06 March 2014, 08:03:16 UTC | Add .travis.yml for Travis-CI This is most of the work of setting up Travis-CI as I suggest in #225. Almost all tests pass with this configuration file: https://travis-ci.org/shoyer/netcdf4-python/builds/20191174 The "testing diskless file capability" test requires a newer build of the netCDF4 library than is available pre-built for Ubuntu, so it's not surprise it fails (I suppose we really ought to build the netCDF library from source). As for the "testing unicode" test on Python 2.7, that also fails when I run the test suite locally on laptop. | 06 March 2014, 08:03:16 UTC |
| 9cc8558 | Jeff Whitaker | 06 March 2014, 02:26:47 UTC | Merge pull request #223 from jswhit/master fix another broken test in python 3.3 | 06 March 2014, 02:26:47 UTC |
| 2be6fca | Jeff Whitaker | 06 March 2014, 02:25:31 UTC | fix another broken test in python 3.3 | 06 March 2014, 02:25:31 UTC |
| c2ace4b | Jeff Whitaker | 06 March 2014, 02:21:59 UTC | Merge pull request #222 from jswhit/master fix broken test in python 3.3 | 06 March 2014, 02:21:59 UTC |
| e9a3039 | Jeff Whitaker | 06 March 2014, 02:20:56 UTC | fix broken test in python 3.3 | 06 March 2014, 02:20:56 UTC |
| d8728b7 | Jeff Whitaker | 06 March 2014, 02:15:26 UTC | Merge pull request #221 from jswhit/master make __unicode__ return a more useful error message for Variable and Dimension when Dataset no longer exists. | 06 March 2014, 02:15:26 UTC |
| 1c95d02 | Jeff Whitaker | 06 March 2014, 02:13:31 UTC | update | 06 March 2014, 02:13:31 UTC |
| 5e1a54c | Jeff Whitaker | 06 March 2014, 02:12:07 UTC | modify __unicode__ for Variable and Dimension to return more useful error message when Dataset object has been garbage collected. | 06 March 2014, 02:12:07 UTC |
| 01cbb86 | Jeff Whitaker | 06 March 2014, 02:00:00 UTC | Merge pull request #219 from jswhit/master potential fix for issue 218 (circular references preventing garbage collection) | 06 March 2014, 02:00:00 UTC |
| fb5c597 | Jeff Whitaker | 04 March 2014, 21:19:18 UTC | take out outdated restriction on diskless files | 04 March 2014, 21:19:18 UTC |
| 339ac7f | Jeff Whitaker | 04 March 2014, 21:11:28 UTC | don't try to print fill info for non-primitive data types | 04 March 2014, 21:11:28 UTC |
| 4810136 | Jeff Whitaker | 03 March 2014, 20:37:10 UTC | update url in setup.py to point to github instead of googlecode | 03 March 2014, 20:37:10 UTC |
| 20517be | Jeff Whitaker | 01 March 2014, 15:31:39 UTC | use nc-config by default | 03 March 2014, 20:31:02 UTC |
| a595562 | Jeff Whitaker | 28 February 2014, 13:12:04 UTC | fix typo in comment | 03 March 2014, 20:31:02 UTC |
| 4f32e9a | Jeff Whitaker | 27 February 2014, 21:05:24 UTC | rename README to README.md | 03 March 2014, 20:31:02 UTC |
| 127b065 | Jeff Whitaker | 27 February 2014, 20:14:15 UTC | remove bogus 'pass' | 03 March 2014, 20:31:02 UTC |
| 1ff35c2 | Jeff Whitaker | 27 February 2014, 20:12:29 UTC | add some comments | 03 March 2014, 20:31:02 UTC |
| a20bc0d | Jeff Whitaker | 27 February 2014, 16:43:16 UTC | update | 03 March 2014, 20:31:02 UTC |
| 90a7324 | Jeff Whitaker | 27 February 2014, 16:30:49 UTC | use weakref.proxy instead of weakref.ref | 03 March 2014, 20:31:01 UTC |
| ba2db70 | Jeff Whitaker | 27 February 2014, 14:45:18 UTC | test for issue 218 | 03 March 2014, 20:29:54 UTC |
| 2488d2b | Jeff Whitaker | 27 February 2014, 14:06:00 UTC | add comment referencing issue 218 | 03 March 2014, 20:29:54 UTC |
| 79599f6 | Jeff Whitaker | 27 February 2014, 13:03:25 UTC | potential fix for issue 218 | 03 March 2014, 20:29:54 UTC |
| 1a27c2f | Jeff Whitaker | 27 February 2014, 13:01:33 UTC | potential fix for issue 218 | 03 March 2014, 20:29:53 UTC |
| 9f59b11 | Jeff Whitaker | 03 March 2014, 20:04:44 UTC | Merge pull request #220 from shoyer/fix-0d-arrays __getitem__ for 0d variables returns 0d arrays (not 1d arrays) | 03 March 2014, 20:04:44 UTC |
| 3d2f9eb | Stephan Hoyer | 03 March 2014, 18:29:21 UTC | Edits in response to comments on PR#220 | 03 March 2014, 18:29:21 UTC |
| cc56739 | Stephan Hoyer | 03 March 2014, 02:04:04 UTC | __getitem__ for 0d arrays returns 0d arrays (not 1d arrays) This fix ensures that indexing a 0-dimensional array returns a 0-dimensional numpy array (consistent with numpy's behavior), not a 1-dimensional array with one element, which is the current behavior. I'm sure there is a more elegant solution to this issue, but I think this is one of the better options short of rewriting the core indexing routines. As an alternative, I did try removing the hack `if nDims == 0: ...` at the start of _StartCountStride but that definitely broke things. Presumably that hack is there for a reason? Note that to fully match how numpy handles 0-dimensional arrays, we would actually raise an error when somebody tries to access or set a 0-dimensional variable with `v[:]`. The right syntax for indexing a 0-dimensional array is somewhat obscure: `v[()]` (using an empty tuple) or `v[...]`. But I avoided this option because it would assuredly break some user code. My fix here will probably break some poorly constructed user code, too, if it relied on the broken 0-d arrays are always 1-d behavior. The right way to handle this would have been to use `np.asscalar`. So there may be a case for saving this until the next major version. | 03 March 2014, 02:04:04 UTC |
| 5448f80 | Ryan May | 25 February 2014, 23:30:13 UTC | Ignore build directory. | 25 February 2014, 23:30:13 UTC |
| 1015980 | Jeff Whitaker | 25 February 2014, 02:44:56 UTC | bump version number to 1.0.9 | 25 February 2014, 02:44:56 UTC |
| d6939b7 | Jeff Whitaker | 19 February 2014, 13:03:01 UTC | add __array__ method to Variable to make unfuncs faster on Variable objects (issue 216). | 19 February 2014, 13:03:01 UTC |
| 49a03fd | Jeff Whitaker | 12 February 2014, 16:59:17 UTC | update | 12 February 2014, 16:59:17 UTC |
| 8e84b53 | Jeff Whitaker | 12 February 2014, 16:10:37 UTC | update download_url | 12 February 2014, 16:10:37 UTC |
| 3c0c7c9 | Jeff Whitaker | 12 February 2014, 13:17:05 UTC | test for issue 215 | 12 February 2014, 13:17:05 UTC |
| c61fc00 | Jeff Whitaker | 12 February 2014, 13:05:34 UTC | fix for issue 215 | 12 February 2014, 13:05:34 UTC |
| dffa80f | Jeff Whitaker | 10 February 2014, 16:29:06 UTC | silence annoying warnings with Cython 0.2. | 10 February 2014, 16:29:06 UTC |
| 74ad480 | Jeff Whitaker | 07 February 2014, 21:54:18 UTC | update | 07 February 2014, 21:54:18 UTC |
| 439fb7b | Jeff Whitaker | 07 February 2014, 21:52:25 UTC | update docs | 07 February 2014, 21:52:25 UTC |
| 52a5814 | Jeff Whitaker | 29 January 2014, 17:37:45 UTC | fix typo | 29 January 2014, 17:37:45 UTC |
| 491389c | Jeff Whitaker | 29 January 2014, 17:29:12 UTC | Dataset.__unicode__ (ncinfo) now underlines variable names | 29 January 2014, 17:29:12 UTC |
| 2c2cfc0 | Jeff Whitaker | 29 January 2014, 17:19:33 UTC | update | 29 January 2014, 17:19:33 UTC |
| a50f186 | Jeff Whitaker | 29 January 2014, 17:10:06 UTC | fix api check for nc_inq_format_extended | 29 January 2014, 17:10:06 UTC |
| e376c0e | Jeff Whitaker | 29 January 2014, 16:58:16 UTC | add disk_format attribute, change file_format attribute to data_model. | 29 January 2014, 16:58:16 UTC |
| 3041867 | Jeff Whitaker | 16 December 2013, 23:23:02 UTC | bump version number to 1.0.8 | 16 December 2013, 23:23:02 UTC |
| 8a90b3b | Jeff Whitaker | 16 December 2013, 23:18:16 UTC | Fix URL | 16 December 2013, 23:18:16 UTC |
| a597c5c | Jeff Whitaker | 29 November 2013, 21:06:59 UTC | regenerate C | 29 November 2013, 21:06:59 UTC |
| 9aa5835 | Jeff Whitaker | 29 November 2013, 16:33:41 UTC | ignore default _FillValue for signed and unsigned byte data types (issue 209) ncinfo now returns fill mode information. | 29 November 2013, 16:33:41 UTC |
| e9edcf2 | Jeff Whitaker | 28 November 2013, 16:24:54 UTC | update comment. | 28 November 2013, 16:24:54 UTC |
| 9917179 | Jeff Whitaker | 28 November 2013, 15:51:35 UTC | regenerate C | 28 November 2013, 15:51:35 UTC |
| b04cd6b | Jeff Whitaker | 28 November 2013, 15:49:27 UTC | issue 209 - check to see if fill value was disabled before masking data equal to fill value. | 28 November 2013, 15:49:27 UTC |
| 70902a6 | Jeff Whitaker | 19 November 2013, 16:04:26 UTC | update | 19 November 2013, 16:04:26 UTC |
| 08d79a7 | Jeff Whitaker | 19 November 2013, 16:03:32 UTC | add variable type info to Dataset.__repr__ | 19 November 2013, 16:03:32 UTC |
| e2849c0 | Jeff Whitaker | 18 November 2013, 17:54:09 UTC | don't recreate netCDF4.c with cython when running 'setup.py sdist' (ported from v1.0.7rel tag) | 18 November 2013, 17:54:09 UTC |
| eeb4a97 | Jeff Whitaker | 18 November 2013, 17:34:34 UTC | update | 18 November 2013, 17:34:34 UTC |
| 7521060 | Jeff Whitaker | 15 November 2013, 16:49:15 UTC | update | 15 November 2013, 16:49:15 UTC |
| 366994f | Jeff Whitaker | 14 November 2013, 16:31:06 UTC | update | 14 November 2013, 16:31:06 UTC |
| 4ac1794 | Jeff Whitaker | 14 November 2013, 16:26:42 UTC | update build instructions. | 14 November 2013, 16:26:42 UTC |
| 3d2ae74 | Jeff Whitaker | 14 November 2013, 16:06:32 UTC | add ability to specify how to link hdf4,jpeg and curl (needed if linking static libnetcdf built with hdf4 and OpenDAP support) | 14 November 2013, 16:06:32 UTC |
| 88276ed | Jeff Whitaker | 14 November 2013, 13:44:52 UTC | update comment. | 14 November 2013, 13:44:52 UTC |
| ba51052 | Jeff Whitaker | 14 November 2013, 13:40:58 UTC | fix for issue 202. | 14 November 2013, 13:40:58 UTC |
