Hello Martin, A Sunday 16 May 2010 15:13:35 Martin Teichmann escrigué: > Hello List, Hello Francesc, > > I worked a bit more on the extensions necessary > to use dimension scales, and now I think it is > ready for inclusion into the main branch. > It's all uploaded into the svn repository, and > merged with the current trunk.
Mmh, you seem to forgot to merge more things from trunk than you actually do. The net result is that I cannot compile your branch. Anyway, I've setup a new branch: http://www.pytables.org/svn/pytables/branches/dimension-scales2 that compiles correctly (it needed with some small additional modifications). Please work with this new branch (I've given you commit privileges to it). At any rate, in order to avoid messes in the future, do not try to merge things from trunk into branches (unless absolutely necessary, e.g. you need a feature in trunk). The normal work-flow is merging from branches into trunk, and not the contrary. > Since 2.2 beta 3 is already out, I guess I'm a bit > late for 2.2, but for the next release I guess it's fine. Yeah, it is a bit late, and in addition, there are some tests that do not work correctly: ====================================================================== ERROR: None (tables.tests.test_attributes.NotCloseTypesTestCase) ---------------------------------------------------------------------- Traceback (most recent call last): File "/home/faltet/PyTables/pytables/branches/dimension- scales2/tables/tests/test_attributes.py", line 915, in test03_setObjectAttributes self.array.attrs.rs = {"ddf":32.1, "dsd":1} File "/home/faltet/PyTables/pytables/branches/dimension- scales2/tables/attributeset.py", line 453, in __setattr__ self._g__setattr(name, value) File "/home/faltet/PyTables/pytables/branches/dimension- scales2/tables/attributeset.py", line 394, in _g__setattr self._g_setAttr(self._v_node, name, stvalue) File "hdf5Extension.pyx", line 447, in tables.hdf5Extension.AttributeSet._g_setAttr (tables/hdf5Extension.c:3691) File "hdf5Extension.pyx", line 264, in tables.hdf5Extension.islist (tables/hdf5Extension.c:2065) KeyError: 0 ====================================================================== ERROR: None (tables.tests.test_attributes.CloseTypesTestCase) ---------------------------------------------------------------------- Traceback (most recent call last): File "/home/faltet/PyTables/pytables/branches/dimension- scales2/tables/tests/test_attributes.py", line 915, in test03_setObjectAttributes self.array.attrs.rs = {"ddf":32.1, "dsd":1} File "/home/faltet/PyTables/pytables/branches/dimension- scales2/tables/attributeset.py", line 453, in __setattr__ self._g__setattr(name, value) File "/home/faltet/PyTables/pytables/branches/dimension- scales2/tables/attributeset.py", line 394, in _g__setattr self._g_setAttr(self._v_node, name, stvalue) File "hdf5Extension.pyx", line 447, in tables.hdf5Extension.AttributeSet._g_setAttr (tables/hdf5Extension.c:3691) File "hdf5Extension.pyx", line 264, in tables.hdf5Extension.islist (tables/hdf5Extension.c:2065) KeyError: 0 ====================================================================== FAIL: None (tables.netcdf3.tests.test_netcdf3.NetCDFFileTestCase) ---------------------------------------------------------------------- Traceback (most recent call last): File "/home/faltet/PyTables/pytables/branches/dimension- scales2/tables/tests/common.py", line 248, in newmethod return oldmethod(self, *args, **kwargs) File "/home/faltet/PyTables/pytables/branches/dimension- scales2/tables/netcdf3/tests/test_netcdf3.py", line 75, in test_repr assert f.__repr__() == outstring AssertionError ---------------------------------------------------------------------- Ran 2949 tests in 43.360s FAILED (failures=1, errors=2) > A little summary of the changes I did: > > - implemented a class "reference" as a new type for > numpy arrays, which represents HDF5 references. I've had a look at it. There a couple of things that I don't like specially: - When dereferrencing, you need to pass the `file` argument to `deref()` method. IMO, this is non-natural. Perhaps it would be possible to add a new attribute in the `reference` class (and perhaps new `__getattr__()` and `__setattr__()` special methods for keeping the file info? - You have chosen the 'r' typecode for the class. But what would happen if the NumPy developers use 'r' in the future? I think I've already expressed my worries about this before, but perhaps it is not as dangerous as I think? > - attributes which are list-of-arrays-of some type (esp. reference) > are stored and loaded as a list of vlens. > - attributes which are a list of strings are stored and > loaded as variable-length strings That's nice, but your implementation seems to interact badly with existing functionality (see tests failures above). > - a unit test for all that > - added some documentation The excerpt about the ABI incompatibility among NumPy 1.3 and 1.4 no longer applies as NumPy crew have decided to declare 1.4 as invalid and released 1.4.1 which is ABI compatible with 1.x series. So you may want to remove the note about this. > so, it would be nice if someone else tested all that > independently from me. Exactly. More people interested in testing the new functionality? Thanks for the contribution! -- Francesc Alted ------------------------------------------------------------------------------ _______________________________________________ Pytables-users mailing list Pytables-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/pytables-users
