Changes from 2.1 to 2.1.1

openFile(), isHDF5File() and isPyTablesFile() functions accept Unicode filenames now. Closes #202 and #214.
When creating large type sizes (exceeding 64 KB), HDF5 complained and refused to do so. The HDF5 team has logged the issue as a bug, but meanwhile it has been implemented a workaround in PyTables that allows to create such large datatypes for situations that does not require defaults other than zero. Addresses #211.
In order to be consistent with how are stored the other data types, Unicode attributes are retrieved now as NumPy scalars instead of Python Unicode strings or NumPy arrays. For the moment, I've fixed this through pickling the Unicode strings. In the future, when HDF5 1.8.x series would be a requirement, that should be done via a HDF5 native Unicode type. Closes #213.

Fixed a memory leak when a lot of queries were made. Closes #203 and #207.
The chunkshape="auto" parameter value of Leaf.copy() is honored now, even when the (start, stop, step) parameters are specified. Closes #204.
Due to a flaw in its design, the File class was not able to be subclassed. This has been fixed. Closes #205.
Default values were not correctly retrieved when opening already created CArray/EArray objects. Fixed. Closes #212.
Fixed a problem with the installation of the nctoh5 script that prevented it from being executed. Closes #215.
[Pro] The iterseq cache ignored non-indexed conditions, giving wrong results when those appeared in condition expressions. This has been fixed. Closes #206.

Enjoy data!

—The PyTables Team