I have still not managed to solve my problem. To my mind the garbage collector is culprit in some way, but I can't figure out how.
I join to this mail a patch to be used on "pytables-1.3-ds-devel ", hoping that you or someone else would have a look at my work, and would find what's going wrong.
The code to create PyTables objects from references is mainly located in the hdf5Extension.pyx, from line 1132 to line 1214. By the way currently I've only dealt with the attribute "DIMENSION_LIST", so I've only tried to create DimensionScales from references. But it could be extended to the "REFERENCE_LIST" without much difficulties (to my mind).
I've also performed some minor modifications to some constructors (added a new attribute called "dataset_id" and some other minor things).
Remark : I've let some "print" statements in the code, as I thought it may help you understand what's going wrong.
I also join to this mail a little script : run it and you should get a segmentation fault (and some other stuff, see the remark above).
Finally, any feedback is welcome !
--
Remi
patch_ds.tar.gz
Description: GNU Zip compressed data
example.tar.gz
Description: GNU Zip compressed data
