Bug in pickling of tables prevents compiling

Joshua Macy — Fri, 19 Jan 2001 18:50:16 -0000

For some (but apparently not all) tables, pickling and unpickling results in a pickle error complaining that the pickle file was truncated. This doesn't seem to have anything to do with whether the files were read/written with the 'b' option, but rather with the exact structure of the table being pickled. For intance, the "Hello World" table constructed "by hand" in the doctest in table.py can be successfully pickled and unpickled, but the same table read from an XML file using rollsax.py gets the pickle error after running compile.py and then roll.py. Apparently there's something different about the master table created by the doctest and the master table returned by the content handler.

At the moment, I'm at a loss as to exactly what the difference is, particularly with such a simple table; I'll probably end up creating a unittest that compares the two master tables field-by-field, but not right now. Since the compile.py utility only exists as a speed optimization for large tables, and so far nobody has complained that it hasn't worked (which probably means nobody uses it), fixing it isn't urgent.

Recent changes to bugs

Bug in pickling of tables prevents compiling