On Wednesday, September 4, 2013 4:29:30 PM UTC-7, Volker Braun wrote: > > Having spent a day to look closer at pickling issues, there seem to be > some dangerous issues. In particular, Python currently has a bug where you > can create pickles that unpickle incorrectly without raising any exceptions > (http://trac.sagemath.org/15156). >
OK, the problem were two sage classes, CachedFunction and Map, which tried to pickle caches and dictionaries (containing caches!) as part of the construction step. Since those can easily contain circular references, that's not a good idea. The solution is to move their initialization to the setstate phase. Of course, this only resolves the circularity part of the problems. Objects that end up unpickling badly because *too much* of their state is deferred to setstate need their `reduce` methods adjusted in the opposite direction. Note that while Python has hooks such as __getinitargs__ and __getnewargs__ to customize reduction behaviour, most sage code seems to prefer to roll its own methods via custom __reduce__ methods. This gives better control but makes it easier to err in the direction of trying to do too much at construction time. -- You received this message because you are subscribed to the Google Groups "sage-devel" group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To post to this group, send email to sage-devel@googlegroups.com. Visit this group at http://groups.google.com/group/sage-devel. For more options, visit https://groups.google.com/groups/opt_out.
