On 27-5-2013 2:39, Roy Smith wrote: > In article <51a28f42$0$15870$e4fe5...@news.xs4all.nl>, > Irmen de Jong <irmen.nos...@xs4all.nl> wrote: > >> On 26-5-2013 22:48, Roy Smith wrote: >> >>> The advantage of pickle over json is that pickle can serialize many >>> types of objects that json can't. The other side of the coin is that >>> pickle is python-specific, so if you think you'll ever need to read your >>> data from other languages, pickle is right out. >> >> That is not entirely true :) I've written a pickle implementation for Java >> and .NET >> that is almost feature complete; it is part of >> http://pythonhosted.org/Pyro4/pyrolite.html > > Very cool > >> Still, pickle may not be the best choice here. > > Perhaps not, but lots of points for the awesomeness factor. >
Thanks for the praise :) There's another interesting thing perhaps to also mention about Pyrolite. Its Pickle implementation obviously maps built in types to their counterparts in Java/.NET, but it is also capable of unpickling custom classes. It defaults to outputting a simple hashtable with the class's properties in it, but you can also provide a custom deserializer to recreate a custom object (it already does this automatically for a few complex types such as DateTime, BigDecimal, and a bunch of Pyro specific types). As the unpicklers are not executing any Java or .NET code dynamically they are not susceptible to the security issue that Python's pickle has. Still: tread with care. (Especially if you use the lib to pickle stuff and unpickle it on the Python side). Also, the one missing feature is memo-ing when pickling so that you can pickle recursive object graphs. It now throws a stack overflow exception instead. Irmen -- http://mail.python.org/mailman/listinfo/python-list
