On Sat, May 29, 2010 at 8:19 PM, Noufal Ibrahim <nou...@gmail.com> wrote:
> On Sat, May 29, 2010 at 4:26 PM, Gaurav Kalra <gvka...@gmail.com> wrote: > > Hi Noufal. > > > > >From Wikipedia: > > JSON only handle basic Python types like strings, integers, and > > collections of basic types, whereas pickle is intended for arbitrary > > objects. > > > > pickle won't be more suitable for what Murugadoss is asking for ? > > For general purpose serialisation, json is better. Pickle is a Python > specific format that will necessitate that the remote end be Python > (or atleast have a sawn off depickler). Also, unpickling involves > something like execution and the data is to be treated as malicious > (check out the warning here - > http://docs.python.org/library/pickle.html). JSON has no such issues > since it's just data . With JSON, the remote end is free to be in > whatever language you want which is a good thing. > > It is possible to jsonify arbitrary types but you'll have to decide > how. I'm not sure there's much point in serialising code itself and > sending it across a line. Well, maybe for something like an RPC > mechanism but even then, it shouldn't be code but some data > representing an instruction of some kind. > I use a few thumb rules when deciding between JSON and Pickle. 1. If data is to be persisted to disk and not sent out to network, use Pickle, especially when only a Python runtime is involved. 2. If the data has to be portable not just across Python runtimes, use JSON. Finally if I dont have much time to decide what the data format should be in terms of portability, safety etc, I always default to JSON, since that is generally safer. > > -- > ~noufal > http://nibrahim.net.in. > > _______________________________________________ > BangPypers mailing list > BangPypers@python.org > http://mail.python.org/mailman/listinfo/bangpypers > -- --Anand _______________________________________________ BangPypers mailing list BangPypers@python.org http://mail.python.org/mailman/listinfo/bangpypers
