This patch does two things:
- it speeds up decoding of JSON objects by doing fewer temporary allocations
- it reduces memory use of decoded JSON objects by reusing key strings
when they appear several times in a JSON string

The problem with this is that it makes json in Python diverge even more
from the simplejson trunk.

Unless Bob applies it to simplejson as well, that is.

I've just noticed that there is a persistent decoder instance in
`json._default_decoder`. To avoid keeping references to past keys
forever, this updated patch clears the memo dict when a string has
finished decoding.

Here is the patch.

I wonder if you can make a patch for 2.6 python branch.

> I wonder if you can make a patch for 2.6 python branch.

No, 2.6 doesn't have the same C accelerator in the first place.

I applied most of this patch to r206 of simplejson trunk

Le dimanche 27 décembre 2009 à 11:10 +0000, Bob Ippolito a écrit :
> Bob Ippolito <bob@redivi.com> added the comment:
> 
> I applied most of this patch to r206 of simplejson trunk

Thank you. Will you port it to CPython yourself or would you prefer
someone else to do it?

Updated patch against py3k.

Committed in r84505.