 |
 |
On Mon, May 18, 2015 at 10:34 PM, Eric Lease Morgan <[log in to unmask]> wrote: > On May 18, 2015, at 9:23 PM, Galen Charlton <[log in to unmask]> wrote: > > >> I have two scripts, attached. They do EXACTLY the same thing > >> in almost EXACTLY the same manner, but the Python script is > >> almost 25 times slower than the Perl script: > > > > I'm no Python expert, but I think that the difference is much more > > likely due to which JSON processor is being used. I suspect your Perl > > environment has the JSON::XS module, which is written in C, is fast, > > and is automatically invoked (if present) by "use JSON;". > > > > In contrast, I believe that the Python "json" library is written in > > Python itself. I tried swapping in cjson and UltraJSON [1] in place > > of "json" in your Python script, and in both cases it ran rather > > faster. > > > > [1] https://github.com/esnme/ultrajson > > > Thank you. After using the Python module ujson instead of json, the speed > of my two scripts is now all but equal. Whew! —Eric > This is kind of weird, since the stdlib json module should be using a C extension (from _json). ultrajson does claim to be faster than simplejson (which is the library that got incorporated into the python 2.6 stdlib as "json"), but I wouldn't have expected it to make that much of a difference. Out of curiosity, does "import _json" work on your Python? I'm wondering if the C extension didn't build for some reason.