from http://pylonshq.com/irclogs/%23pylons/%23pylons.2009-03-06.log.html#t2009-03-06T08:39:13 :
ianbickingokay; I was just planning on doing decoding in WebOb zepolenianbicking: please don't08:43<http://pylonshq.com/irclogs/%23pylons/%23pylons.2009-03-06.log.html#t2009-03-06T08:43:28> zepolenurls don't have encoding ianbickingzepolen: sure they do, just like POST vars etc08:44<http://pylonshq.com/irclogs/%23pylons/%23pylons.2009-03-06.log.html#t2009-03-06T08:44:18> zepolenbut implicitly assuming they are utf8 is imho wrong08:44<http://pylonshq.com/irclogs/%23pylons/%23pylons.2009-03-06.log.html#t2009-03-06T08:44:40> ianbickingit'd be the same policy as other unicode in webob08:45<http://pylonshq.com/irclogs/%23pylons/%23pylons.2009-03-06.log.html#t2009-03-06T08:45:08> ianbickingyou set request.charset or request.default_charset, and you start getting unicode08:45<http://pylonshq.com/irclogs/%23pylons/%23pylons.2009-03-06.log.html#t2009-03-06T08:45:20> zepolenwhat if you dont want unicode? can you set request.charset = None? ianbickingzepolen: yeah zepolenok then08:48<http://pylonshq.com/irclogs/%23pylons/%23pylons.2009-03-06.log.html#t2009-03-06T08:48:24> zepolenyou probably think im crazy to use utf8 instead of unicode, but the performance gain isn't trivial08:49<http://pylonshq.com/irclogs/%23pylons/%23pylons.2009-03-06.log.html#t2009-03-06T08:49:26> zepolenno decoding requests, no encoding template renders, no decoding (and then reencoding) database results08:50<http://pylonshq.com/irclogs/%23pylons/%23pylons.2009-03-06.log.html#t2009-03-06T08:50:01> steve9001no .__cmp__() or .startswith() or ...08:50<http://pylonshq.com/irclogs/%23pylons/%23pylons.2009-03-06.log.html#t2009-03-06T08:50:49> zepolennot really, i just decode when i need to do string manipulation/comparison08:51<http://pylonshq.com/irclogs/%23pylons/%23pylons.2009-03-06.log.html#t2009-03-06T08:51:21>*** vpol has joined #pylons08:51<http://pylonshq.com/irclogs/%23pylons/%23pylons.2009-03-06.log.html#t2009-03-06T08:51:57> steve9001i don't have to worry about decoding any time i do something that depends on the value. it's a tradeoff I guess, performance versus ease of programming and code clarity08:51<http://pylonshq.com/irclogs/%23pylons/%23pylons.2009-03-06.log.html#t2009-03-06T08:51:57> zepolenit's actually clearer - you never have to guess what something is 08:52<http://pylonshq.com/irclogs/%23pylons/%23pylons.2009-03-06.log.html#t2009-03-06T08:52:46> zepolenbut the issue is more py2.5's fault08:53<http://pylonshq.com/irclogs/%23pylons/%23pylons.2009-03-06.log.html#t2009-03-06T08:53:02> zepoleni'll probably use unicode with py3's better distinction between unicode and bytes08:53<http://pylonshq.com/irclogs/%23pylons/%23pylons.2009-03-06.log.html#t2009-03-06T08:53:58>*** anilm has joined #pylons08:54<http://pylonshq.com/irclogs/%23pylons/%23pylons.2009-03-06.log.html#t2009-03-06T08:54:22> zepolenbut that's years away08:54<http://pylonshq.com/irclogs/%23pylons/%23pylons.2009-03-06.log.html#t2009-03-06T08:54:31> steve9001yeah i think i'm looking forward to it though08:54<http://pylonshq.com/irclogs/%23pylons/%23pylons.2009-03-06.log.html#t2009-03-06T08:54:48> zepolenin the mean time im happy that the app is 2-3x faster using straight utf808:55<http://pylonshq.com/irclogs/%23pylons/%23pylons.2009-03-06.log.html#t2009-03-06T08:55:16> zepolenthere are so many places it helps you don't even fathom08:55<http://pylonshq.com/irclogs/%23pylons/%23pylons.2009-03-06.log.html#t2009-03-06T08:55:59> zepolenstoring unicode strings in memcache means a pickling/depickling+encoding on output overhead08:56<http://pylonshq.com/irclogs/%23pylons/%23pylons.2009-03-06.log.html#t2009-03-06T08:56:35> steve9001i profiled my app a bit with repoze the other day and the function that is called more than any other is isinstance() to check if a string is unicode. i think from the webhelpers.html.literal function
