| # |
Apr 19th 2017, 13:37 |
 dereuromark |
prefix the cache with sth that identifies him, this way you can invalidate only a part of the cache that needs updating |
| # |
Apr 19th 2017, 13:37 |
inoas |
"sounds reasonable" ;-) |
| # |
Apr 19th 2017, 13:36 |
inoas |
all caching breaks down ;p |
| # |
Apr 19th 2017, 13:36 |
inoas |
so every time some dude saves a document revision |
| # |
Apr 19th 2017, 13:36 |
inoas |
haha |
| # |
Apr 19th 2017, 13:36 |
hmic |
i'd just purge it for now :p |
| # |
Apr 19th 2017, 13:36 |
hmic |
you could do that, sure |
| # |
Apr 19th 2017, 13:36 |
hmic |
cahing is easy, invalidating a cache is hard |
| # |
Apr 19th 2017, 13:36 |
inoas |
so on write you would lookup the cache and remove the keys that are wrong |
| # |
Apr 19th 2017, 13:35 |
hmic |
you need something to invalidate the cache, whenever you use one |
| # |
Apr 19th 2017, 13:35 |
inoas |
I don't like "admin panel clear cache" buttons |
| # |
Apr 19th 2017, 13:35 |
hmic |
like i said in the beginning: yes! absolutely! |
| # |
Apr 19th 2017, 13:35 |
inoas |
however I am concerned about timeouts |
| # |
Apr 19th 2017, 13:35 |
hmic |
yes, if the cache is not warmed up, first access will be slow. who cares? |
| # |
Apr 19th 2017, 13:35 |
inoas |
and then simply write the cache |
| # |
Apr 19th 2017, 13:35 |
inoas |
so you would actually let the requests hit the application |
| # |
Apr 19th 2017, 13:34 |
hmic |
so you use a cache! |
| # |
Apr 19th 2017, 13:34 |
inoas |
a fast read proxy |
| # |
Apr 19th 2017, 13:34 |
inoas |
because hitting the database is slow |
| # |
Apr 19th 2017, 13:34 |
hmic |
why do you? |
| # |
Apr 19th 2017, 13:34 |
inoas |
okay, then I want a proxy ;p |
| # |
Apr 19th 2017, 13:34 |
inoas |
https://github.com/makallio85/yaml-route doesn't get purged |
| # |
Apr 19th 2017, 13:34 |
hmic |
a cache is a cache and its not persistent by definition |
| # |
Apr 19th 2017, 13:34 |
hmic |
for you too! |
| # |
Apr 19th 2017, 13:34 |
inoas |
yeah that's not where I want to go |
| # |
Apr 19th 2017, 13:33 |
hmic |
it can get purged whenever |
| # |
Apr 19th 2017, 13:33 |
inoas |
read again, not for me |
| # |
Apr 19th 2017, 13:33 |
hmic |
ionas: it's a cache! |
| # |
Apr 19th 2017, 13:33 |
inoas |
redis continues to live |
| # |
Apr 19th 2017, 13:33 |
inoas |
hmic but APCu breaks down when you restart the webserver |
| # |
Apr 19th 2017, 13:33 |
crazycoder |
i should set _joinData hidden on the fly |
| # |
Apr 19th 2017, 13:32 |
hmic |
but the fastest i can get, which is php-process memory (APCu) followed by memcache |
| # |
Apr 19th 2017, 13:32 |
crazycoder |
$_hidden on entity works but i cannot change it globally. |
| # |
Apr 19th 2017, 13:32 |
crazycoder |
hey hmic! |
| # |
Apr 19th 2017, 13:32 |
crazycoder |
it is really really bad but if there is no alternative i should do that: https://gist.github.com/anonymous/3f55105fe22a4db8dd017285d140f383 |
| # |
Apr 19th 2017, 13:32 |
inoas |
Instead? Memcache? |
| # |
Apr 19th 2017, 13:31 |
hmic |
redis is a quite good cache for heavier things, i'd personally not use it for routes |
| # |
Apr 19th 2017, 13:30 |
inoas |
like https://github.com/makallio85/yaml-route but in redis |
| # |
Apr 19th 2017, 13:30 |
inoas |
redis would be not a simple cache but more like the router-storage |
| # |
Apr 19th 2017, 13:30 |
inoas |
and then build-cache-on-write |
| # |
Apr 19th 2017, 13:29 |
inoas |
I am considering: hit cache first, if not found, hit cake-stack? |