| Time |
S |
Nick |
Message |
| 07:43 |
|
|
aditsu_ joined ##friendlyjava |
| 07:44 |
|
aditsu_ |
hi, would you recommend using ThreadLocal for SimpleDateFormat in a web application? |
| 07:44 |
|
aditsu_ |
(to have a separate SimpleDateFormat per thread) |
| 07:56 |
|
aditsu_ |
similar question for DecimalFormat.. |
| 08:12 |
|
aditsu_ |
hmm, ThreadLocal is not serializable |
| 08:27 |
|
pulsar |
why would you need a threadlocal variable to be serializable? if i use tl in a webapp, i usually put the (de)init into a filter and make 10000% sure i do not forget to clear it after the request is done. |
| 08:29 |
|
aditsu_ |
because I was thinking of making it non-static in a converter class, and the converter class has to be serializable; anyway, one option is to make the threadlocal static and hold a map |
| 08:39 |
|
aditsu_ |
pulsar: something like this: http://dpaste.com/2RR8TP6 |
| 08:40 |
|
pulsar |
ic |
| 08:43 |
|
pulsar |
just out of curiosity, why do you need the time/date/number formats to be TL? I am thinking @context and httpsession/headers/cookies in a similar helper class to determine the format associated with the request / user. |
| 08:44 |
|
aditsu_ |
I'd like to reuse them (for performance reasons) and they are not thread safe |
| 08:45 |
|
aditsu_ |
or do you think the performance impact is negligible? |
| 08:45 |
|
pulsar |
pretty much. yeah. |
| 08:46 |
|
aditsu_ |
if so, I could just create it on the spot in getDateFormat |
| 08:46 |
|
pulsar |
if in doubt, i tend to go with the cacheloader from guava |
| 08:46 |
|
pulsar |
keying will be an issue though, so you wont gain much |
| 08:47 |
|
aditsu_ |
http://askldjd.com/2013/03/04/simpledateformat-is-slow/ |
| 08:48 |
|
aditsu_ |
actually, that doesn't really talk about the object creation |
| 08:48 |
|
pulsar |
yeah, that what i was thinking. |
| 08:49 |
|
pulsar |
depens on how often you need to format dates |
| 08:49 |
|
pulsar |
and how it relates to the overall application |
| 08:49 |
|
pulsar |
from my experience, dateformat will be the last of your concerns in a web application |
| 08:50 |
|
pulsar |
and you could also cache the formatted strings by wrapping the df in a custom class with a TL cache or something |
| 08:50 |
|
pulsar |
but all of that would fall into the pre-optimization category |
| 08:50 |
|
pulsar |
99% of the tiime it is a waste of latter. |
| 08:52 |
|
aditsu_ |
this answer http://stackoverflow.com/a/22100487/179864 says "Creating an instance of SimpleDateFormat is very expensive" but it references the same link I posted above o_O |
| 08:53 |
|
aditsu_ |
let me do a quick benchmark :p |
| 08:54 |
|
pulsar |
i would make the benchrmark when you have something to compare order of magnitude of the numbers against |
| 08:54 |
|
pulsar |
like db queries / persistence layer doing something |
| 08:54 |
|
pulsar |
i still bet, that it will be a couple of orders of magnitude less significant than your persistence layer |
| 08:57 |
|
aditsu_ |
I can create 50k SimpleDateFormat's and add them to a list in less than 0.5s, but 100k are taking around 2.5s for some reason :o |
| 08:58 |
|
aditsu_ |
maybe gc kicking in or something? |
| 08:58 |
|
aditsu_ |
anyway, sounds fast enough |
| 09:03 |
|
pulsar |
yeah, you need to watch the gc spaces to understand the numbers. large number of objects does usually overflow the youngspace |
| 09:03 |
|
pulsar |
java is pita for synthetic benchmarks usually. |
| 09:04 |
|
pulsar |
and even non synthetic, chances are that some tweaks to the gc / memory ratios will give you a hunge performance boost are not that rare in a massive multithreaded application |
| 09:28 |
|
aditsu_ |
oh great, I got my new fancy date converter ready and it's not using it for some reason :/ |
| 09:28 |
|
aditsu_ |
debug time.. |
| 09:31 |
|
aditsu_ |
ah, a matter of java.util.Date.class vs java.sql.Timestamp.class |
| 12:03 |
|
|
Black-Ridder joined ##friendlyjava |
| 12:03 |
|
Black-Ridder |
bonjour :) |
| 12:04 |
|
Black-Ridder |
j'ai un petit pépin avec un code qui plante sous android studio http://pastebin.com/cdeU3b0F |
| 12:04 |
|
Black-Ridder |
dès que j'appelle cette fonction l'application plante |
| 14:40 |
|
aditsu |
25 minutes.. |
| 15:46 |
|
pdurbin |
pardon my French |
| 16:26 |
|
|
mr_lou joined ##friendlyjava |
| 22:02 |
|
|
pdurbin_m joined ##friendlyjava |
| 22:02 |
|
pdurbin_m |
Java EE Platform on Twitter: "New web site for JSON-B launched! #JavaEE8 https://t.co/E43ar51gHS": https://mobile.twitter.com/Java_EE/status/776094327592198144 |
