-
-
Notifications
You must be signed in to change notification settings - Fork 507
Conversation
Not to add a new npm dependency, I can use https://gist.github.com/brettz9/4212262, Where dya want me to put it @mcollina ? |
You can define it as a function somewhere, but don't polyfill array. |
Why a fail? Can't see travis logs... :\ |
@davedoesdev What happens to QlobberDedup on node < 4 with no ES6 |
david@david-Latitude-E6440:~$ nvm use 0.12
Now using node v0.12.11 (npm v2.14.9)
david@david-Latitude-E6440:~$ node --version
v0.12.11
david@david-Latitude-E6440:~$ node
> Set
[Function: Set] |
Mosca has 10, 12, 4 & 5: https://github.com/mcollina/mosca/blob/master/.travis.yml |
qlobber already polyfills |
|
Will drop 0.10 support @mcollina ? |
Why does this PR need to use |
Seems |
https://github.com/caolan/async#collections-1
|
hmm... I should check why I was unable to iterate over them with |
async 1.5.x doesn't support them, updated async to latest |
Hi @behrad is performance acceptable now or are there still problems? |
Startup really improved @davedoesdev, I believe it was an effective change, Thank you for your nice and quick feedbacks. However we have stills performance issues in Mosca with lots of connecting/disconnecting clients (~10/seconds) I'm trying to check timeout management, sub trie add/removes and the way Mosca is working at last, in isolated tests. |
I'm using this in my production, Will it be merged in @mcollina ? |
@behrad FYI I just published qlobber 0.7.0, which now uses ES6 Maps internally for the trie. It seems to give roughly 25% speedup from the benchmarks. |
Is there any upgrade notes I should care about? Replaced Sets with Maps? |
No - changes are all internal. |
this is ready @mcollina |
Awesome, I will assemble a new major as soon as I have a bit of time. |
Already merged in v2, about to be released. |
This eliminates match/add pairs when adding topics to tries, and mosca is behaving more performant, Fixes #428, and I believe it also should fix #426