by

Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
  • 08:41

    back2dos on master

    Make slice tink_json compatible. (compare)

  • 08:40

    back2dos on master

    Improve inference for custom wr… (compare)

  • Sep 15 10:03
    back2dos commented #76
  • Sep 15 09:59
    back2dos commented #76
  • Sep 15 09:26
    kevinresol commented #76
  • Sep 15 09:03
    back2dos opened #77
  • Sep 15 09:02
    back2dos commented #76
  • Sep 15 08:29
    kevinresol commented #76
  • Sep 15 08:06
    back2dos opened #76
  • Sep 14 09:59
    cedx commented #77
  • Sep 10 09:11
    back2dos commented #45
  • Sep 10 08:21
    kevinresol commented #45
  • Sep 10 08:20
    kevinresol commented #45
  • Sep 09 23:01
    qkdreyer opened #45
  • Sep 08 10:51
    back2dos opened #75
  • Sep 08 08:09

    back2dos on master

    Compatibility with tink_core v2. (compare)

  • Sep 07 11:13
    kevinresol edited #119
  • Sep 07 11:13
    kevinresol edited #119
  • Sep 07 11:13
    kevinresol opened #119
  • Sep 07 11:00

    kevinresol on generated

    Use tink_spatial Get rid of geojson completely Extract spatial value in a more… and 32 more (compare)

Peter Achberger
@Antriel
Seems like iterating Map is around 5 times slower than iterating an array.
Juraj Kirchheim
@back2dos
with an empty loop body, I guess that's accurate ;)
did you measure js.lib.Map or Map?
Peter Achberger
@Antriel
I measured JS directly.
new Map()
Maybe I messed up, not sure: https://jsben.ch/7IxI0
I suppose all I can do is use Map for lookups, but keep iterating and deleting from an array.
Juraj Kirchheim
@back2dos
Peter Achberger
@Antriel
indexOf probably gets a bit slow with thousands of entities. :D
Juraj Kirchheim
@back2dos
I wonder if an entity can be in multiple collections
Peter Achberger
@Antriel
It can.
Alright, nice bench, but not keys() but values() and val.val. Only twice slower.
That might actually be worth it.
Juraj Kirchheim
@back2dos
ah, right
Peter Achberger
@Antriel
Best is, I can implement a bit more of basic functionality, write some dumb bots, let it run for 10 minutes...
And then just replay that scenario as benchmark while changing the implementation.
Juraj Kirchheim
@back2dos
yeah, that might be a bit more indicative than a microbenchmark ^^
Peter Achberger
@Antriel
Indeed. :) And if I will have too much time (I don't ^ 3), I could replace the whole ECS with ECX and see what happens.
Just gonna need to bench on some different PC, my laptop is so memory bandwidth starved that when I was testing my networking code, what I thought was taking the most time ended up being negligible on a server.
Alright, I feel better now. Scary morning.
Thanks guys. :)
state.observe().bind({direct:true}, ...) seems to be some old syntax.
Juraj Kirchheim
@back2dos
ah, right ... now it's bind(..., tink.state.Scheduler.direct)
or well, with appropriate imports more like bind(..., direct)
Peter Achberger
@Antriel
And what's the difference between state.observe().bind() and state.bind(), and how does .nextTime() fall into it all?
But I guess I can just replace it with an array, it's just a few lines.
Juraj Kirchheim
@back2dos
the implementation of state.bind() is observe().bind() ^^
but State doesn't have all the APIs of Observable ... nextTime is only defined on the latter, hence .observe().nextTime()
Peter Achberger
@Antriel
Right, thanks. :)
Juraj Kirchheim
@back2dos
sure ;)
Peter Achberger
@Antriel
Hm, seems like tink_cli doesn't work with the new core. And I'm way too behind to properly debug it. :(
Juraj Kirchheim
@back2dos
could you be more specific? ^^
doesn't compile? just hangs at runtime?
Peter Achberger
@Antriel
Right, sorry. Doesn't really trigger any command.
Juraj Kirchheim
@back2dos
hmm, ok ... there's probably some next/map/flatMap somewhere that needs to become eager
Peter Achberger
@Antriel
Brief debugging seems to indicate it hangs on some promise.next(..., <never firing callback>).
Kevin Leung
@kevinresol
what is the exact breaking change in core regarding laziness?
Peter Achberger
@Antriel
How cool it would be to have something like inquirer.js with crossplatform Haxe. One can dream... :)
Juraj Kirchheim
@back2dos
everything is lazy ^^
Kevin Leung
@kevinresol
iirc only Future.async was eager?
Juraj Kirchheim
@back2dos
IIRC map/flatMap on FutureTrigger were always eager
Peter Achberger
@Antriel
With macro-based auto-build into tink_web simple frontend too. Could be doable with proper abstraction of commands and data display. And macro-in-macro maybe.
Kevin Leung
@kevinresol
don't think tink_cli has used triggers...
can you please raise an issue?
I will have a look
Peter Achberger
@Antriel
Sure, gonna make some minimal example to verify I'm not doing something silly.
Peter Achberger
@Antriel
The default example seems to run. Gonna check what's different in mine.
Alright, it was my fault. I've removed .handle(..) on the startpoint Cli.process(). It used to work, but now it's lazy I guess.
Sorry for the panic.
Kevin Leung
@kevinresol
good to hear