cfbolz changed the topic of #pypy to: PyPy, the flexible snake (IRC logs: https://botbot.me/freenode/pypy/ ) | use cffi for calling C | mac OS and Fedora are not Windows
danieljabailey has joined #pypy
mattip has joined #pypy
mattip has quit [Ping timeout: 276 seconds]
jamesaxl has quit [Quit: WeeChat 2.2]
dddddd has quit [Remote host closed the connection]
jcea has quit [Remote host closed the connection]
_whitelogger has joined #pypy
mattip has joined #pypy
mattip__ has joined #pypy
mattip has quit [Ping timeout: 250 seconds]
forgottenone has joined #pypy
mattip__ has left #pypy [#pypy]
tayfun26 has joined #pypy
glyph has quit [Quit: End of line.]
glyph has joined #pypy
_whitelogger has joined #pypy
_whitelogger has joined #pypy
jamesaxl has joined #pypy
Garen_ has joined #pypy
Garen has quit [Ping timeout: 245 seconds]
jacob22__ has quit [Quit: Konversation terminated!]
fryguybo1 has joined #pypy
fryguybob has quit [Ping timeout: 252 seconds]
smvv has joined #pypy
dddddd has joined #pypy
themsay has quit [Ping timeout: 276 seconds]
lritter has joined #pypy
jcea has joined #pypy
nimaje has quit [Ping timeout: 252 seconds]
nimaje has joined #pypy
forgottenone has quit [Ping timeout: 240 seconds]
demonimin has quit [Ping timeout: 252 seconds]
moei has joined #pypy
demonimin has joined #pypy
demonimin has joined #pypy
<dstufft>
Is there a good place to look/track 3.6 support for PyPy3?
<fijal>
dstufft: "nearly done"? ;-)
<dstufft>
fijal: that's pretty cool! I'm thining about poking at Warehouse/PyPI on PyPy again. I think it could be a solid win for speeding up our XMLRPC code (short story, xmlrpc has a big ole list, appends a bunch of str to that list, then joins them. That is taking up ~3k-4k ms atm)
<dstufft>
We're currently on CPython 3.6
<fijal>
ask mattip or ronan they will know
<dstufft>
gonna poke at it a bit a tleast, I don't *think* we're using too many 3.6-isms in Warehouse, besides f strings which I think PyPy 3.5 supports anyways
<Alex_Gaynor>
fijal: I already told dstufft it'll be way faster withh StringBuilder :-)
<dstufft>
Sounds like a thing I'd submit a patch to PyPy for
<dstufft>
since xmlrpc is part of the stdlib
<dstufft>
(unless PyPy doesn't patch the stdlib, I dunno)
<Alex_Gaynor>
yeah. carrying patches to the pure python parts of the stdlib is annoying, but it'd be reasonable
<fijal>
it does in places
<fijal>
since some parts of stdlib are like OMG why?
<dstufft>
worst case, technically the renderer is pluggable, so I can make my own
<fijal>
dstufft: I'm happy to help you optimize at some stage
<fijal>
but not right now, i'm on a plane as of very soon
<fijal>
which means I should likely board or something
<dstufft>
fijal: have a good flight!
adamholmberg has joined #pypy
<dstufft>
oh man, keyword only arguments *and* trailing comma in a function sig don't work together (apparently this is true in CPython 3.5 as well?!) That's super weird, no idea that edge case even existed
<_aegis_>
what do you mean
<dstufft>
def foo(*, bar,): -> Invalid Syntax, but def foo(bar,): and def foo(*, bar): are fine
<dstufft>
(This isn't PyPy specific, just noticed it because PyPy is still 3.5)
<dstufft>
Alex_Gaynor: fijal uhhhh, I think PyPy might be a 100x speedup for XMLRPC
<dstufft>
I need to go check my original numbers
<Alex_Gaynor>
lol
<dstufft>
I ran the list_packages() api in a loop and was getting < 30ms
<dstufft>
I'm pretty sure it was like, 3k-4k ms on Cpython
<Alex_Gaynor>
pypy: it's kind of fast
<dstufft>
that's without any StringBuilder unless PyPy'x xmlrpclib already does that
<Alex_Gaynor>
dstufft: And I think switching to StringBuilder is probably worth another 10%+
<dstufft>
Ah right, my 3k number was with profling turned on, the <30ms PyPy number was with Profling turned off. It's actually a 10x speedup :P
<dstufft>
apparently profiling sucks for perf!
<gsnedders>
take away: never profile code
adamholmberg has quit [Remote host closed the connection]
adamholmberg has joined #pypy
adamholmberg has quit [Ping timeout: 252 seconds]
adamholmberg has joined #pypy
<fijal>
10x ain't bad
<dstufft>
fijal: Yea, 10x on our slowest routes (which constitute the bulk of the traffic that actually *hits* the Python service)