#twisted

Tickets pending review: https://tm.tl/#9299, #9305, #9333, #9337, #9338, #9340, #9343 (rodrigc), #9350 (ktdreyer), #9352, #9210 (markrwilliams), #9358, #8966 (the0id), #9100, #9118 (the0id), #4964 (jameshilliard), #9138, #9176, #7495

is twisted.protocols.basic.LineReceiver ok to use with clients or is it just for servers?

hmm, nmd. if I'm reading the docs right it should be fine. problem must be elsewhere

Yeah it's good for both.

ok, so new question then. what am I doing wrong with sending data here: https://github.com/ntchambers/cherrybomb/blob/m... I get this error: http://ix.io/Dqm

I think it's self.sendLine

but i didnt lookat your code. hold on

it is

writeLine is my own thing that calls sendLine

ah, gotcha

nchambers: I don't understand why you are pretty much alias'ing sendLine to writeLine

honestly it was just to separate everything to see if I could spot the issue

I'm kinda confused why it says self.transport doesnt have a write method

gah, let me download

yeah I don't understand what its complaining about

Ehhh I dont have the other files

well

just separate out the other values

get rid of writeLine

I was getting the error before that

Don't care. Isolate variables.

make it as simple as you can

print out values

I had it and again have it as simple as possible

self.sendLine('whatever') still throws the same error

print prints out expected values

I'm looking at code I wrote and uhh it works fine

I'm really confused

my ident server uses the same code more or less, and it works fine

self.sendLine("PROTOCTL EAUTH={}".format(self.factory.name))

Maybe you overwrote something?

oh my protocol can't access factory

could that be the problem?

pretty sure it cant be that

I have no idea what I'm doing wrong then

comment out everything in Irc and in __init__ jujst put self.sendLine("whatever")

see if it still breaks then

yep still breaks

isn't the .transport attribute added after __init__ runs?

o.o

ohhh is that it?

maybe move the call into connectionMade

so is it because

Ah, excellent point

Yeah I didn't even make that connection. connectionMade is typically used as the initializer for Protocols

also nchambers, just so you know, there's a specific IRC module for twisted

meh

I'd rather implement my own

thanks for the heads up though

well

I'm sorta-kinda implementing an IRC server

twisted kinda doesnte ven give the bare minimum to do this

but it does make it a lot easier

what with the prefix and params arguments

I recommend not reinventing the wheel

thanks for the recommendation. Still not going to use it

but I appreciate the heads up

why tho

because I want to write my own?

you are doing this: https://en.wikipedia.org/wiki/Begging_the_question

You can not say "I am feeding the dog because I am feeding the dog"

I am asking why you want to write your own.

I dont' really care your actual reasoning. I just hope it's for a good reason

because I do. I didn't realize I needed a reason to use twisted

again, that's beggining the question

And you can simply say "because I enjoy the challenge"

not really. you're just pestering for a reason when none exists

or "to learn how to make my own protocol"

you're overthinking this

Everything anyone does is for a reason.

you're welcome to that opinion

So...you're doing it for a shitty reason.

or I'm not really doing it for any reason at all and you're overthinking it

Making more work for yourself, deliberately slowing yourself down

Hey man, it doesn't matter to me. Don't get so defensive when someone is trying to help you. It's an important trait as a programmer to recognize when you're making unnecessary work for yourself, is all.

seeya

I'm not being defensive at all. I'm giving you the answer and you clearly don't want to accept it

if it doesn't matter to you, just stop asking then :)

reinventing the wheel is awesome.

if people didn't reinvent the wheel, we'd have bugatis on stone discs.

Good point, haha

some of my earliest hobby projects (as a self taught coder) were pointless reinvent-the-wheel type things. screensavers, irc clients, email clients, etc.

some people learn well with hands-on experience, even if they're "doing it wrong" (i definitely was, in hindsight), etc.

also, happy festivus, etc.

Yes, I agree with that. I think doing it as a learning exercise is valid.

But projects done as learning exercises tend not to be coded as well as the already-invented-wheel

Tickets pending review: https://tm.tl/#9299, #9305, #9333, #9337, #9338, #9340, #9343 (rodrigc), #9350 (ktdreyer), #9352, #9210 (markrwilliams), #9358, #8966 (the0id), #9100, #9118 (the0id), #4964 (jameshilliard), #9138, #9176, #7495

totally agreed.

though occasionally they do provide a sandbox for experimenting with new approaches to already-"solved" problems, etc.

it is also possible to do things for fun

rather than insisting that everything must accomplish some economically productive use either directly or indirectly (e.g. learning)

however this attitude is distinctly suspicious and un-American

yes for learning or fun, either answer i would have been more satisfied with than

"because i wanted to"

Tickets pending review: https://tm.tl/#9299, #9305, #9333, #9337, #9338, #9340, #9343 (rodrigc), #9350 (ktdreyer), #9352, #9210 (markrwilliams), #9358, #8966 (the0id), #9100, #9118 (the0id), #4964 (jameshilliard), #9138, #9176, #7495

seanicus: it is still kind of rude to demand people explain their motivations and then judge them for not meeting your standards

What is a reliable method of checking twisted's version, so that I can fail my app is it is too old?

sveinse: pinning from below in your setup.py

tos9: sure, but that moves the resposibility down to the package level. Do I understand this that is implies that one usually don't do application based version checks?

sveinse: correct

using linter is a two edged sword :o :). Especially when twisted use camel cases...

njs: not demanding, just asking, then subsequently getting annoying that he gave me a nonanswer

hi, i'm researching about the use of PyPy and twisted, and reached this 5 year-old bug: https://twistedmatrix.com/trac/ticket/6278

someone knows if there's been any advances in either getting benchmarks for this or if PyPy might have changed in any way that made the issue less problematic?

(i don't really understand the whole issue at hand yet :)

drebs: Don't know if anything's changed, but I (we've) been running Twisted on PyPy in production for awhile now, and others (like twisted's webpage itself) for even longer, if you're asking a more general question.

tos9, do you make heavy use of inlineCallbcacks?

it seems that this is the possible problematic point, because of the way defer.returnValue is implemented

Will callLater(0,fn) run the fn immediately, as in when callLater(0,..) returns the fn has been run? Or does it schedule the fn to run immediately once the execution comes back to the event loop?

drebs: I don't, I try to avoid it for other reasons (it makes diagnosing errors harder)

But one or two places in prod code use it, and those places are likely hot, so yeah it is in some sort of use

sveinse: I forget, but I think the latter

sveinse: should be reasonably easy to check if callLater special cases 0 (or with a debugger)

tos9, thanks a lot for the info

np

Tickets pending review: https://tm.tl/#9299, #9305, #9333, #9337, #9338, #9340, #9343 (rodrigc), #9350 (ktdreyer), #9352, #9210 (markrwilliams), #9358, #8966 (the0id), #9100, #9118 (the0id), #4964 (jameshilliard), #9138, #9176, #7495

Is there a global way to catch or hook errors that occurs in the functions of callLater or loopingCall? I.e. be notified of unhandled exceptions in Deferred objects?

The challenge with the behaviour is that a Deferred that fails, dumps its traceback to stderr and stops, but overall the reactor continues as nothing has happened. I the application could be notified, it could take appropriate action

sveinse: if you know how to do that then just add the appropriate errbacks :)

sveinse: the problem with this is that you can't know if a Deferred is actually dead or not, until it gets garbage collected and is found to be failed.

because code could always add an errback later, and then the Deferred would have a value again

teratorn: firstly, in callLater() or loopingCall() I'm oblivious to the backend implementation using Deferreds. Secondly, I do agree, you could and perhaps should try catch everything that can go wrong, but that creates quite bloated code in a large application

It would be nice if there were a reactor hook when a errback is being collected without a handler. This way the application could be informed that something is not right

erm, Deferred has an __del__ method that prints the unhandled error in Deferred...

that is where Unhandled error in Deferred comes from. it happens at GC time