DDraceNetwork - developer

ChillerDragon, are you there?

Learath2

[ddnet-scripts:master] 1 new commit

9fcf4db Fix string interpolation mistakes - Learath2

@deen ^^ when you are around pls

@Learath2 hi

o/

done

@deen can you check error.log? I think I broke something with that

07:14

Ah, that should have been an elseif not an else if

07:15

hm, shouldn't need to be

def-

[ddnet-scripts:master] 1 new commit

dea35dc Fix teehistorian/index.php - def-

@Learath2 ^

oh the stupid closing brackets, I wonder why my IDE didn't complain at all

^[0-9a-fA-F]{8}\-(?:[0-9a-fA-F]{4}\-){3}[0-9a-fA-F]{12}$ is probably a better regex for uuids

you don't need to escape - outside of [] and not at the start/end of [] (edited)

^[0-9a-fA-F]{8}-(?:[0-9a-fA-F]{4}-){3}[0-9a-fA-F]{12}$ even better

[\da-fA-F]

[[:xdigit:]]

ew

09:40

lucky with grep u can use -E

you don't like posix character classes?

no

09:41

its ugly

09:41

imo pcre is the best regex

PCRE has posix character classes

but most of time u dont need them

09:43

This crate provides a library for parsing, compiling, and executing regular expressions. Its syntax is similar to Perl-style regular expressions, but lacks a few features like look around and backreferences.

rust regex

You never need them, they can make regex easier to read sometimes

why most languages cant implement a full pcre

I think php and ruby are pretty complete pcre implementations

backreferences are bad™ because they make the expressions no longer regular in the cs sense

09:49

that means they are no longer guranteed to run in linear time

09:49

(PCRE backreferences have exponential worst case)

hm, is it because with regexes are supposed to be equivalent to some DFA and with backreferences we can no longer map it to a DFA?

yes

s/with//

the name comes from CS, "regular languages" are exactly those that can be accepted by a DFA

I wonder how regex engines handle backreferences

backtracking

backtracking as in actually going back to when the match first happened?

no, backtracking as in "brute force with backtracking"

09:56

try to greedily match the regex, if it doesn't match, backtrack to the first position where you made a choice and make the next possible choice

hm, what choice are they making?

(a*)*

09:57

aaaaaaab

ah, how much to match

yes

09:58

(this is an example where PCRE chokes btw)

09:58

yes, this one is artificial, but people have been bitten by exponential regexes before (e.g. stackoverflow)

I see, I wonder if someone came up with a smart way to do non-regular expressions decently efficiently

https://stackstatus.net/post/147710624694/outage-postmortem-july-20-2016

Stack Exchange Network Status

Outage Postmortem - July 20, 2016

On July 20, 2016 we experienced a 34 minute outage starting at 14:44 UTC. It took 10 minutes to identify the cause, 14 minutes to write the code to fix it, and 10 minutes to roll out the fix to a...

This is what happens when you get too comfortable with features like these

with linear-time regex engines, you don't have to worry about it

I always assume a regex match is an order of magnitude more expensive then a more traditional approach

10:03

Sure that means I use less regex in my code, but whatever, I'd rather have predictable performance

linear-time regexes give you that predictable performance

yes, but say we use a linear-time regex engine on ddnet, then I decide to parse something in js and copy over that regex

yea

suddenly my linear-time regex can become an exponential cost

10:05

impressed by the response time by SX btw

34 minutes is impressive, indeed

They probably have some builtin profiling somewhere in the code, it'd be impossible to track it down blind

10:06

or maybe some tracing of calls into the underlying engine

my nginx will pass PHP_AUTH_USER just fine, the one on ddnet.tw wont

10:39

now it decided it'll start to pass it along

10:39

wtf?

hmmm

ah another http header oddity

10:45

yay to not upgrading decades old software because dinosaur forgotten servers we should still be able to connect to

elaborat

10:47

e

hm, now that I think about it it's more of a PHP oddity

10:49

as soon as you send 1 header, php will no longer touch the headers