awful programming: feature flags and suffering

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > YOSPOS > awful programming: feature flags and suffering

«‹›1121 »

Carthag Tuek: Oct 15, 2005; Tider skal komme,
tider skal henrulle,
sl�gt skal f�lge sl�gters gang

Nomnom Cookie posted:

would you accept 128 MBps

i would accept them,

# ? Dec 21, 2019 12:55

Adbot: ADBOT LOVES YOU

# ? Jun 8, 2024 00:07

Carthag Tuek: Oct 15, 2005; Tider skal komme,
tider skal henrulle,
sl�gt skal f�lge sl�gters gang

i think i weigh almost a megagibigram? but bits instead of bytes? you know what i mean? circa

# ? Dec 21, 2019 12:59

jesus WEP: Oct 17, 2004

Krankenstyle posted:

i think i weigh almost a megagibigram? but bits instead of bytes? you know what i mean? circa

i know a belt won�t go all the way circa your waist

# ? Dec 21, 2019 13:01

Carthag Tuek: Oct 15, 2005; Tider skal komme,
tider skal henrulle,
sl�gt skal f�lge sl�gters gang

jesus WEP posted:

i know a belt won�t go all the way circa your waist

it do

idk its a pretty long belt, like 90 cm. rack belts are lik 75-80 cm :fatbert:

# ? Dec 21, 2019 14:27

Soricidus: Oct 21, 2010; freedom-hating statist shill

put a girdle round your waist in forty inches

# ? Dec 21, 2019 14:31

Doom Mathematic: Sep 2, 2008

A centbimeter is 1/128th of a meter.

# ? Dec 21, 2019 14:31

Phobeste: Apr 9, 2006; never, like, count out Touchdown Tom, man

Doom Mathematic posted:

A centbimeter is 1/128th of a meter.

i think cebimeter mostly because it's more upsetting to say

# ? Dec 21, 2019 14:42

Carthag Tuek: Oct 15, 2005; Tider skal komme,
tider skal henrulle,
sl�gt skal f�lge sl�gters gang

i prefer to call it 7.8 debimeters

# ? Dec 21, 2019 14:47

Carthag Tuek: Oct 15, 2005; Tider skal komme,
tider skal henrulle,
sl�gt skal f�lge sl�gters gang

i got it: i weigh one eigth of a megabig

# ? Dec 21, 2019 14:50

pokeyman: Nov 26, 2006; That elephant ate my entire platoon.

Krankenstyle posted:

i prefer to call it 7.8 debimeters

that's what i owe my credibitors

# ? Dec 21, 2019 16:04

Carthag Tuek: Oct 15, 2005; Tider skal komme,
tider skal henrulle,
sl�gt skal f�lge sl�gters gang

https://www.youtube.com/watch?v=d2tBhaVEWGM

# ? Dec 21, 2019 16:18

Carthag Tuek: Oct 15, 2005; Tider skal komme,
tider skal henrulle,
sl�gt skal f�lge sl�gters gang

pro tip: put milk in the gears, it makes them slower

call it lactasafFUC call it sabotage but milky

# ? Dec 21, 2019 16:20

gonadic io: Feb 16, 2011; >>=

rethinking making this vs code extension now

# ? Dec 22, 2019 16:50

Soricidus: Oct 21, 2010; freedom-hating statist shill

gonadic io posted:

rethinking making this vs code extension now

lol

webdev gonna webdev

# ? Dec 22, 2019 16:53

matti: Mar 31, 2019

regex is the worst PL

# ? Dec 22, 2019 17:27

gonadic io: Feb 16, 2011; >>=

i'm going mad here dealing with escaping of escaping

my grammar has

code:

		"keywords": {
			"patterns": [{
				"name": "keyword.other.dominionsmod",
				"match": "\\b#icon\\b"
			}]
		},

but I see these results: (no indentation or whitespace before the top 2 lines)

i thought \b, or word boundary, included start of line and whitespace. all the examples of keywords say to use it.

# ? Dec 22, 2019 18:01

gonadic io: Feb 16, 2011; >>=

Even weirder, my numbers are defined as

code:

		"numbers": {
			"patterns": [{
				"name": "constant.numeric.integer.decimal.dominionsmod",
				"match": "\\b[[:digit:]]+\\b"
			}]
		},

and those work just fine

# ? Dec 22, 2019 18:13

gonadic io: Feb 16, 2011; >>=

okay turns out that the example I was following was just wrong and textmate recommends you do this instead:

code:

		"keywords": {
			"patterns": [{
				"name": "keyword.other.dominionsmod",
				"match": "\\s*#icon\\b\\s*"
			}]
		},

so i don't know if \b behaves differently compared to normal regex, or if i never understood what it was supposed to do in the first place, but whatever

# ? Dec 22, 2019 18:40

Jabor: Jul 16, 2010; #1 Loser at SpaceChem

The issue there is that # is not a "word character". \b looks for the boundary between words and non-words, so your first regex is looking for #icon immediately after some word characters.

If you don't care about getting some superfluous matches then you could just look for #icon.

# ? Dec 23, 2019 00:33

Powerful Two-Hander: Mar 10, 2004; Mods please change my name to "Tooter Skeleton" TIA.

gently caress regexes

# ? Dec 23, 2019 00:55

akadajet: Sep 14, 2003

Powerful Two-Hander posted:

gently caress regexes

truth

# ? Dec 23, 2019 01:41

gonadic io: Feb 16, 2011; >>=

I would absolutely love a better way to specify blazingly fast DFAs. Maybe I should give nom another go although in this case it wasn't an option since I was firmly in the vscode ecosystem.

# ? Dec 23, 2019 02:24

Presto: Nov 22, 2002; Keep calm and Harry on.

gonadic io posted:

i'm going mad here dealing with escaping of escaping

I wrote a line once that had like literally dozens of backslashes in it. It was a command string that was getting passed around from shell to subshell to sub-subshell, with whole thing wrapped in a Python script.

I put a comment in saying basically I can't explain how this works.

# ? Dec 23, 2019 04:36

Xarn: Jun 26, 2015

gonadic io posted:

I would absolutely love a better way to specify blazingly fast DFAs. Maybe I should give nom another go although in this case it wasn't an option since I was firmly in the vscode ecosystem.

Are you sure that you even get DFAs?

This is your regularly scheduled reminder that most popular regex implementations use backtracking.

# ? Dec 23, 2019 10:38

jesus WEP: Oct 17, 2004

Powerful Two-Hander posted:

gently caress regices

# ? Dec 23, 2019 10:38

gonadic io: Feb 16, 2011; >>=

Xarn posted:

Are you sure that you even get DFAs?

This is your regularly scheduled reminder that most popular regex implementations use backtracking.

It was my understanding that if you stick to the regular parts of the language and don't use lookahead or lookbehind you were fine. These aren't user controlled.

# ? Dec 23, 2019 10:52

Xarn: Jun 26, 2015

An implementation needs special casing for that. I am not saying it doesn't, I don't use whatever VSCode thingy you are writing for, but it is worth keeping it in mind.

# ? Dec 23, 2019 11:21

Zlodo: Nov 25, 2006

i did a syntax highlighting thing for vscode once and it was awful

they reused a system developed for the textmate editor (that i never heard of other than for this) which itself used a regex library developed by some random dude

the docs for vscode syntax highlighting points you to the textmates docs which themselves contain a copy of that regex lib docs bc it was hosted on geocities and its long gone

# ? Dec 23, 2019 11:26

gonadic io: Feb 16, 2011; >>=

Zlodo posted:

i did a syntax highlighting thing for vscode once and it was awful

they reused a system developed for the textmate editor (that i never heard of other than for this) which itself used a regex library developed by some random dude

the docs for vscode syntax highlighting points you to the textmates docs which themselves contain a copy of that regex lib docs bc it was hosted on geocities and its long gone

this was my experience too. somebody is releasing a textmate 2 so all the textmate 1 docs and half the search results point to it but, unfortunately, it is empty and has nothing but seo inducing headers.

# ? Dec 23, 2019 11:28

gonadic io: Feb 16, 2011; >>=

Xarn posted:

An implementation needs special casing for that. I am not saying it doesn't, I don't use whatever VSCode thingy you are writing for, but it is worth keeping it in mind.

Does it though? I've done (non optimised) implementations before and you have a function that can return a list of results but (with my regexes) that list will always have length 1. An NFA that only ever has one state surely isn't _that_ much slower than a proper DFA?

Hmm, now I kind of want to test what speeds I can get out of an underfeatured actually regular regex library

# ? Dec 23, 2019 11:55

MrMoo: Sep 14, 2000

Woohoo, I�ve gone and done it again. Webdev world hits scrolling ticker technology again. So first version was using WebAnimations for CPU controlled GPU animation but suffered from Chome overheads, getting better and more popular now: Apple uses the tech quite frequently.

Version 2.0 as covered in a long projects post moved to WebGL using PIXI.js to paint to a canvas and animate in GL land. This is on NYSE floor and outside on broad street and works well if you have a nice speedy CPU to GPU memory bus.

Version 3.0 rips out the canvas and uses multi-signed distance fields (MSDF) from a paper by Valve that uses a font texture atlas with smarty pants magic to make text scale up biggly. So now there are no canvas elements in play, everything traditional GL stuff and can have shaders thrown around at will. So I took the Unreal Bloom Pass and made a nice neon effect sports ticker, even works on Firefox.

http://ahyoomee.miru.hk/zignage-sports7/build/default/www/3840-bloom.html

edit: adding craptacular gif:

Here's the MSDF font image:

The server side software is the same spectacular junk as NYSE, a Node.JS message router, :lol:

together with a dedicated feed handler for a sports data service. I went a bit OT and separated out the data schema for each sport and implemented a runtime agent to convert each one as a child process so that it can be dynamically updated without restarting the main process.

Client side software is a multi-threaded monster, the renderer thread uses Polymer and LitHTML x LitElement as a basic container for a webcomponent. The graphics are all in THREE.js with modified MSDF code to work with ES6 imports, and Bloom Pass to work with global THREE include. The renderer thread talks to a series of worker threads that each manage the business logic behind each tile in the ticker, all using ES10 with polyfill for TC39 extension. Basically the main ticker subscribes to list of games for each sport and creates a webcomponent for each game that then subscribes to that games data and manages its display. The comms is on an additional thread via websocket shared between all the worker threads and communicating via Google's Comlink, a convenient wrapper around PostMessage and friends.

MrMoo fucked around with this message at 06:41 on Dec 24, 2019

# ? Dec 23, 2019 19:12

Nomnom Cookie: Aug 30, 2009

gonadic io posted:

Does it though? I've done (non optimised) implementations before and you have a function that can return a list of results but (with my regexes) that list will always have length 1. An NFA that only ever has one state surely isn't _that_ much slower than a proper DFA?

Hmm, now I kind of want to test what speeds I can get out of an underfeatured actually regular regex library

an NFA that's only ever in 1 state at a time is a DFA

# ? Dec 23, 2019 20:21

FlapYoJacks: Feb 12, 2009

Powerful Two-Hander posted:

gently caress regexes

I use regex for two things and two things only:

Find a pattern and everything after it:
IE: foo.*

Find all blank lines:
^(?:[\t ]*(?:\r?\n|\r))+

# ? Dec 23, 2019 20:30

akadajet: Sep 14, 2003

ratbert90 posted:

I use regex for two things and two things only:

Find a pattern and everything after it:
IE: foo.*

Find all blank lines:
^(?:[\t ]*(?:\r?\n|\r))+

wow, you found 2 bad use cases for regexes

# ? Dec 23, 2019 20:38

brand engager: Mar 23, 2011

# ? Dec 23, 2019 22:55

FlapYoJacks: Feb 12, 2009

akadajet posted:

wow, you found 2 bad use cases for regexes

Oh lord, I don't use them in code, I use them in editors. I'm not a monster.

# ? Dec 23, 2019 23:04

gonadic io: Feb 16, 2011; >>=

Nomnom Cookie posted:

an NFA that's only ever in 1 state at a time is a DFA

yes, this was in response to:

gonadic io posted:

It was my understanding that if you stick to the regular parts of the language and don't use lookahead or lookbehind you were fine. These aren't user controlled.

Xarn posted:

An implementation needs special casing for that. I am not saying it doesn't, I don't use whatever VSCode thingy you are writing for, but it is worth keeping it in mind.

the NFA engine itself wasn't necessarily built with the knowledge that I am not going to give it a non-regular regex

# ? Dec 23, 2019 23:14

Tanners: Dec 13, 2011; woof

MrMoo posted:

Woohoo, I�ve gone and done it again. Webdev world hits scrolling ticker technology again. So first version was using WebAnimations for CPU controlled GPU animation but suffered from Chome overheads, getting better and more popular now: Apple uses the tech quite frequently.

Version 2.0 as covered in a long projects post moved to WebGL using PIXI.js to paint to a canvas and animate in GL land. This is on NYSE floor and outside on broad street and works well if you have a nice speedy CPU to GPU memory bus.

Version 3.0 rips out the canvas and uses multi-signed distance fields (MSDF) from a paper by Valve that uses a font texture atlas with smarty pants magic to make text scale up biggly. So now there are no canvas elements in play, everything traditional GL stuff and can have shaders thrown around at will. So I took the Unreal Bloom Pass and made a nice neon effect sports ticker, even works on Firefox.

http://ahyoomee.miru.hk/zignage-sports7/build/default/www/3840-bloom.html

Neat

# ? Dec 23, 2019 23:35

Jabor: Jul 16, 2010; #1 Loser at SpaceChem

a backtracking regex implementation won't use an nfa or dfa regardless of what type of regex you give it.

you can create pathological regexes that exhibit exponential behaviour on a backtracking implementation that are completely regular

# ? Dec 24, 2019 02:16

Adbot: ADBOT LOVES YOU

# ? Jun 8, 2024 00:07

Nomnom Cookie: Aug 30, 2009

https://www.regular-expressions.info/catastrophic.html

# ? Dec 24, 2019 04:28

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > YOSPOS > awful programming: feature flags and suffering

«‹›1121 »