awful programming: feature flags and suffering

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

«‹›1116 »

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

does our resident stan.yaml consider themselves an experienced yaml engineer?

# ? Jan 14, 2023 05:03

Adbot: ADBOT LOVES YOU

# ? May 2, 2024 06:44

12 rats tied together: Sep 7, 2006

no, it's not possible to "engineer yaml" because it is just a markup language. it doesn't do anything by itself

person who authored this article is likely just a dumbass although i don't recall that part of it so maybe mocking yaml engineers is the 1 correct thing in it

# ? Jan 14, 2023 06:20

Share Bear: Apr 27, 2004

html

# ? Jan 14, 2023 06:20

Bloody: Mar 3, 2013

I updated some appveyor yaml today and it was a breeze. just copy pasted chunks from another project ezpz

# ? Jan 14, 2023 06:22

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

cinci zoo sniper posted:

https://ruudvanasseldonk.com/2023/01/11/the-yaml-document-from-hell

this guys name is Ruud Van Asseldonk

e: I hadn't actually read the article until now and wow that is the dumbest poo poo I ever read

Powerful Two-Hander fucked around with this message at 09:51 on Jan 14, 2023

# ? Jan 14, 2023 09:46

Sagacity: May 2, 2003; Hopefully my epitaph will be funnier than my custom title.

a perfectly cromulent Dutch name

# ? Jan 14, 2023 09:47

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

12 rats tied together posted:

what's the difference between the yaml 1.1 schema assuming implicitly that nn:nn is for time, and your ideal scenario here?

how can a generic wire format/markup language have a schema that is simultaneously not implicit, even though !!str is apparently an unreasonable ask, but also supports values that are of an arbitrary type

this feels like a fake complaint. the article starts off by examining base 60 numbers that end up in fields in the docker api. this happens because they use the generic fallback schema from 2005 and because, in that schema, they don't specify a type for their values, and the schema resolves them to something that was not intended by the human

is the problem that there even is a fallback schema? your ideal yaml fails to serialize unless you specify that you're using tags.yaml.org:dockerContainer? i don't necessarily disagree, just trying to understand

?????

by �schema� I mean like an xml schema or a json schema. a thing that defines the content for a particular type of document, like what keys are valid and what types their values have.

it may be named in the document but ideally the application will know which schema to apply because it�s the one that describes the data the application knows how to use. it may even be generated automatically from the objects being serialised, but if so it will still be provided to the parser to ensure values are parsed correctly, or at least to identify and explain cases where something was not the expected type.

yaml per se does not have a schema at all. a document with explicit types like you use has an implicit single-use schema, but that�s worthless for validation because it just describes what�s there, not what *should* be there.

use a drat schema, whatever serialisation format you choose.

# ? Jan 14, 2023 12:03

Xarn: Jun 26, 2015

12 rats tied together posted:

i have good news for you then, the failsafe schema in yaml 1.2 (2009) only has string, sequence, and mapping

you can consult your library for details on how to parse documents with this schema

Genuinely glad to hear that, but we don't ingest yaml and 3rd party places that do use various incompetent libraries :shrug:

# ? Jan 14, 2023 13:49

mystes: May 31, 2006

12 rats tied together posted:

no, it's not possible to "engineer yaml" because it is just a markup language. it doesn't do anything by itself

Bridges aren't turning complete either

# ? Jan 14, 2023 14:47

jony neuemonic: Nov 13, 2009

Bloody posted:

I would simply use xml

hell yeah brother

# ? Jan 14, 2023 15:48

matti: Mar 31, 2019

Powerful Two-Hander posted:

this guys name is Ruud Van Asseldonk

lol

# ? Jan 14, 2023 15:58

CPColin: Sep 9, 2003; Big ol' smile.

matti posted:

lul

Fixed

# ? Jan 14, 2023 17:26

12 rats tied together: Sep 7, 2006

Soricidus posted:

?????

by �schema� I mean like an xml schema or a json schema. a thing that defines the content for a particular type of document, like what keys are valid and what types their values have.

it may be named in the document but ideally the application will know which schema to apply because it�s the one that describes the data the application knows how to use. it may even be generated automatically from the objects being serialised, but if so it will still be provided to the parser to ensure values are parsed correctly, or at least to identify and explain cases where something was not the expected type.

yaml per se does not have a schema at all. a document with explicit types like you use has an implicit single-use schema, but that�s worthless for validation because it just describes what�s there, not what *should* be there.

use a drat schema, whatever serialisation format you choose.

It does have schemas, a wire format only speaks byte array, it's not possible to parse without a schema. The thing in 1.1 that goes "22:22? Must be base 60" is a schema. PyYAML (which the article author was complaining about) has several loader classes that have their own schemas embedded. The golang lib, according to the article author, chose to implement its own custom version of the 1.1 schema.

A rare example of a good yaml lib, the crystal stdlib has an entire namespace for interacting with YAML schemas and instructions for registering your own types, in a handy macro. But even PyYAML, widely regarded as absolute dogshit, provides these instructions in the section "Constructors, representers, resolvers", sadly with a much worse API.

If your definition of "a schema" is "a file that defines a schema", these absolutely exist in YAML, but they are of course the same thing. How else would an XML parser load additional schema machinery from an xsd file? Is it schema files all the way down?

# ? Jan 15, 2023 00:32

Dijkstracula: Mar 18, 2003; You can't spell 'vector field' without me, Professor!

(editing to be more precise)

I don't see a user-defined schema used in that crystal-lang url: the other poster is talking about the reverse of:

code:

require "yaml"

string = YAML.build do |yaml|
  yaml.mapping do
    yaml.scalar "foo"
    yaml.sequence do
      yaml.scalar 1
      yaml.scalar 2
    end
  end
end
string # => "---\nfoo:\n- 1\n- 2\n"

ie. a variation on parse(data) that takes a user-defined schema indicating "the thing I'm parsing must contain a scalar named "foo", a sequence of pairs of scalars" or whatever. By contrast, here's an example of an XML schema validating a particular class of XML document. Sounds like you're confusing this with the grammar for YAML itself maybe?

12 rats tied together posted:

If your definition of "a schema" is "a file that defines a schema", these absolutely exist in YAML, but they are of course the same thing. How else would an XML parser load additional schema machinery from an xsd file? Is it schema files all the way down?

Not if something self-describing is at the bottom.

Dijkstracula fucked around with this message at 00:57 on Jan 15, 2023

# ? Jan 15, 2023 00:44

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

Ultimately, the type that a piece of data should be is not defined by the config file itself - it's defined by what the application intends to do with that data. Sometimes that's explicit in the form of a schema, sometimes that's implicit in that the application will just do the wrong thing if it's not the data type that the application author intended.

If you allow your parsing library to choose data types for a parse based on what's in the file being parsed rather than based on what the application intends to do with that data, the only thing you gain is the possibility for type mismatches causing undesired behaviour. There's no upside.

# ? Jan 15, 2023 01:20

12 rats tied together: Sep 7, 2006

Dijkstracula posted:

By contrast, here's an example of an XML schema validating a particular class of XML document.

As the link indicates, this is an XSD file, which is XML that can be interpreted to configure machinery (additional types, and restrictions on values for those types) for other XML documents. There used to exist other ways of validating XML than XSD, some of them might even be around still.

Some ways you might produce a file that defines an additional set of types (in yaml: "application specific tags") and restrictions on values for these types would be: json-schema, kwalify, rx, yamale. There are a bunch of others in various levels of popularity. Most of them are not very popular, none of them are YAML in the same way that XSD is not XML.

edit: I guess json-schema is pretty popular actually.

Jabor posted:

Ultimately, the type that a piece of data should be is not defined by the config file itself - it's defined by what the application intends to do with that data. Sometimes that's explicit in the form of a schema, sometimes that's implicit in that the application will just do the wrong thing if it's not the data type that the application author intended.

If you allow your parsing library to choose data types for a parse based on what's in the file being parsed rather than based on what the application intends to do with that data, the only thing you gain is the possibility for type mismatches causing undesired behaviour. There's no upside.

`<first-name>Benjamin</first-name>' and '!first-name Benjamin'. Why is one of these bad and one of them is good?

# ? Jan 15, 2023 01:29

Dijkstracula: Mar 18, 2003; You can't spell 'vector field' without me, Professor!

12 rats tied together posted:

As the link indicates, this is an XSD file, which is XML that can be interpreted to configure machinery (additional types, and restrictions on values for those types) for other XML documents. There used to exist other ways of validating XML than XSD, some of them might even be around still.

Some ways you might produce a file that defines an additional set of types (in yaml: "application specific tags") and restrictions on values for these types would be: json-schema, kwalify, rx, yamale. There are a bunch of others in various levels of popularity. Most of them are not very popular, none of them are YAML in the same way that XSD is not XML.

this is fine but I don't think it's invalidating anything we're saying :confused:

# ? Jan 15, 2023 01:31

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

12 rats tied together posted:

`<first-name>Benjamin</first-name>' and '!first-name Benjamin'. Why is one of these bad and one of them is good?

Do you actually think those are equivalent in how they're used?

# ? Jan 15, 2023 01:34

mystes: May 31, 2006

Dijkstracula posted:

this is fine but I don't think it's invalidating anything we're saying

Post the schema for your posts and we'll find out

# ? Jan 15, 2023 01:38

Brain Candy: May 18, 2006

Dijkstracula posted:

Not if something self-describing is at the bottom.

there's a turtle underneath your turtle and just because you stopped looking at this one does not make it the bottom

# ? Jan 15, 2023 01:43

12 rats tied together: Sep 7, 2006

Jabor posted:

Do you actually think those are equivalent in how they're used?

If OPs assertion was "the way that yaml is typically abused by morons and the freedom of the grammar that enables it, is bad" I would have simply posted the :hai:

emoji.

# ? Jan 15, 2023 01:48

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

12 rats tied together posted:

As the link indicates, this is an XSD file, which is XML that can be interpreted to configure machinery (additional types, and restrictions on values for those types) for other XML documents. There used to exist other ways of validating XML than XSD, some of them might even be around still.

Some ways you might produce a file that defines an additional set of types (in yaml: "application specific tags") and restrictions on values for these types would be: json-schema, kwalify, rx, yamale. There are a bunch of others in various levels of popularity. Most of them are not very popular, none of them are YAML in the same way that XSD is not XML.

edit: I guess json-schema is pretty popular actually.

`<first-name>Benjamin</first-name>' and '!first-name Benjamin'. Why is one of these bad and one of them is good?

the equivalent yaml to <first-name>Benjamin</first-name> is �first-name: Benjamin�. in both cases that is how you represent the value �Benjamin� being assigned to the first-name field.

idk how the hell �!first-name Benjamin� would be used. do you really have a dedicated first-name type that can be used for any value of any field?

# ? Jan 15, 2023 15:10

Antigravitas: Dec 8, 2019; Die Rettung fuer die Landwirte:

https://github.com/marciniuk/undertale/blob/master/scripts/SCR_TEXT.gml

# ? Jan 16, 2023 17:28

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

Antigravitas posted:

https://github.com/marciniuk/undertale/blob/master/scripts/SCR_TEXT.gml

oof thats like the code i wrote when i was 12

# ? Jan 16, 2023 20:16

Chalks: Sep 30, 2009

Carthag Tuek posted:

oof thats like the code i wrote when i was 12

game maker is a truly horrible language. it's like it's designed by someone who'd only ever used javascript, but didn't really understand even that.

# ? Jan 16, 2023 20:30

Plorkyeran: Mar 22, 2007; To Escape The Shackles Of The Old Forums, We Must Reject The Tribal Negativity He Endorsed

that's "decompiled" code and not actually the code in its original form. the giant switch is it packing all of the separately editable screens into a state machine and all the magic numbers were hopefully named constants

# ? Jan 16, 2023 20:48

Antigravitas: Dec 8, 2019; Die Rettung fuer die Landwirte:

# ? Jan 16, 2023 21:01

12 rats tied together: Sep 7, 2006

Soricidus posted:

the equivalent yaml to <first-name>Benjamin</first-name> is �first-name: Benjamin�. in both cases that is how you represent the value �Benjamin� being assigned to the first-name field.

idk how the hell �!first-name Benjamin� would be used. do you really have a dedicated first-name type that can be used for any value of any field?

no, i was being terse, if i were to reproduce the XSD example i would have !bookstore probably.

Antigravitas posted:

i used to work on a video game that has "what happens when someone uses an ability" expressed in a 200k LoC switch case statement. i dont recall the case count because it was "case skill_id" and the skill_ids had a firewall-rule-like system where ids 0 through 16k were reserved for patch 1, and so on

games are moments away from falling apart basically all the time. even the ones that look like they arent

# ? Jan 16, 2023 21:27

tef: May 30, 2004; -> some l-system crap ->

i have bad news about how the sausage is made lads

# ? Jan 16, 2023 23:27

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

tef posted:

i have bad news about how the sausage is made lads

yeah the casing is pretty gross

# ? Jan 17, 2023 00:07

mystes: May 31, 2006

I only use artificial sausage cases to test my spaghetti code

# ? Jan 17, 2023 00:29

Plank Walker: Aug 11, 2005

mystes posted:

I only use artificial sausage cases to test my spaghetti code

drake no: mockFoobar
drake yes: impossible�Foobar

# ? Jan 17, 2023 05:13

Share Bear: Apr 27, 2004

i thought it was widely well known that many game code bases are horrors because a lot of it gets generated by (usually custom, but sometimes game maker and unity) tooling

# ? Jan 17, 2023 17:38

FlapYoJacks: Feb 12, 2009

Who cares if undertales code is bad. It�s a good game.

# ? Jan 17, 2023 17:42

tef: May 30, 2004; -> some l-system crap ->

Share Bear posted:

i thought it was widely well known that many game code bases are horrors because a lot of it gets generated by (usually custom, but sometimes game maker and unity) tooling

it's because indie/small dev games are developed experimentally, regularly testing out new ideas and chunks. so investing in a big "correct" architecture ends up being a huge time sink, even though random people can look at the final process and go "i could do it neater than that"

i mean, so could the gamedev, if they knew they were going to write that in the first place.

the other thing is that games are rolled up like a big ball of mud. what starts as a two condition switch for events turns into a hundred as things get added to the game, and there really isn't much benefit to cleaning it up. when the game's done, you move on. it's only when people write their own engines that the tradeoff for better code up front makes any sense

well, it would if the people who write game engines ever get around to writing a game

# ? Jan 17, 2023 17:51

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

FlapYoJacks posted:

Who cares if undertales code is bad. It’s a good game.

10x programmer spotted

# ? Jan 17, 2023 17:52

DELETE CASCADE: Oct 25, 2017; i haven't washed my penis since i jerked it to a phtotograph of george w. bush in 2003

tef posted:

it's because applications are developed experimentally, regularly testing out new ideas and chunks. so investing in a big "correct" architecture ends up being a huge time sink, even though random people can look at the final process and go "i could do it neater than that"

i mean, so could the application developer, if they knew they were going to write that in the first place.

the other thing is that applications are rolled up like a big ball of mud. what starts as a two condition switch for events turns into a hundred as things get added to the app, and there really isn't much benefit to cleaning it up. when the app's done, you move on. it's only when people write their own libraries that the tradeoff for better code up front makes any sense

well, it would if the people who write libraries ever get around to writing an app

# ? Jan 17, 2023 17:56

Twerk from Home: Jan 17, 2009; This avatar brought to you by the 'save our dead gay forums' foundation.

tef posted:

well, it would if the people who write game engines ever get around to writing a game

It'd be awesome if Epic made a single player game to show off their awesome engine. Maybe they could even name the game after the engine?

# ? Jan 17, 2023 17:56

Sagacity: May 2, 2003; Hopefully my epitaph will be funnier than my custom title.

I bet it would take them about a fortnite

# ? Jan 17, 2023 18:01

Adbot: ADBOT LOVES YOU

# ? May 2, 2024 06:44

12 rats tied together: Sep 7, 2006

if you work at a place that has their own engine where they've done it "right" the engine will be so much code that you can't possibly check it all out at the same time or understand more than a fifth of it at once. also they wont have done it all right, some of it will be wrong, and other parts of it will be a different, mutually exclusive type of wrong

# ? Jan 17, 2023 18:03

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

«‹›1116 »