type stuff | Factorio | Page 1

clever sapphire May 6, 2023, 3:19 AM

#

i'm threading us

waxen hearth May 6, 2023, 3:19 AM

#

good decision

clever sapphire May 6, 2023, 3:19 AM

#

because i'm sure there will be more

waxen hearth May 6, 2023, 3:19 AM

#

also I forgot extend needs to have data as a first param

clever sapphire May 6, 2023, 3:19 AM

#

oh yeah i think you can just call it self and it'll understand

#

that's the conventional name for a this in lua

#

and what : does anyway

waxen hearth May 6, 2023, 3:20 AM

#

yeah I've used self a few times

#

well it's working, for the things I haven't defined yet

#

I haven't made MiningDrill yet

clever sapphire May 6, 2023, 3:21 AM

#

that's not literal tables

#

i mean like it shoudl be able to understand the giant literatl one-big-table like base files have

#

if all the smarts is there

waxen hearth May 6, 2023, 3:23 AM

#

oh do you mean like if someone does this

data:extend{{
  type = "car",
  -- car properties
},{
  type = "item"
  -- item properties
}}

clever sapphire May 6, 2023, 3:23 AM

#

yeah

#

that's basically the holy-grail and my understanding was we're still waiting on some LuaLS smarts to handle the narrowing once it has seen the type

#

it woudl also apply to various things further down the tree

waxen hearth May 6, 2023, 3:24 AM

#

it seems to be working

clever sapphire May 6, 2023, 3:24 AM

#

nifty

#

then i guess now all we're waiting on is the data set

waxen hearth May 6, 2023, 3:25 AM

#

I'm hoping the prototype docs will be much easier to auto generate this from

clever sapphire May 6, 2023, 3:25 AM

#

i expect so

waxen hearth May 6, 2023, 3:25 AM

#

when therenas is done with it

#

for now, it's manual labor for me :P

#

on second thought

#

that's not a car property

clever sapphire May 6, 2023, 3:32 AM

#

yeah i think LuaLS does a slightly incorrect thing of basically pretending subclasses are union-ish

#

(subclass is the A:B relation, A:B,C is just A:B and A:C, since it's both that would correctly be an intersection of two subclasses, but LuaLS seems to treat it more like a subclass of unions, A:(B|C))

waxen hearth May 6, 2023, 3:35 AM

#

yeah nothing I'm trying seems to have anything to that effect

clever sapphire May 6, 2023, 3:36 AM

#

i wonder if it has a never type yet, and if the absence of that is related to this being not a thing

#

never is an empty set of values, which is useful surprisingly often for representing things which are impossible

#

such as the return value of a function that never returns

waxen hearth May 6, 2023, 3:37 AM

#

I don't think the annotations docs are completely up to date either

clever sapphire May 6, 2023, 3:37 AM

#

(because it throws, for example)

#

or the result of string & number

waxen hearth May 6, 2023, 3:38 AM

#

interesting

clever sapphire May 6, 2023, 3:38 AM

#

which is another way to spell an empty set of values

#

or, more usefullly, your big Prototype type & your literal table so far, when it becoems too broken to match anything

#

(rather than narrowing to a specific subtype)

#

like if you had type="flargle" in it

waxen hearth May 6, 2023, 3:39 AM

#

lol

#

not sure why but flargle is making me laugh out loud

clever sapphire May 6, 2023, 3:40 AM

#

the loader would have a similar reaction i'm sure

#

but yeah, in ts (and many other languages with mature type systems) all this stuff "just works" and it's super easy to get spoiled by that

#

i have occasionally considered porting more of teh debugger's bits to ts2lua also, but not until their sourcemaps are functional :p

#

i think i would get more benefit than most though, i'm more likely to be handling an any that i found laying around somewhere than "regular" mod code would be

waxen hearth May 6, 2023, 3:43 AM

#

yeah makes sense

clever sapphire May 6, 2023, 3:44 AM

#

which is where fancy narrowing really shines

waxen hearth May 6, 2023, 3:45 AM

#

one of the things I'm trying is removing any ---@field type string from prototype definitions that way each one has a specific one like ---@field type "car"

clever sapphire May 6, 2023, 3:46 AM

#

yeah, i do tend to find better results (even in ts) with it unspecified at the higher level, because it makes "any other string" always make an invalid type

waxen hearth May 6, 2023, 3:51 AM

#

I'm still finding non-car related fields even though nothing has a generic string in the type field, hmm

clever sapphire May 6, 2023, 3:51 AM

#

yeah i think LuaLS just still doesn't do that correctly

#

last time i experimented heavily its narrowing was extremely limited

#

like how we have to rewrite .object_name so it looks like it's a global function call so we can pretend it's an extended type

waxen hearth May 6, 2023, 3:53 AM

#

lol that's jank

clever sapphire May 6, 2023, 3:53 AM

#

because only "things like type" get type narrowing in ifs

#

(that's waht __object_name is, and it's not real)

waxen hearth May 6, 2023, 3:54 AM

#

makes sense

#

yeah nothing I tried worked, LLS just isn't there for that yet I guess

#

so is there actually a difference between @alias thing thing.0|thing.1 and @class thing:thing.0,thing.1 besides being able to extend from the latter

clever sapphire May 6, 2023, 4:08 AM

#

the class is an intersection not a union

waxen hearth May 6, 2023, 4:08 AM

#

right right

clever sapphire May 6, 2023, 4:09 AM

#

but i think LLS doesn't have proper intersections which makes it... a mess

waxen hearth May 6, 2023, 4:09 AM

#

what does it do instead

clever sapphire May 6, 2023, 4:09 AM

#

waxen hearth on second thought

this, apparently

waxen hearth May 6, 2023, 4:10 AM

#

considering they were exactly the same I don't think it does intersections at all

clever sapphire May 6, 2023, 4:10 AM

#

intersections are how it narrows the stated type & the value so far

#

or at least, how it's supposed to work

#

the autocomplete that it should offer you is properties taht are not yet present but would fit on objects that are in the interesection of the stated type and the type of the literal value seen so far

waxen hearth May 6, 2023, 4:11 AM

#

but it's doing a union instead

clever sapphire May 6, 2023, 4:11 AM

#

its' doing... something 🤷

#

i'm not sure what it actually does cleanly fits to type algebra

waxen hearth May 6, 2023, 4:12 AM

#

for my own sake, union is comparable to binary OR and intersection is comparable to binary AND

clever sapphire May 6, 2023, 4:12 AM

#

yes

#

it's literally that but on sets

#

you have to think of it as sets of values, the type itself is not actually a thing

#

it's just the circle you draw on the diagram to talk about the values

waxen hearth May 6, 2023, 4:13 AM

#

yeah I'm not thinking about it like a type anymore

#

sometimes it's easier to say that though

clever sapphire May 6, 2023, 4:16 AM

#

bonus: object types are just sugar for intersections of simpler objects with one field each. "all the objects that have this field, and also have that field"

waxen hearth May 6, 2023, 4:18 AM

#

lmao

clever sapphire May 6, 2023, 4:18 AM

#

but basically, everywhere they should have gone "oh that's intersections" LLS did something weirder instead to cover just the one case

#

so they have lots of things that are kinda but not quite intersections if you abuse them just right, but no &

#

and i assume still no never because that's part of the eliminating fields from narrowing

#

(unspecified fields on an object that doesn't accept extra fields are never, and you don't offer them for autocomplete once they narrow to never)

waxen hearth May 6, 2023, 4:25 AM

#

makes sense

#

you think it's worth opening an issue on this?

clever sapphire May 6, 2023, 4:30 AM

#

if you feel like trying to explain it

waxen hearth May 6, 2023, 4:30 AM

#

I'm good at explaining things

#

if alias is just a union though, wouldn't it just not work in the first place?

waxen hearth May 6, 2023, 4:33 AM

#

waxen hearth so is there actually a difference between `@alias thing thing.0|thing.1` and `@c...

like in order to get the behavior we want it would have to use the second definition, right

clever sapphire May 6, 2023, 4:33 AM

#

i think my impression is that they're kind of aware of teh problem but that it's going to take a lot of work to fix it all

waxen hearth May 6, 2023, 4:33 AM

#

I'm not too familiar with the scene, all I know is that "it just works (unless it doesn't)"

waxen hearth May 6, 2023, 4:34 AM

#

waxen hearth like in order to get the behavior we want it would have to use the second defini...

like if it already worked perfectly

clever sapphire May 6, 2023, 4:34 AM

#

iirc sumneko's english isn't great which tends to frustrate me pretty quickly

#

the thign you want is a union of two non-overlapped types

#

so spellign it as subclasses is just wrong anyway

#

you just want to then be able to intersect that union with more things to extend it further

waxen hearth May 6, 2023, 4:35 AM

#

so the alias should work if it was done "correctly"

clever sapphire May 6, 2023, 4:35 AM

#

what you want is &

#

but that requires reworking all the things for peroper intersections, basically

waxen hearth May 6, 2023, 4:36 AM

#

right, because | is union\or and intersection doesn't exist yet

#

where does it do "intersection" again? type resolution and such

clever sapphire May 6, 2023, 4:36 AM

#

intersection-like-things are all over the place

#

the subclass of multiple parents should be an intersection but seems to be a subclass-of-union-ish

waxen hearth May 6, 2023, 4:37 AM

#

navigating the source code is near impossible

clever sapphire May 6, 2023, 4:37 AM

#

the autocomplete shoudl be based on the intersection of value-so-far and stated-type

#

i feel like there's more that i've named earlier but i don't remember now

#

and if you could write an intersection explicitly then you could just do alias & {more:stuff}

waxen hearth May 6, 2023, 4:41 AM

#

would this syntax exist at all or is this just wrong
---@alias baz foo&bar

clever sapphire May 6, 2023, 4:41 AM

#

that just doesnt' exist at all

#

but that's how you would spell an intersection if it did

waxen hearth May 6, 2023, 4:42 AM

#

I mean I know it doesn't exist now, but is this how it should be done if it did

clever sapphire May 6, 2023, 4:42 AM

#

at least i hope so

#

that's how it is basically everywhere else

waxen hearth May 6, 2023, 4:43 AM

#

I should just look at how TS does things for reference

#

that's the golden standard

clever sapphire May 6, 2023, 4:43 AM

#

yes

#

at least for typing on top of a dynamic language it seems to be yeah

waxen hearth May 6, 2023, 4:43 AM

#

my first real understanding of types being... this, is not helping

clever sapphire May 6, 2023, 4:44 AM

#

https://www.typescriptlang.org/docs/handbook/intro.html

Handbook - The TypeScript Handbook

Your first step to learn TypeScript

waxen hearth May 6, 2023, 4:44 AM

#

I mean I knew what types were before, but as far as using them in complicated ways

clever sapphire May 6, 2023, 4:44 AM

#

https://www.typescriptlang.org/docs/handbook/2/types-from-types.html

Documentation - Creating Types from Types

An overview of the ways in which you can create more types from existing types.

#

i guess they consider a lot of this "everyday types" https://www.typescriptlang.org/docs/handbook/2/everyday-types.html

Documentation - Everyday Types

The language primitives.

waxen hearth May 6, 2023, 5:02 AM

#

this is a lot of things

clever sapphire May 6, 2023, 5:27 AM

#

i think the only thing lua does that ts doesn't have a reasonable spelling for already is multiple-expressions, distinct from TS's tuple types which map more directly to tuple-shaped-tables

waxen hearth May 6, 2023, 5:30 AM

#

I see

clever sapphire May 6, 2023, 5:32 AM

#

oh and operators

#

becasue js doesn't do operator overloads

#

but there's an obvious derivation of those from the way they did new i guess

waxen hearth May 6, 2023, 10:02 PM

#

the downside to typing things is that all of my previous code is not chock full of errors for things I haven't typed out yet

#

like type = "custom-input" is invalid because I haven't done that one yet lol

clever sapphire May 7, 2023, 1:24 AM

#

yeah that tends to happen when adding types to codebases that previously didn't have tehm

waxen hearth May 14, 2023, 10:28 PM

#

this is odd

#

scrolling down too far in a multi line union makes the string highlighting go away

#

is that a bug with LuaLS or is that something that can't really be worked around

clever sapphire May 14, 2023, 10:31 PM

#

hard to say without tryign to inspect the actual lsp messages going bakc and forth

waxen hearth May 14, 2023, 10:31 PM

#

reporting it anyway

clever sapphire May 14, 2023, 10:31 PM

#

i would lean towards luals bug though

waxen hearth Jun 2, 2023, 7:44 PM

#

---@meta

---@class IconSpecification:Icon.0,Icon.1,Icon.2

---@class Icon.0
---@field icons IconData.0[]
---@field icon_size SpriteSizeType
---@field icon_mipmaps? uint8

---@class Icon.1
---@field icons IconData.1[]
---@field icon_mipmaps? uint8

---@class Icon.2
---@field icon FileName
---@field icon_size uint8
---@field icon_mipmaps? uint8

---@class IconData.1:IconData.0
---@field icon_size SpriteSizeType

---@class IconData.0
---@field icon FileName
---@field tint? Color
---@field shift? vector
---@field scale double
---@field icon_mipmaps? uint8

#

for bilka

chilly mountain Jun 14, 2023, 6:01 PM

#

waxen hearth ```lua ---@meta ---@class IconSpecification:Icon.0,Icon.1,Icon.2 ---@class Ico...

Hello type people, thinking about how to do the icon specification. In this example, do the .0 things have any special meaning?

clever sapphire Jun 14, 2023, 6:03 PM

#

i just had to generate individual names for them for sumneko so that's the index it was in the union definition

chilly mountain Jun 14, 2023, 6:03 PM

#

I see, so it's just part of the name?

clever sapphire Jun 14, 2023, 6:04 PM

#

yeah type names in sumneko can include a dot and i occasionally abuse that to make related things look related

chilly mountain Jun 14, 2023, 6:05 PM

#

Sweet

clever sapphire Jun 14, 2023, 6:05 PM

#

and i only do that if the union member didn't have its own name already some ordinary way

chilly mountain Jun 14, 2023, 6:05 PM

#

I feel like I'm missing some context though, how did that definition come to exist in the first place?

clever sapphire Jun 14, 2023, 6:06 PM

#

err

chilly mountain Jun 14, 2023, 6:06 PM

#

(Bilka just pointed me here and said it'd be interesting)

clever sapphire Jun 14, 2023, 6:06 PM

#

well i guess codegreen probably just wrote that one by hand?

#

but that style of thing is what i do with the unions of anonymous types in runtime json

chilly mountain Jun 14, 2023, 6:06 PM

#

Ahh alright

#

Sounded like you were already playing with prototype stuff

clever sapphire Jun 14, 2023, 6:07 PM

#

no, i have't touched it yet

#

the adhd project list is an adventure lately

chilly mountain Jun 14, 2023, 6:07 PM

#

ha no worries, the format isn't 100% set in stone yet anyways (well, one property in particular)

#

I'm still not sure how I'll translate IconSpecification into the new system. Mostly because I don't want to repeat all its fields everywhere it is used.

waxen hearth Jun 14, 2023, 6:09 PM

#

I've been writing out a lot of the typedefs manually whenever I use something that's missing

clever sapphire Jun 14, 2023, 6:09 PM

#

i forget, does the json have some means of listing a class/table type's parent type?

waxen hearth Jun 14, 2023, 6:09 PM

#

sumneko isn't smart enough to do type narrowing though so some of the complicated stuff like the iconspecification is functionally useless

clever sapphire Jun 14, 2023, 6:09 PM

#

if so, you can break the common stuff out into a subtype that all the variations inherit

#

(multiple parents is fine too)

chilly mountain Jun 14, 2023, 6:11 PM

#

That's what I was just thinking as well. Slight issue with it is that prototypes can only have one parent currently, so the IconSpec would need to be 'inserted' in the right spot, if it'd even be possible with a single definition of it.

waxen hearth Jun 14, 2023, 6:11 PM

#

that's a toughie

clever sapphire Jun 14, 2023, 6:11 PM

#

types are hard

chilly mountain Jun 14, 2023, 6:11 PM

#

Maybe proper multiple parent support would be warranted, but I don't know if there's any other uses for it, so seems a bit heavy-handed

waxen hearth Jun 14, 2023, 6:12 PM

#

there's a few other things it's used for

clever sapphire Jun 14, 2023, 6:12 PM

#

i feel like you'll end up using it more once you ahve it

waxen hearth Jun 14, 2023, 6:12 PM

#

not off the top of my head but I do remember typing out some stuff like that

chilly mountain Jun 14, 2023, 6:12 PM

#

probably

clever sapphire Jun 14, 2023, 6:12 PM

#

yeah i can't think specifically where but i'm pretty sure there's more

chilly mountain Jun 14, 2023, 6:12 PM

#

Wouldn't be too hard to implement, thinking about it.

waxen hearth Jun 14, 2023, 6:12 PM

#

sprite/animation layers

chilly mountain Jun 14, 2023, 6:12 PM

#

Just a bit tedious

clever sapphire Jun 14, 2023, 6:13 PM

#

(and sumneko already understands multiple inheritance, so it's easy on this end)

chilly mountain Jun 14, 2023, 6:13 PM

#

Ah that's good to know.

#

@tacit pine Get in here! (during work hours)

clever sapphire Jun 14, 2023, 6:13 PM

#

waxen hearth ```lua ---@meta ---@class IconSpecification:Icon.0,Icon.1,Icon.2 ---@class Ico...

wait, shoudln't that top level type be a union not a class with multiple parents?

waxen hearth Jun 14, 2023, 6:14 PM

#

can't extend off of unions

clever sapphire Jun 14, 2023, 6:14 PM

#

hrm

waxen hearth Jun 14, 2023, 6:14 PM

#

I remember having to extend off of it for something

#

dark_background icon

#

something like that

clever sapphire Jun 14, 2023, 6:15 PM

#

that's annoying because the class with multiple parents is technically an intersection and maybe one day sumneko will do those properly and it'll break that

waxen hearth Jun 14, 2023, 6:15 PM

#

hopefully it does

#

lack of type narrowing is awful, please make union behave like union and not intersection, sumneko

chilly mountain Jun 14, 2023, 6:16 PM

#

My hope is Rai's language server project pulls through at some point, don't have much hope for LuaLS working properly ever sadly.

kindred solar Jun 14, 2023, 6:17 PM

#

TIL this thread exists

waxen hearth Jun 14, 2023, 6:18 PM

#

LuaLS is great for the things it does work with, but it really falls flat in some of the fundamentals

#

welcome rai

kindred solar Jun 14, 2023, 6:19 PM

#

I mentioned this yesterday, but I think I'm going to prioritize the language server over my pesticide project because while I can use the debugger through vscode just fine, LuaLS is horrible in both places and is actively annoying me every day.

chilly mountain Jun 14, 2023, 6:19 PM

#

He has been summoned! Guess you have a keyword notifier thing?

kindred solar Jun 14, 2023, 6:19 PM

#

Yeah I have highlights for my name 😄

waxen hearth Jun 14, 2023, 6:19 PM

#

do you have one for rai specifically lmao

kindred solar Jun 14, 2023, 6:20 PM

#

Yes

chilly mountain Jun 14, 2023, 6:20 PM

#

So nosy

#

:p

waxen hearth Jun 14, 2023, 6:21 PM

#

example of type narrowing not existing

chilly mountain Jun 14, 2023, 6:23 PM

#

While we're on it, I have something else I wanted to ask about. I have a global variable called util, and LuaLS insists on combining it with the core lualib stuff. Core util is not required anywhere. I know the easy solution is to use a different name for my variable, but I really don't want to. Is this a LuaLS issue, or is there something you could do about it jarg?

kindred solar Jun 14, 2023, 6:23 PM

#

It's because fmtk has the core util in the library

#

What we really need is a way to have different language server sessions for data vs. control

waxen hearth Jun 14, 2023, 6:23 PM

#

you might be able to define a custom class and make it specifically that instead

chilly mountain Jun 14, 2023, 6:24 PM

#

Hm that could work. I was mainly bringing it up in case it was an oversight on fmtk's side, or at least fixable by it in some way.

waxen hearth Jun 14, 2023, 6:24 PM

#

could name it utils instead

chilly mountain Jun 14, 2023, 6:24 PM

#

that could also work, but eh I'm picky

kindred solar Jun 14, 2023, 6:25 PM

#

It's not an oversight, it's intentional. Unfortunately fmtk can't tell the server to only require it some of the time, so it requires it all the time.

waxen hearth Jun 14, 2023, 6:25 PM

#

you could also remove util from fmtk's auto require

chilly mountain Jun 14, 2023, 6:25 PM

#

So it can't be tied to having require("util") somewhere in the project?

kindred solar Jun 14, 2023, 6:25 PM

#

One thing I intend to do for my server is have a way to specify "only include these globals if these files are required" and specify specific root files to start from (data.lua and control.lua)

clever sapphire Jun 14, 2023, 6:25 PM

#

it probably is tied to seeing that soemwhere

waxen hearth Jun 14, 2023, 6:26 PM

#

util exists in data stage as a global though

clever sapphire Jun 14, 2023, 6:26 PM

#

but it may have also seen it through data files somewhere

chilly mountain Jun 14, 2023, 6:26 PM

#

It's not in my project, but maybe vanilla data files then?

clever sapphire Jun 14, 2023, 6:26 PM

#

yeah

chilly mountain Jun 14, 2023, 6:27 PM

#

Unfortunate, but I can live with it. Just wanted to make sure there's not something silly that would fix it.

waxen hearth Jun 14, 2023, 6:27 PM

#

base makes util a global in data stage, LuaLS can't separate files out, and fmtk is telling LuaLS that util is a global

chilly mountain Jun 14, 2023, 6:27 PM

#

Makes sense

waxen hearth Jun 14, 2023, 6:27 PM

#

all combined together to give you this issue

chilly mountain Jun 14, 2023, 6:27 PM

#

Guess the real crime is util.lua creating a global variable.

#

Would be fun to break all the mods by undoing that.

kindred solar Jun 14, 2023, 6:28 PM

#

Globals are evil. One particularly annoying one is that one of the core lualib files defines a player global 😡

waxen hearth Jun 14, 2023, 6:28 PM

#

a "workspace" is meant to be one set of files that all interact, so we're the ones doing weird stuff with multiple stages of things

#

globals are like the only things that break when doing this

chilly mountain Jun 14, 2023, 6:29 PM

#

Feels like it should have the concept of a single file being the entry point, and only follow requires from there.

kindred solar Jun 14, 2023, 6:29 PM

#

Yeah, that's exactly what I intend to do in mine

clever sapphire Jun 14, 2023, 6:30 PM

#

that will cause some trouble with globals defined in the parent file and used in the children

chilly mountain Jun 14, 2023, 6:30 PM

#

How so?

clever sapphire Jun 14, 2023, 6:30 PM

#

because the child file would have no context to see them

#

so it woudl think they dont' exist

#

ls for it would

kindred solar Jun 14, 2023, 6:31 PM

#

I haven't written the server yet, we don't know what it will or won't do 🙃

clever sapphire Jun 14, 2023, 6:31 PM

#

well, following from what you just said

#

if each file only sees what it itself requiers, then it has no context for what its require-parent has set up for it

waxen hearth Jun 14, 2023, 6:31 PM

#

chilly mountain Feels like it should have the concept of a single file being the entry point, an...

only doing this means globals defined before requires won't be seen in the required files

chilly mountain Jun 14, 2023, 6:32 PM

#

It could collect all globals along the way and inject them afterwards (no idea if that's work)

waxen hearth Jun 14, 2023, 6:32 PM

#

the solution is to not use globals

kindred solar Jun 14, 2023, 6:32 PM

#

clever sapphire if each file only sees what it itself requiers, then it has no context for what ...

Since globals are, well, global, I intend to put them in a separate place that all files will reference for diagnostics.

clever sapphire Jun 14, 2023, 6:32 PM

#

chilly mountain It could collect all globals along the way and inject them afterwards (no idea i...

thats' basically waht sumneko is doing now

clever sapphire Jun 14, 2023, 6:32 PM

#

kindred solar Since globals are, well, _global_, I intend to put them in a separate place that...

so now util is global to all files again

chilly mountain Jun 14, 2023, 6:32 PM

#

Except it shouldn't even parse the data stage files from my control.lua file starting point right?

kindred solar Jun 14, 2023, 6:33 PM

#

All files in that workspace, yes. The whole problem with LuaLS is that data stage and control stage are not separated, so it shows globals that aren't there.

waxen hearth Jun 14, 2023, 6:33 PM

#

being able to separate files into different scopes would be great

chilly mountain Jun 14, 2023, 7:03 PM

#

So getting back to IconSpecification, talked to Bilka about it, and the inheritance solution won't work for a multitude of reasons. One 'proper' solution would be to have a system where you can 'extend' a prototype with a Type, so basically you could say 'This prototype has all the properties from this type'. But that would add complication to the whole stack, just for this particular situation.
We're kinda leaning towards just duplicating the underlying docs text everywhere and being done with it. The real issue is that the IconSpec implementation is very silly, and so the docs will also be silly. If you guys can think of another solution though, we'd be happy to consider it.

kindred solar Jun 14, 2023, 7:11 PM

#

Solution: make it less silly for 1.2

tacit pine Jun 14, 2023, 7:12 PM

#

congrationlations, you have been selected as the winner for some free work

kindred solar Jun 14, 2023, 7:12 PM

#

No not like that runs

chilly mountain Jun 14, 2023, 7:17 PM

#

Oh I will have lots of 1.2 stuff for you at some point :p

#

But also this would break literally every mod

#

Also the docs will be for 1.1 as well so it wouldn't even help me

waxen hearth Jun 14, 2023, 7:33 PM

#

it would be nice if icon_specification was a table field in the prototype instead of the prototype extending off of it

#

but that would break literally every existing mod

chilly mountain Jun 14, 2023, 7:34 PM

#

yeah exactly

tacit pine Jun 14, 2023, 7:35 PM

#

I think the goal should be to not make the layering and scaling not suck too

#

but like, not an easy task, hence my (badly typoed aka why I shouldnt be working right now) joke

waxen hearth Jun 21, 2023, 7:52 PM

#

what happens if I have a class that extends from another class but has a field of the same name

#

does it override the base class?

clever sapphire Jun 21, 2023, 7:54 PM

#

i'm not suer what sumneko actually does but what it should do is the child definition wins (and must be a subset of the parent's definition)

waxen hearth Jun 21, 2023, 7:54 PM

#

I'm wondering if I can have a field set to false in one, then the other sets it to true and has additional properties

clever sapphire Jun 21, 2023, 7:55 PM

#

well, no, because now the child is not a valid member of teh parent

#

you can only narrow a field by redefining, not expand

waxen hearth Jun 21, 2023, 7:56 PM

#

I'm going to try it anyway and see if sumneko explodes

#

it works, surprisingly

#

#

and when it's generic it gives you both options

#

still no type narrowing in general, but it's better than nothing

#

when you make a field optional with ? is that shorthand for |nil on the end of it

clever sapphire Jun 21, 2023, 8:04 PM

#

yes

#

for fields there's no distinction between ? and |nil

#

but for arg/return lists there can be

#

since those do have a "not present"/"end of list" that is distinct from nil, especially when interacting with cfunctions

waxen hearth Jun 21, 2023, 8:08 PM

#

interesting

clever sapphire Jun 21, 2023, 8:08 PM

#

for example, script.on_event will throw if the third arg is nil-but-present when the first one isn't a single event

waxen hearth Jun 21, 2023, 8:09 PM

#

ah I see

#

so the ? would be "or not present" in that case

clever sapphire Jun 21, 2023, 8:10 PM

#

but for fields there's no such distinct "not-present" case

waxen hearth Jun 21, 2023, 8:10 PM

#

in script.on_event's definition

clever sapphire Jun 21, 2023, 8:10 PM

#

yeah the ? is the not-present case when it is meaningfully different

waxen hearth Jun 21, 2023, 8:10 PM

#

neat

clever sapphire Jun 21, 2023, 8:10 PM

#

which means youc an also do fun things like string|nil? in an arg list

waxen hearth Jun 21, 2023, 8:11 PM

#

lol what the heck

#

either nil or not present makes sense though

clever sapphire Jun 21, 2023, 8:11 PM

#

you shouldn't but because they mean different things, you can

waxen hearth Jun 21, 2023, 8:11 PM

#

"just because you can..."

waxen hearth Jun 21, 2023, 8:13 PM

#

waxen hearth

as it turns out this whole thing is useless in my specific scenario because all of the potential additional fields could be nil

#

I wish there was an easy way to define either/or properties

#

for all of the "this field must be present if x" in the prototype definitions

waxen hearth Jul 5, 2023, 5:21 PM

#

can you put a question mark on a @type

clever sapphire Jul 5, 2023, 5:22 PM

#

what do you mean?

waxen hearth Jul 5, 2023, 5:24 PM

#

this is inside of a @class literal with the actual table definition, this can be either uint or nil

---@type uint?
train_id = nil

clever sapphire Jul 5, 2023, 5:25 PM

#

i think so?

#

wouldn't it be a @field inside a class though?

waxen hearth Jul 5, 2023, 5:26 PM

#

full context

---@class PlayerData
global.players[index] = {
    ---@type LuaPlayer
    player = player,
    ---@type table<string, any>
    flags = {},
    ---@type LuaEntity[]
    rename = {},
    ---@type uint|nil
    default_train_limit = 1,
    ---@type boolean
    restore_automatic = true,
    ---@type uint?
    train_id = nil
}

#

seems to work though

clever sapphire Jul 5, 2023, 5:27 PM

#

i'd have expected those to be tagged with field but yeah that should be fine too 🤷

waxen hearth Jul 5, 2023, 5:27 PM

#

field expects a name

clever sapphire Jul 5, 2023, 5:27 PM

#

ah right

chilly mountain Jul 5, 2023, 5:36 PM

#

As I understand ? and |nil is the same

clever sapphire Jul 5, 2023, 5:44 PM

#

Except in arg/return lists, yeah

chilly mountain Jul 5, 2023, 5:56 PM

#

Oh what‘s the difference there?

clever sapphire Jul 5, 2023, 5:57 PM

#

? is for "can be not-present" vs "can be nil"

chilly mountain Jul 5, 2023, 5:57 PM

#

Which is functionally the same in Lua right? Guess it has semantic difference for the language server

clever sapphire Jul 5, 2023, 5:57 PM

#

it is distinct in lua also depending on how you unpack them

#

in particular when interacting with vararg functions, like all cfunctions

#

foo(nil) and foo() produce different contents of ... in foo

chilly mountain Jul 5, 2023, 5:58 PM

#

Ah I see

clever sapphire Jul 5, 2023, 5:59 PM

#

you have to try a bit harder to contrive a situation where it's obviously distinct in returns but they're teh same mechanism, so the same thing applies

#

but in "normal" lua with all named args and assignments stright to named vars the magic list lenght adjustments hide it all

#

and of course "not-present" can only be at/after the end of the list of present values

#

i'ts only distinguishable by explicitly checking the count though

chilly mountain Jul 5, 2023, 6:06 PM

#

Appreciate the explanations

clever sapphire Jul 5, 2023, 6:08 PM

#

the .../args/returns-tuple type is definitely the neglected one in lua, which makes it all a bit weirder to see since all interactions with it are at least a little clunky

waxen hearth Jul 26, 2023, 12:23 AM

#

so sumneko updated or something and now it says I have missing fields on every single prototype

clever sapphire Jul 26, 2023, 12:23 AM

#

uh oh

waxen hearth Jul 26, 2023, 12:24 AM

#

from the changelog:
NEW diagnostic: missing-fields

waxen hearth Jul 26, 2023, 12:42 AM

#

we might finally get type narrowing

#

https://github.com/LuaLS/lua-language-server/issues/2214

clever sapphire Jul 26, 2023, 12:48 AM

#

oh hell yeah

#

i think i'm gonna try and work on taht and the PR for the plugin and some other bits tomorrow then

#

i've been dragging on working on all the things for a while, but i'm feeling functional again maybe...

waxen hearth Jul 26, 2023, 1:22 AM

#

very nice

#

also, with this new diagnostic:
stuff like LuaGuiElement.add_param is going to have to change

#

currently it's just type: string and other fields

waxen hearth Jul 26, 2023, 6:30 AM

#

thanks

Cannot assign { player: LuaPlayer, [string]: LuaGuiElement } to parameter { player: LuaPlayer, [string]: LuaGuiElement }.

tacit pine Jul 26, 2023, 10:51 AM

#

the type narrowing was released a few hours ago

#

with the prototype doc migration progressing quite well it looks like we'll have good language server support for data stage soon

#

and I'm waiting on this release https://community.kde.org/Schedules/KDE_Gear_23.08_Schedule for trying it with a different text editor and that's only a month away

chilly mountain Jul 26, 2023, 12:36 PM

#

waxen hearth also, with this new diagnostic: stuff like `LuaGuiElement.add_param` is going to...

I‘ll change type to a union of literals. If there‘s more of these you run into, please post them into the forum thread.

waxen hearth Jul 26, 2023, 1:53 PM

#

@chilly mountain well, I meant that it's going to have to be a union of classes, each with their own definition for each type

chilly mountain Jul 26, 2023, 1:53 PM

#

Sure, that's what the LS can do, but the type param should still be a union to enable that :) It already splits out the properties depending on type, so the LS should be able to handle it somehow right?

waxen hearth Jul 26, 2023, 1:55 PM

#

from what I can see, no it can't

#

currently LuaGuiElement.add_param is one class with all of the fields of the different types of elements

#

so it says stuff like column_count is required, even though type is not a table element

chilly mountain Jul 26, 2023, 1:56 PM

#

Right, it needs to narrow it somehow. Which is the same kinda thing needed for prototypes, if I understand correctly.

waxen hearth Jul 26, 2023, 1:58 PM

#

here's an example of what I did for the energy source type for prototypes, there's a base class for the shared parameters, and then a bunch of individual classes for the unique parameters, and then a union all the way at the top for the actual type (there's more below for heat/void/fluid, didn't want to image wall)

chilly mountain Jul 26, 2023, 1:59 PM

#

And does the language server understand that as it should?

#

(I use that exact pattern a lot in the prototype docs, btw)

waxen hearth Jul 26, 2023, 2:00 PM

#

it should, if type narrowing exists now then absolutely

chilly mountain Jul 26, 2023, 2:01 PM

#

Well that's good then. I guess what I've been trying to say is that the runtime docs JSON for LuaGuiElement will not split it out into classes like that, but if type is the proper union of literals, FMTK will be able to do that part itself I assume.

waxen hearth Jul 26, 2023, 2:04 PM

#

yeah I guess with variant parameter groups

chilly mountain Jul 26, 2023, 2:04 PM

#

Yep

#

I don't love those as a solution, but they won't change in the short term.

waxen hearth Jul 26, 2023, 2:05 PM

#

if that's the case, then I think that's something that could be done now, type string union or not

chilly mountain Jul 26, 2023, 2:05 PM

#

Yes, it's just that it wouldn't know about values of type that don't have additional parameters.

#

Since they don't figure in that list.

waxen hearth Jul 26, 2023, 2:06 PM

#

ahhh I see

chilly mountain Jul 26, 2023, 2:06 PM

#

Hence why I'm looking at turning all those type parameters into unions of literals now

#

Which they just should be anyways.

waxen hearth Jul 26, 2023, 2:08 PM

#

for clarity's sake, this is the original issue for anyone that comes across this

waxen hearth Jul 26, 2023, 4:57 PM

#

@clever sapphire are you auto generating the serpent.options typedef? none of the fields are marked as optional

waxen hearth Jul 26, 2023, 6:25 PM

#

also, I seem to have two places for runtime-api files, one in the .vscode workspace folder and one in appdata, should I delete one of those?

clever sapphire Jul 26, 2023, 6:46 PM

#

yeah the one in workspace is old, shouldn't exist anymore

#

and serpent is just a manual lib file,

waxen hearth Jul 26, 2023, 7:02 PM

#

cool, thanks

#

serpent.options should be all optional fields and not required, not sure if that's something you can update for everyone else

clever sapphire Jul 26, 2023, 7:03 PM

#

yeah when it cools down a bit and my AC catches up again...

clever sapphire Jul 26, 2023, 7:49 PM

#

waxen hearth <@288714200296194048> are you auto generating the serpent.options typedef? none ...

https://github.com/justarandomgeek/vscode-factoriomod-debug/commit/d782da5758f9d4bf87349b5cbec7d70560bed790

GitHub

serpent options optional · justarandomgeek/vscode-factoriomod-debug...

#

wtf why only this one?!?

waxen hearth Jul 26, 2023, 7:59 PM

#

lmao

clever sapphire Jul 26, 2023, 8:06 PM

#

sure seems like i need to reverse the union for all the position types for that though lol

waxen hearth Jul 26, 2023, 8:07 PM

#

weird

clever sapphire Jul 26, 2023, 8:25 PM

#

were there mroe things for me to look at immediately or is it just functions that need their params split up in the json first?

waxen hearth Jul 26, 2023, 9:18 PM

#

I don't think so

#

I'm in the process of writing a script to detect every instance of variant parameters without a union of string literals for one of the regular parameters

#

but that's for therenas, not you :P

chilly mountain Jul 26, 2023, 9:38 PM

#

It's not that many actually

#

Mostly filter concepts, which are very crappy in general

#

And create_entity, which is a whole mess anyways

waxen hearth Jul 26, 2023, 10:03 PM

#

yeah that's not even based on a type

#

I had it working a tiny bit but I go's type system is kind of ass sometimes

clever sapphire Aug 1, 2023, 10:53 PM

#

@chilly mountain can we make it officially stated that union types used as return types always actually mean the first member type? i'm pretty sure that would let me adjust the generated bits to have it be teh strict type on returns and the loose type everywhere else. ~~pretty sure it's true for all the current ones.~~

#

oh wait, hrm. maybe that's not true....

#

i have a note from somebody to try to do taht on all the position types but then i'm now thinkign about other places where the return type depends on what you ask for instead... 🤔

chilly mountain Aug 1, 2023, 11:04 PM

#

I‘d need some examples 😅 I‘m not aware of any such rule. For positions (and/or Vector-like things?) specifically, it should always return the table, not tuple, format when reading, but accept either. But even that is not quite always true if I recall correctly. It‘s on the todo list to clean up the mess for 1.2.

#

The mess kinda comes from read-write attributes having one type, so it needs to be the more permissive one. The proper solution is to allow for separate read and write types, which would solve your issue I think.

clever sapphire Aug 1, 2023, 11:05 PM

#

basically all the named or numbered tables, beucase it only pushes the named versions

#

because the LuaHelpers to push them are only for that version

chilly mountain Aug 1, 2023, 11:06 PM

#

Yes, but I‘m not sure if that‘s true 100% of the time.

clever sapphire Aug 1, 2023, 11:06 PM

#

it would be unless someone did the extra work to make the other kind

chilly mountain Aug 1, 2023, 11:06 PM

#

If they didn‘t use the helpers, yes

#

I have some notes on this somewhere, but it‘s late, I can look into it tomorrow.

chilly mountain Aug 2, 2023, 2:46 PM

#

Hm I'm not sure, my notes are unfortunately confusing. But yeah there's no formal rule that it always uses the first union member. This will be addressed in the long term by separate read and write types.

clever sapphire Aug 2, 2023, 2:47 PM

#

well, no, because that still won't address functions which return position, for example

#

... which i'm pretty sure exist, but now i'm having to look to try to fnd a specific example

#

ah here's one https://lua-api.factorio.com/latest/LuaForce.html#LuaForce.get_spawn_position

#

but also other union concepts, like SimpleItemStack from https://lua-api.factorio.com/latest/LuaEquipmentGrid.html#LuaEquipmentGrid.take though that one is even in the "wrong" order since returns woudl always be the full ItemStackDefinition afaik (though, now that that has a name, maybe it shoudl just be used as the return type directly?)

chilly mountain Aug 2, 2023, 2:52 PM

#

Those are just kinda lazy documentation writing, they should be the exact type that is returned.

#

There is no ambiguity in the code for method returns, so there shouldn't be any in the docs.

clever sapphire Aug 2, 2023, 2:52 PM

#

yeah i was just thinking maybe we should just not use union types in returns at all when it's like that...

#

only when it actually coudl be many things (though, in many cases overload signatures woudl be better for that instead)

chilly mountain Aug 2, 2023, 2:53 PM

#

I can write down a note to go over these and use the more specific type.

#

Yeah overload is a different thing, which I also want to add at some point.

clever sapphire Aug 2, 2023, 2:54 PM

#

yeah i forget off hand but there's a couple places where the signatures we wrote in plain types are kinda weird because it was a hack around not having overloads

#

i think with this narrowing stuff there's a good chance sumenko will get good overload resolution soon too (since it's a very similar analysis)

chilly mountain Aug 2, 2023, 2:56 PM

#

Ah yeah that'd be nice. It's on the list to add overloads for the next version of the runtime docs.

clever sapphire Aug 2, 2023, 2:57 PM

#

because technically overloads are just unions of function signatures, so ti just has to narrow based on the args filled in

waxen hearth Aug 10, 2023, 8:12 AM

#

thank you language server

#type stuff