A WIP wrapper for the GitHub REST API.
#GitHub.gml
native fable
unkempt rover
native fable
I'll be adding more stuff to this as time goes on. For now, it can retrieve either a list of releases from a repo or the latest release
native fable
I've added functions for all release and release asset endpoints.
native fable
I've added a tonne of functions for various issues endpoints, and I have done some maintainence work behind the scenes. Once I have added some other endpoints I will make a pre-release version with documentation
If anyone wants some specific endpoint categories fast-tracked, please let me know
near quiver
If anyone wants some specific endpoint categories fast-tracked, please let me kn...
Any chance your request structure could take in a callback and error (setter functions after request is made) which would automatically be called based on the response? And it would just pass in the body. I'm doing something similar with promise library, however that library only makes the error back optional. Where as here I see there being a benefit to either approach, one is simpler to write, the other is more versatile
Thinking something like
var r = GitHub.issue_method()
.set_callback(func(){})
.set_errorback(func(){});
And obviously if it's undefined you just simply wouldn't call it
native fable
Yeah good idea
Been looking for a better way to handle receiving the data
Noted
Also has room for status handling too
.status404(function () {})
Gets called upon a 404 status code
near quiver
I would consider that an error back, but it has its specific uses
Maybe if 404 isn't set then default to error back? And if neither are set obviously just don't call anything
Personally anything which isn't returning what you're expecting I always consider an error, and pass in all information so the user can decide if they want to try another request or if they would rather fail safely out
native fable
That’s a fair point
near quiver
bahhh
also partially unrelated but someone really needs to make an extention for gml which allows for macro overwrites safely.
in concept
native fable
I dislike the idea of using a macro for an API key here
Just because of safety
It’s fine for testing but you ideally want to be getting your key from your own web server and using expiring keys and regenerating them every once in a while
I have it in a file to test and so that I don’t accidentally commit my auth tokens
I’ll also mention this in the docs but I recommend you only use fine-grained tokens
native fable
request.setCallback(function (_requestBody, _requestObject) {
if (_requestObject.httpStatus != 204)
{
show_debug_message(_requestBody);
}
else
{
show_debug_message("No data");
}
});
request.setErrorback(function (_requestBody, _requestObject) {
if (_requestObject.httpStatus == 404)
{
show_debug_message("Object not found.");
}
else
{
show_debug_message("Some other error.");
}
});
this is the usage
it doesnt expect a return value
native fable
@near quiver finished all the issues endpoints except for issue dependencies (no idea how its used or what its for)
gonna start working on gist now
near quiver
<@408240530816892928> finished all the issues endpoints except for issue depende...
That sounds like issues which are blocked by other issues. Specifically for use in feature request. But I might be wrong
native fable
working on gists
native fable
That sounds like issues which are blocked by other issues. Specifically for use ...
that sounds about right
native fable
oh thats the wrong embed
@near quiver first beta release
please let me know if you encounter any issues
near quiver
Hell yeah!
native fable
if you spot anything wrong with the docs let me know, i wrote a script to generate them for me based on the jsdoc comments
native fable
github has oauth, im gonna see if i can get oauth working with this library
native fable
ive got the basic steps working, need to figure out a simple system to actually request oauth from the user
native fable
i got it "working"
like it works but its bad
and this is the code that makes it work
i need to think of a better implementation
proven fjord
Oauth is so annoying 🤣
native fable
yeah
i think ive got a good system in place
// Create OAuth
oauth = new GitHubOAuth(_clientID);
// Create auth request
oauthRequest = oauth.requestAuthentication();
// Set the request callback
oauthRequest.setCallback(function (_resultBody, _request)
{
// Check that the request has made it through
if (_request.httpStatus == 200)
{
// Show debug message to login
show_debug_message($"Please visit: \"{_resultBody.verification_uri}\" and use the code: \"{_resultBody.user_code}\". This code will expire in {_resultBody.expires_in} seconds.");
// Set our authentication callback (runs when authentication is successful)
oauth.setAuthenticationCallback(function(_resultBody, _request) {
show_message(_resultBody);
});
// Now we poll the authentication
oauth.pollAuthentication(_resultBody.device_code, _resultBody.interval + 1, 20);
}
});
GitHubOAuth is doing most of the work here
there is minimal error checking so I will need to add a bunch in
but, if you were using this library would you be happy with this functionality?
near quiver
Alub you sound like a teenage girl who just got her first boyfriend, so eager to make people happy. It's fine my guy!
native fable
i like active feedback
also, feel free to test out the authentication
its on the oauth-testing branch
proven fjord
writing docs is a pain in the ass
Sure is
deep tide
here's how to do web page-based auth: https://github.com/JujuAdams/GitHub.gml/commit/46cfca52f4ef1cd4bed9bd432be42b47cc530d3f
not 100% robust, i didn't handle all the error states, but the basic idea works
native fable
Insane work, I did not know that the web flow would work in GM
deep tide
the first step is to imagine it is possible and then work backwards from there
native fable
Round of applause to you
deep tide
it needs a lot of tidying up for it to be acceptable as production code but ... yknow ... i can't be bothered lol
do you want a PR for this?
native fable
Yeah you can do! Thanks!
And tbh, I dunno if I even consider my code production acceptable
deep tide
it's pretty good!
only thing i saw that exceeded my raised eyebrow threshold was putting the config file under (System)
you probably wanna make that more obviously editable
everything else is nice and clear and does what it should do, as far as i can tell
native fable
it's pretty good!
Thank you
native fable
only thing i saw that exceeded my raised eyebrow threshold was putting the confi...
Yeah, done that as part of the refactor work I did last night, I will move that when I next work on it
There’s a couple more endpoint categories I want to implement (like another probably 60 actual endpoints, most important being users), a bunch of full scale testing and perhaps some example code, as well as studying the web flow code
Once all that’s done I’ll probably consider a v1.0 release
deep tide
hypothetically, could this tool be used to automatically list releases and then pull .yymps from GitHub?
because this + the ImGui work + an automatic import tool = community package manager
native fable
Yeah it can do that right now probably
I’d need to fully test the whole process
But theoretically, yes
It would probably need some sort of index to grab repos from to automatically show a bunch
But searching via owner and repo is ez
deep tide
oh that's a solveable problem
you create an index repo
i'd imagine this tool can also read PRs
native fable
I haven’t implemented PR endpoints just yet, I can add that to the fast-track though
deep tide
please and thank you
so then people can PR URLs into a master list. package manager can then have a "safe" and "unsafe" modes where the "safe" mode only looks at the main branch on the repo and the "unsafe" mode also scans the PRs
filthy, filthy stuff
native fable
We love JSON in this household
Yeah I have no idea how .yymps’ actually work
That would be an interesting problem to solve
deep tide
they're .zip files 🙂
deep tide
having a little poke around now ... looks like a normal project file with bits missing
hopefully a merge is as easy as copy-pasting assets
native fable
I wonder if stuff like the resource order file and other junk auto updates
native fable
Never really taken a look at how project files work, so I don’t know how assets are indexed
All I know is that they use JSON and in gms1.x they used XML(?????)
deep tide
yeah GMS1 was bad
GMS2.0 was also bad in a different way
GMS2.3 is much better
i'll let you enjoy the experience of poking around, shan't spoil the fun
native fable
I appreciate it thank you, I will peruse once I’m home
native fable
back home, working on making some changes to your PR before i merge
it works really well
deep tide
i was a little surprised it doesn't need port forwarding
oooh haven't tried on Android actually
native fable
i was a little surprised it doesn't need port forwarding
yeah that is actually quite surprising
native fable
oh no
deep tide
there'll be ways around that but i have no idea what they'd be
lemme test on Mac real quick
yep we're good
Linux is probably fine too
deep tide
ah it'll be because the app gets backgrounded
native fable
i need to think about how to check for a device-flow timeout
deep tide
i'll leave that to you. a timeout will be helpful for both flows
native fable
yeah thats for sure
i wonder if web-flow has some sort of timeout time like device-flow does?
deep tide
nope
if you think about it it's just a button on a website
that button redirects you to another website
(second website being localhost for us)
the access token will eventually time out though
native fable
thats true, ill look into the lifespan of those tokens
though i feel like they probably last a couple days at the minimum
added programmer definable HTML for web-flow auth
potentially risky business
deep tide
you might find you need to handle subsequent GETs from the browser
at the moment we're serving one "file" but if you want to do something prettier then you'll need to let the browser request other assets
you might also want to change the logic away from using the localhost root to something like http://localhost:52499/please-return-to-game/
when the HTTP GET comes in from the browser you'll be able to pick up /please-return-to-game/
then subsequent GETs from the browser can be redirected more easily
(also you'll want to keep the browser socket open for a while instead of closing it immediately)
kinda fun building a basic HTTP server in GM
native fable
so many things to consider
ill throw all this into an issue and i can deal with them later
near quiver
finally got some time to put this to work, should i use the current release or would you prefer i test the oauth @native fable?
native fable
the latest release has device-flow authentication
you can test with the current main branch commit which has jujus changes and some other stuff
all that new stuff is undocumented so youll need to figure it out or ask away here
native fable
im currently working on actual tests that arent just hacked together
near quiver
small issue ive already ran into, setters like this are common to be builder structures
thought about making a pr for this but its such a small thing
native fable
im not entirely sure, but the github docs do differentiate the two
you can leave either undefined and it wont set those values
near quiver
nice!
native fable
https://github.com/AlubJ/GitHub.gml/issues/12
fixed and pushed
near quiver
while you're at it, its a small thing but helpful
when a callback or predicate function is an argument its good practice that the decription states the order of arguments of that function so people dont need to pull up the docs just to make sure if its index, element or element, index
looking at you yyg...
oh wait it does
native fable
ive added them in square brackets to make it more clear
near quiver
i actually looked here first i guess i was looking for something else
thats a good idea, stand out to the eye a bit more
native fable
pushed that too
near quiver
i swear i looked though 😭
native fable
dont worry about it lol
there are so many endpoints ive gotta add
anyway, im happy with the current state of the library
ofc theres still more to do
near quiver
Yeah im very happy this exists, the number of times ive had to write unit tests this will help a lot...
native fable
im glad it has its uses
near quiver
what would you say is the best way to parse issues to find one with a matching title?
just .getIssues()?
near quiver
yeah i see its got some filter options
native fable
if the number of open issues is more than 100 youll need to request a second page or more
near quiver
and eventually it will 404 yea?
native fable
unsure
near quiver
on way to find out page=999
also i noticed that in some places labels is an array and others its an array-like string
GitHubIssue is an array of strings
.getIssues is a string
native fable
if youre uploading or updating data, itll be an array
if youre getting data itll be an array-like string
its just because of the different ways its handled in the API
near quiver
ah gotcha
native fable
uploading or updating data requires that everything is sent in the body, whereas getting is a optional query parameter
i may edit those functions to accept an array or array-like string
and deal with it
native fable
and eventually it will 404 yea?
okay, paginated stuff returns back a link in the header that has information about the pagination
ill need to figure that stuff out
native fable
Yeah, next thing on the list is to parse the response headers and grab some vital information about the response
After that I can return to implementing more endpoints
native fable
got response headers returning as a struct in _request.responseHeaders
adding some helper functions to GitHub to get the last rate limit data
.getRateLimitLimit .getRateLimitUsed . getRateLimitRemaining .getRateLimitReset
returns back
5000
16
4984
28/11/2025 01:01:41
the github docs are not clear on this, but i dont know if the rate limits are per-oauth app / key or per-user that authenticated through said app
near quiver
I should consider now before it's a problem how I'll handle caching
native fable
you could 100% just store the raw request json as a data file
should i have a .cacheResult method in GitHubRequest?
near quiver
Nah, because it's likely user preferences on how to cache
Like I would want a file, others would want it in memory
native fable
yeah thats fair
native fable
ive added expire detection in device-flow authentication
i dont think that web-flow actually ever times out, but i may add a user-definable time-out time
oauth.setAuthenticationTimeoutCallback
this callback will fire when an authentication times out either by the max poll attempts being hit or the expire time expires
native fable
the web-server now has a shutdown time attached to it
if the server needs the be shutdown for any reason, it goes through a time source to delay the shutdown by a user configurable macro
native fable
repo content retrieval
native fable
I’m gonna take a break from this project for a little while
near quiver
you've done great work!
deep tide
would you accept PRs if there's something i need and end up making?
native fable
Yeah, I’ll keep reviewing PRs and accepting them etc
It’s open to all to contribute to
deep tide
k
native fable
Very intrigued on what you plan to do
deep tide
right now i'm mostly interested in auth to get around rate limits
if i'm gonna scale up Hotglue then the 60 requests per minute is gonna be a problem
native fable
The docs mention a rate limit of 5000 per hour when authenticated, tho I dunno if this is tied to a single user auth token or the client tokens
native fable
It’s a little annoying to require a user to sign in to actually do useful stuff with hotglue
deep tide
not sure what you mean
if someone doesn't want to sign in then all that happens is GitHub requests will need to be spaced out
everything else is still fine. the API endpoints i'm interested in are, at the moment, all public
native fable
not sure what you mean
I have immediately forgot what I mean too, ignore me
Currently at work and it’s a draaaag
near quiver
if i'm gonna scale up Hotglue then the 60 requests per minute is gonna be a prob...
I would also be interested in this. As gmlc will surely be submitting far more then 60 with in the span of an hour. If you need some assistance on this let me know
native fable
The simple answer is user authentication via device flow if you don’t wanna try to hide a client secret. Device flow only requires the client ID
I need to do more testing with multiple accounts to actually verify if it’s 5000 requests per hour per app or per authenticated user
native fable
The simple answer is user authentication via device flow if you don’t wanna try ...
Oh yeah, web-flow is only really supported for desktop, I don’t have the means to test any other platforms tho I’d imagine it won’t work on console
But 9 time out of 10 most apps that require API access on console do device flow
Think YouTube for TV
As far as I can tell, the login link is static, so you could easily have a QR code that directs to the login screen
deep tide
You can use a personal access token to make API requests. Additionally, you can authorize a GitHub App or OAuth app, which can then make API requests on your behalf.
All of these requests count towards your personal rate limit of 5,000 requests per hour.
so it's 5k per person per hour
which is plenty
native fable
Ah okay cool
deep tide
@native fable have you figured out how to view private repos for the auth'd user?
native fable
That should just come naturally, as long as the auth perms are set to “repo” I think
The auth scope
native fable
Yeah, I’ll be home in about 30 minutes and i can spin up a test for you
native fable
<@355405927425245194> have you figured out how to view private repos for the aut...
are you getting a 404 or 403 error when making a request to that users repo?
yeah, im able to retrieve private repos owned by the currently authenticated user
.requestAuthenticationViaWebPage(["repo", "read:user"]); same perms as hotglue
deep tide
i'm looking to get a list of repos for the user including private ones
the user/repos/ endpoint
couldn't get that to work
native fable
yeah, im able to retrieve private repos owned by the currently authenticated use...
yeah, this was a response from that endpoint
native fable
let me have a look
just a bog standard oauth app using webflow
are you testing using a github app as opposed to an oauth app?
native fable
ill pull hotglue tomorrow evening and have a prod around and see what i can do
deep tide
are you testing using a github app as opposed to an oauth app?
hhhmmmm maybe!
it was indeed a regular GitHub app
i will retest tomorrow
native fable
Morning
Yeah I’ve been using an oauth app for all my testing with webflow authentication
After looking at the GitHub app setup page there’s extra perms that can be set that oauth doesn’t have
GitHub wants auth to use GitHub app instead of oauth apps and I’m honestly not sure what the benefit is if any at all
When I get home later tonight i can take a proper look GitHub apps
deep tide
looks like using an OAuth app specifically does the trick
native fable
I’m sure that a GitHub app could work but needs to have specific perms attached, unsure but i will test later tonight
deep tide
i tried with the Metadata permission but no luck. i think the problem was that i'm using the OAuth flow instead of the GitHub App flow
native fable
Interesting
I may look into how app flow works, though I’d imagine I couldn’t make it secure enough for production
Or be able to get gamemaker to register that a browser wants to focus back to it like how GitHub desktop works
And probably other things, but like you said, gotta imagine it works then work backwards
deep tide
URI registration is a solved problem for Windows. however, having a running application elegantly pick up subsequent URI invocations isn't solved
deep tide
i came up with a workaround a few years ago but it's ugly stuff
native fable
Yeah, I wouldn’t imagine there would be an elegant solution
I’m curious to see how it actually works for native apps
native fable
Interesting
native fable
oh damn it's almost 9 years ago to the day i did URI registration
That’s really cool
That’s actually something I’m gonna look into
Might be able to do it using a c++ extension, hopefully I can get it working on Mac and Linux for a nice cross platform library
deep tide
Mac requires some configuration in the Xcode project itself https://developer.apple.com/documentation/xcode/defining-a-custom-url-scheme-for-your-app
native fable
Yuck
I dislike that
Linux is somewhat similar to windows but does not require registry edits because there isn’t one for Linux
native fable
Alright, researched all 3 desktop OSs URI handling and I have a plan for how to make it work
This looks like it’ll be a super fun project, and it is actually useful for actual games
mygame://join?room=Y81H7 or something
deep tide
Hotglue now has a URI registration solution for Windows
feel free to use it as reference, or ignore it entirely
whatever you figure out for Mac i will 100% steal
native fable
Ooh amazing! Did you get it to work with already running instances?
deep tide
yeah
so the trick is to spin up a network socket on a particular port when the application runs
when you execute the URI, you boot up a new instance of the game. when that new instance is created, that instance will fail to create a socket because of a port conflict
you then create a new socket on whatever port is available and send a UDP packet to the original instance and then self-destruct
the original instance then receives the URI string via the sent packet and then you can process it
you can also do this with saving files on disk as well. i'd prefer to run some of this logic in a batch file but i haven't gotten around to working out the process for that
native fable
Oh that’s clean!
I was for sure thinking I’d have to deal with multiple instance handling through a DLL
deep tide
that network socket thing is a nice trick. it's not flawless, however
file saving would be a lot more stable. there will be a way to do it but i've got a meeting with a client in a few minutes so i'll need to figure it out later
native fable
Me personally, I hate the idea of saving a file to disk to communicate between processes
I agree that it would probably be more stable
I wonder how performant checking the hash of a file is
My first thought would be check save data for a rando file, check its hash and compare with the last hash, if they’re different, register URI
Either that or check if that file exists at all, and delete after reading
deep tide
delete after reading is the way forwards
fortunately in this situation it's an infrequent operation to trigger the URI
native fable
That’s a positive
I’m very cautious around a lot of file writing, I’d rather not be the cause to someone’s SSD failing
deep tide
Hotglue is a tool that reads and writes hundreds of files per operation
i am a criminal, sir
native fable
🔫 stick em up
deep tide
native fable
i have an idea on how to get macos URI handling to work, i just cannot test it because i have not paid for my apple developer account yet
and thus i cannot build for macos
yet
the idea is using a build script to add the URL scheme into the apps plist, and resigning the app afterwards
native fable
I think I’m gonna work on bug testing all of my current methods for the GitHub API, make sure everything is up to scratch. Then I’ll do a first release (with admittedly not as many endpoints as I’d like)
deep tide
good luck!
native fable
Thank you kindly
near quiver
would it be possible to hook this up to automatically update the ide's version number from github release tags?
or you think that would be better off as a pre build script instead?
native fable
I just woke up, can you elaborate?
