Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Introduce JSDoc-based Type System with checkJs flag #5318

Merged
merged 44 commits into from
Dec 6, 2023

Conversation

mkslanc
Copy link
Contributor

@mkslanc mkslanc commented Sep 17, 2023

Issue #, if available:

Description of changes:
This PR introduces a comprehensive type system for our plain JavaScript project, leveraging the power of JSDocs and TypeScript's checkJs feature.

Details:

  • All primary JavaScript files have been adorned with JSDoc type annotations to ensure type safety without converting the project to TypeScript.
  • Configured tsconfig.json to enable checkJs, allowing TypeScript to check our plain JavaScript files for type errors.
  • Refactored some parts of the codebase to address type inconsistencies and to ensure more robust typing.

Benefits:

  • Improved code clarity with explicit type annotations.
  • Early detection of type-related errors, leading to more robust code.
  • Provides a foundation for potential migration to TypeScript in the future, should the need arise.
  • Enhances the developer experience by providing type hints and autocompletion in supported editors.

P.S.: I left //@ts-expect-error with TODO in different places in codebase, where I think some bugs could occur
P.P.S.: not whole classes/ functions covered with types right now, somewhere I left any types, because it already become quite big change. Could be addressed in separate PR.

  • Updated JSDoc's for ace-code base
  • Pre-build ace.d.ts generator for ace-code and ace-builds
  • Fixed modes declarations
  • Resolved TODO's

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

@whazor
Copy link
Contributor

whazor commented Sep 18, 2023

Nice work!! I agree on the any types, it would be best for the first PR to only add the start point, and we could slowly improve types over time.


text.style.opacity = "0";
parentNode.insertBefore(text, parentNode.firstChild);

var copied = false;
var pasted = false;
/**
*
* @type {false | {[key: string]: any}}}
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Any opinion about this style vs Record<string, any>?

return types.indexOf(token.type || token) > -1;
};

CstyleBehaviour.recordAutoInsert = function(editor, session, bracket) {
CstyleBehaviour["recordAutoInsert"] = function(editor, session, bracket) {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I would rather aim to make it work as it is, without trying to please TS language server too much by rewriting everything into ["..."]. You probably just need to define a proper typedef.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Generally I agree with you, but lets leave it for further improvements (especially whole mode/* part)

src/mode/text.js Outdated Show resolved Hide resolved
@kungfooman
Copy link
Contributor

Very nice work! 👍

I just happen to develop a tool right now which parses JavaScript projects with JSDoc via Babel and then adding automatic type assertions at runtime. I use it so far only for the PlayCanvas game engine and it helped me to squeeze out a bunch of bugs already and I'm looking to support more projects. However, it's under heavy construction right now because I prepare for an official release to make it work for all kinds of different projects:

https://github.com/kungfooman/RuntimeTypeInspector.js

(You can watch the first video to see how it works)

Once I have more time, I will look into the integration here. Very happy to see better JSDoc across all kinds of projects 💪 💯 🥇

@mkslanc
Copy link
Contributor Author

mkslanc commented Sep 18, 2023

@nightwing, @whazor, @akoreman, @InspiredGuy I'm hoping to gather thoughts and suggestions on a few decisions we're considering regarding new type system, because it would influence everyone's experience:

  1. What is the preferred method for storing interface types such as Point/Position? Should we place them in ace.d.ts or utilize @typedef within the ace source files?
  2. Do we intend to generate the corresponding .d.ts files after each PR merge, or is it solely for internal development to improve type assistance?
  3. If we are generating .d.ts files, should the main declaration file depend on this generated declaration instead of the source code?
  4. What would be the most effective method for creating declarations for ace-builds?
  5. Regarding code style, should we use @typedef at the beginning of a file for type aliasing, such as:
/**
 * @typedef {import("../ace").Ace.Point} Point
 */

For each type sourced from ace.d.ts? Or, should we maintain it as import("../ace").Ace.Point for every parameter since we cannot create a type alias for the namespace, to the best of my knowledge?

@kungfooman
Copy link
Contributor

For each type sourced from ace.d.ts? Or, should we maintain it as import("../ace").Ace.Point for every parameter since we cannot create a type alias for the namespace, to the best of my knowledge?

Typedef type aliasing has pretty much two issues:

If you don't care about two mouse clicks and just wanna put a ts-ignore over such export * from '...', you can use typedefs.

@marinsokol5
Copy link
Contributor

Hey, nice work!! 🚀

I think checkJs flag is a great addition to the codebase as indeed it gives us that nice extra verification step that stuff are as we type them to be.

I am kind of lost though with what we want to achieve with the generated types:

  • We generate .d.ts files from JSDoc and we put them into types folder but we don't publish this folder, so we still fully rely on ace.d.ts instead. So why do we generate them?
  • Like do we expect consumers to use an import from types like import {CommandBarTooltip} from "ace-code/types/ext/command_bar";
  • Or do we expect them to set "allowJs": true, in their tsconfig.json and then import from source by import {CommandBarTooltip} from "ace-code/src/ext/command_bar";

Do you know of any other popular open source project relying on JSDoc?
I am also curious on if the rely on types exported from their js files or do they maintain .d.ts files.

@mkslanc
Copy link
Contributor Author

mkslanc commented Sep 22, 2023

  • We generate .d.ts files from JSDoc and we put them into types folder but we don't publish this folder, so we still fully rely on ace.d.ts instead. So why do we generate them?

It could provide better intellisense for internal development purposes, but right now I also don't see a lot of benefits on generating types except the case when it would help to convert project to plain typescript.

  • Like do we expect consumers to use an import from types like import {CommandBarTooltip} from "ace-code/types/ext/command_bar";
  • Or do we expect them to set "allowJs": true, in their tsconfig.json and then import from source by import {CommandBarTooltip} from "ace-code/src/ext/command_bar";

Personally, I anticipate consumers won't have to change their usage much, which is why I retained the type aliases in ace.d.ts. The primary idea is to provide better types for external users and make it easier for new contributors to understand the codebase.

Do you know of any other popular open source project relying on JSDoc? I am also curious on if the rely on types exported from their js files or do they maintain .d.ts files.

I don't know of many projects using this approach, as it's often a temporary solution before migrating to full TypeScript. However, some examples include:

P.S.: many cases of uses mixins (EventEmitter and OptionsProvider mostly) in aces classes make us to provide

declare module "ace-code/src/editor" {
    export const Editor: {
        new(renderer: import("./src/virtual_renderer").IVirtualRenderer, session?: import("./src/edit_session").IEditSession, options?: Object): import("./src/editor").IEditor;
    };
}

or

import {Ace} from "../ace";
import {IEditSession} from "../src/edit_session";
import {IVirtualRenderer} from "../src/virtual_renderer";
export const Editor: {
    new(renderer: IVirtualRenderer, session?: IEditSession, options?: Object): Ace.Editor;
};

for any public ace module (actually everything in src dir)
and those who uses direct imports from ace-code codebase like import {Editor} from "ace-code/src/editor"; . At least I couldn't find better solution right now. If this is deemed an acceptable solution, I can write a generator based on the TS Language Service.

@nightwing nightwing marked this pull request as ready for review September 26, 2023 00:33
@mkslanc
Copy link
Contributor Author

mkslanc commented Nov 2, 2023

Also, according to different structure of ace-code and ace-builds it's a bit challenging to solve everything and not break something in process :)

@marinsokol5
Copy link
Contributor

marinsokol5 commented Nov 2, 2023

@mkslanc Looks like you got quite far in the process indeed!

Can we keep it without Step 3 though?
So can we introduce JSDoc + checkJS mechanism without the .d.ts updates right now, so no significant changes to ace.d.ts nor auto generating it yet from JSDoc.
And then we have the separate discussion for how do we plan to maintain our types (do we auto-generate them only or do we have separate .d.ts or a mix of both).

That way we improve the situation right now and merge the majority of this PR but lower the risk of introducing consumer changes.

@mkslanc
Copy link
Contributor Author

mkslanc commented Nov 2, 2023

@marinsokol5 I think it shouldn't be hard to separate those parts - I will leave only JSDoc's changes + tsconfig in that PR and create separate PR with generator. But maybe we will need to separate "typing" in "package.json" for internal and external (published) usage. I need to do few tests to be sure of that

@mkslanc
Copy link
Contributor Author

mkslanc commented Nov 2, 2023

@marinsokol5:

  1. For ace-code, we will use the newly corrected ace.d.ts for both external and internal usage. (It worked surprisingly smoothly for me.)
  2. For ace-builds: Renamed old ace.d.ts to ace-internal.d.ts to prevent breaking changes.
  3. The generator will be in a separate PR, and it seems we could use it only for ace-builds to generate one large declaration file, similar to what Monaco does.

Please test it thoroughly. I don't want to introduce any regressions to the codebase. :)

P.S.: from my side I tested it in different packages, where we used ace-builds and ace-code (ace-linters, ace-layout, ace-samples) and it worked.

@marinsokol5
Copy link
Contributor

marinsokol5 commented Nov 3, 2023

@mkslanc

Is it possible to keep ace.d.ts as it is before this PR? I understand that types will be a bit outdated and we won't leverage our JSDoc exports, but it would be just temporary duplication. We can open that PR right after we merge this one, and then we figure out the auto-generation and what to do ace-code vs ace-build. Just for easier reviewing and a more specific focus of the review.

@mkslanc
Copy link
Contributor Author

mkslanc commented Nov 3, 2023

@marinsokol5 It's possible, but I will need to change all import("../ace").Ace. to import("../ace-internal").Ace. If this is acceptable, I will make changes

@marinsokol5
Copy link
Contributor

@mkslanc Is it possible to not have ace-internal.d.ts at this moment in time, just ace.d.ts?

@mkslanc
Copy link
Contributor Author

mkslanc commented Nov 3, 2023

@marinsokol5 Unfortunately, no. Types in JSDoc's are highly dependent on ace.d.ts declarations and especially on declare module "./src/virtual_renderer" with interfaces like that - export interface VirtualRenderer extends Ace.EventEmitter<Ace.VirtualRendererEvents>, Ace.OptionsProvider<Ace.VirtualRendererOptions>. You could try to remove any declare module statement (most representable would be removing editor/edit_session/virtual_renderer) and see drastic increase of errors in type checking.

@mkslanc
Copy link
Contributor Author

mkslanc commented Nov 3, 2023

@marinsokol5 Also we could set "checkJs" to false, but it would nullify all advantages of type-checking

@nightwing
Copy link
Member

@mkslanc sadly we need to generate the .d.ts files otherwise this example is failing ajaxorg/ace-samples@master...ts, and we may even need the generated file to pass strict check which is specified in the failing test. but not sure about that.

Copy link

codecov bot commented Dec 4, 2023

Codecov Report

Attention: 33 lines in your changes are missing coverage. Please review.

Comparison is base (e8674fc) 87.58% compared to head (07b0241) 87.59%.

Files Patch % Lines
src/keyboard/gutter_handler.js 78.12% 7 Missing ⚠️
src/lib/event.js 68.18% 7 Missing ⚠️
src/editor.js 37.50% 5 Missing ⚠️
src/autocomplete.js 80.95% 4 Missing ⚠️
src/ext/inline_autocomplete.js 88.88% 2 Missing ⚠️
src/virtual_renderer.js 50.00% 2 Missing ⚠️
src/ext/static_highlight.js 83.33% 1 Missing ⚠️
src/keyboard/textinput.js 83.33% 1 Missing ⚠️
src/lib/report_error.js 0.00% 1 Missing ⚠️
src/line_widgets.js 75.00% 1 Missing ⚠️
... and 2 more
Additional details and impacted files
@@            Coverage Diff             @@
##           master    #5318      +/-   ##
==========================================
+ Coverage   87.58%   87.59%   +0.01%     
==========================================
  Files         583      583              
  Lines       46369    46389      +20     
  Branches     7016     7018       +2     
==========================================
+ Hits        40610    40636      +26     
+ Misses       5759     5753       -6     
Flag Coverage Δ
unittests 87.59% <85.65%> (+0.01%) ⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

@mkslanc
Copy link
Contributor Author

mkslanc commented Dec 4, 2023

@mkslanc sadly we need to generate the .d.ts files otherwise this example is failing ajaxorg/ace-samples@master...ts, and we may even need the generated file to pass strict check which is specified in the failing test. but not sure about that.

@nightwing Separate internal and external declaration files, should make it for now. I think generator should be separate PR, especially considering the fact that we need to run it in strict mode

src/selection.js Outdated
@@ -77,7 +69,7 @@ class Selection {

/**
* Returns an object containing the `row` and `column` current position of the cursor.
* @returns {Object}
* @returns {import("../ace-internal").Ace.Point}
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Could you convert this to a typedef, and do the same for other instances where same import() is used several times in a file?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, sure, but users would need to click twice to see the real type declaration. If this is ok, I would process with changes across codebase

@nightwing nightwing force-pushed the better-types branch 2 times, most recently from 79a25f8 to c6398d6 Compare December 5, 2023 23:19
@nightwing nightwing merged commit a7412b7 into ajaxorg:master Dec 6, 2023
2 of 3 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

5 participants