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

SelectControl: Infer value type from options #64069

Merged
merged 8 commits into from
Aug 6, 2024
Merged

Conversation

mirka
Copy link
Member

@mirka mirka commented Jul 29, 2024

Updated take on #60293 (props @mikeybinns)

What?

Tighten the TypeScript type checking on SelectControl, so the value type matches in the value, options, and onChange props.

See Dev Notes at the bottom for how this looks like to a consumer.

Why?

It can catch potential bugs.

How?

Adding a generic that can either be inferred from the options array (if it is immutable), or be set explicitly.

Testing Instructions

See some of the static type tests in the unit test file, and play around by adding your own cases and see what errors.


✍️ Dev Note

Stricter type checking in SelectControl

In TypeScript codebases, SelectControl will now be able to type check whether the value type matches in the value, options, and onChange props.

This will be inferred automatically if the options array is declared as immutable.

<SelectControl
	// @ts-expect-error "baz" is not in `options` array
	value="baz"
	options={ [ // array literal that cannot be mutated
		{
			value: 'foo',
			label: 'Foo',
		},
		{
			value: 'bar',
			label: 'Bar',
		},
	] }
/>;
const options = [
	{
		value: 'foo',
		label: 'Foo',
	},
	{
		value: 'bar',
		label: 'Bar',
	},
] as const; // const assertion to mark as readonly

<SelectControl
	// @ts-expect-error "baz" is not in `options` array
	value="baz"
	options={ options }
/>;

If the options array is mutable, value will remain a simple string type.

This is effectively the same type checking behavior as before.

const options = [
	{
		value: 'foo',
		label: 'Foo',
	},
	{
		value: 'bar',
		label: 'Bar',
	},
];

<SelectControl
	// No error, because `value` has type `string`
	value="baz"
	options={ options }
/>;

You can also set an explicit value type.

<SelectControl< 'foo' | 'bar' >
	// @ts-expect-error "baz" is not in `'foo' | 'bar'`
	value="baz"
	// @ts-expect-error "baz" is not in `'foo' | 'bar'`
	options={ [ { value: 'baz', label: 'Baz' } ] }
/>;

@mirka mirka added [Type] Enhancement A suggestion for improvement. [Package] Components /packages/components labels Jul 29, 2024
@mirka mirka self-assigned this Jul 29, 2024
Copy link
Member Author

Choose a reason for hiding this comment

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

The changes we have to make in this file are what make me a bit hesitant about adding this value type inference. The stricter types do not break runtime code obviously, but we will be forcing some consumers to deal with this new strictness, which may seem like overkill depending on who you ask.

I think we can get away with it, because the strictness will only kick in if the options array is defined as immutable (either a literal array in the component call, or with an as const). For example, there is a potential bug in CustomGradientPicker but it isn't caught with the new strictness because GRADIENT_OPTIONS is mutable.

We don't have a ton of data points in the type checked realm of this repo, but it seems like TimePicker is the only "false positive" in the sense that it didn't surface a potential bug. (I just want to be sure that this type tightening is a net positive and not a nuisance 😅)

Copy link
Contributor

Choose a reason for hiding this comment

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

I agree with what you're saying. It would likely be a good thing to document how to specify a loose type (by using SelectControl<string>()) so that users who don't want to tighten their other code (in this case, the format function) can still have the same level of leniency as they previously had, and then this change empowers users to use tighter types if they like.

Though, saying that, I'm not sure where this would be documented. In the SelectControl docs as a note perhaps?

Copy link
Contributor

Choose a reason for hiding this comment

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

I just want to be sure that this type tightening is a net positive and not a nuisance 😅

This is definitely a big aspect to keep in mind. Overall, I think this change is a net positive — I hope that in the future we'll have a way to always apply the strictness, and not only when the options array is immutable.

We should probably write a dev note for this change, to help consumers of the package deal with this change correctly (especially the ones who need to support multiple versions of Wordpress / Gutenberg).

Copy link
Member Author

Choose a reason for hiding this comment

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

I added a Dev Note so we at least have something to link to if anybody asks. If we do get some questions, we might need to document it somewhere more permanent. But in general I don't think we should have "TypeScript documentation" aside from our actual type files.

Comment on lines +88 to +92
value={
orderBy === undefined || order === undefined
? undefined
: `${ orderBy }/${ order }`
}
Copy link
Member Author

Choose a reason for hiding this comment

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

This was an actual bug.

Comment on lines 69 to 76
const newValues = selectedOptions.map(
( { value } ) => value as T
);
props.onChange?.( newValues, { event } );
return;
}

props.onChange?.( event.target.value, { event } );
props.onChange?.( event.target.value as T, { event } );
Copy link
Member Author

Choose a reason for hiding this comment

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

These type assertions should be safe, as they will always match the options types.

Copy link
Contributor

Choose a reason for hiding this comment

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

Technically, this isn't quite true because the select could be edited by someone using dev tools, however, I think this is fine because

  1. If you edit a website with dev tools, it's your fault you broke the code, not the code maintainer
  2. The likelihood of this happening is next to none

Copy link
Contributor

Choose a reason for hiding this comment

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

Would you be willing to use OptionType instead of T here? T doesn't really mean anything, and while it's a common practice in TypeScript, I see it as equivilent to naming a variable like const a = 'something'.

Copy link
Member Author

Choose a reason for hiding this comment

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

I changed them to V (for value — it's not actually the option type), does that read a bit better? We don't have official naming conventions for this project yet, so eventually we may decide on a prefixed convention like TValue. But to write out a full name like Value, I think it's harder to signal that it's a generic. And ValueType is redundant because everything is a type 😅 We can revisit, but I hope V is a good compromise for now.

@@ -24,7 +24,7 @@ type SelectControlBaseProps = Pick<
Pick< BaseControlProps, 'help' | '__nextHasNoMarginBottom' > & {
onBlur?: ( event: FocusEvent< HTMLSelectElement > ) => void;
onFocus?: ( event: FocusEvent< HTMLSelectElement > ) => void;
options?: {
options?: readonly {
Copy link
Member Author

Choose a reason for hiding this comment

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

Adding a readonly here so it allows immutable arrays. (Otherwise it would be a type error when passing an immutable array.)

* @default false
*/
multiple?: false;
value?: NoInfer< T >;
Copy link
Member Author

Choose a reason for hiding this comment

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

NoInfer<> is a new utility type added in TS 5.4, which tells TS not to use this field to infer the generic type. We need this because we want it to infer from the values in the options array, not the value prop.

This kind of "don't infer" signaling used to be done with T & string, but does not work anymore in TS 5.4+.

Copy link
Member

Choose a reason for hiding this comment

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

Can it be demonstrated when the NoInfer can be useful? A new test case that fails without it? I don't understand very well what it does.

Copy link
Contributor

Choose a reason for hiding this comment

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

Sure, Without NoInfer Typescript will infer the generic type from the value prop as well as the options prop.

What this means is it won't detect that your value is not one of the provided options, and therefore won't provide an error for it.

I've created a simple TS Playground example to illustrate the difference for you. Hope that helps 🙂

Copy link
Member

Choose a reason for hiding this comment

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

Thanks, this is nice 🙂 I didn't realize that passing something as value can expand the T type to also include the type of value. But we want the type to be checked against options instead.

Copy link
Member

Choose a reason for hiding this comment

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

Nice to see a clever usage of NoInfer 👍 I had read about it, but struggled to think of a good use case for it.

Copy link
Contributor

Choose a reason for hiding this comment

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

Also found a good article about NoInfer: https://www.totaltypescript.com/noinfer

Copy link
Contributor

@mikeybinns mikeybinns left a comment

Choose a reason for hiding this comment

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

I have some personal nit-picks, but ultimately approve it because I'd be fine if those were ignored :)

Comment on lines 69 to 76
const newValues = selectedOptions.map(
( { value } ) => value as T
);
props.onChange?.( newValues, { event } );
return;
}

props.onChange?.( event.target.value, { event } );
props.onChange?.( event.target.value as T, { event } );
Copy link
Contributor

Choose a reason for hiding this comment

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

Technically, this isn't quite true because the select could be edited by someone using dev tools, however, I think this is fine because

  1. If you edit a website with dev tools, it's your fault you broke the code, not the code maintainer
  2. The likelihood of this happening is next to none

Comment on lines 69 to 76
const newValues = selectedOptions.map(
( { value } ) => value as T
);
props.onChange?.( newValues, { event } );
return;
}

props.onChange?.( event.target.value, { event } );
props.onChange?.( event.target.value as T, { event } );
Copy link
Contributor

Choose a reason for hiding this comment

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

Would you be willing to use OptionType instead of T here? T doesn't really mean anything, and while it's a common practice in TypeScript, I see it as equivilent to naming a variable like const a = 'something'.

@mikeybinns
Copy link
Contributor

One thing that's missing from my PR is the automated tests, perhaps it's worth copying those over?

* Otherwise, the value received is a single value with the new selected value.
*/
onChange?: (
value: T,
Copy link
Contributor

Choose a reason for hiding this comment

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

Should we also use the NoInfer utility for the onChange's arguments, in case a consumer of the component explicitly types onChange?

Copy link
Member Author

Choose a reason for hiding this comment

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

Good one, I added the constraints (and type tests) for those too.

Copy link
Contributor

Choose a reason for hiding this comment

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

I just want to be sure that this type tightening is a net positive and not a nuisance 😅

This is definitely a big aspect to keep in mind. Overall, I think this change is a net positive — I hope that in the future we'll have a way to always apply the strictness, and not only when the options array is immutable.

We should probably write a dev note for this change, to help consumers of the package deal with this change correctly (especially the ones who need to support multiple versions of Wordpress / Gutenberg).

mirka and others added 3 commits August 2, 2024 03:24
@mirka mirka added the has dev note when dev note is done (for upcoming WordPress release) label Aug 1, 2024
@mirka
Copy link
Member Author

mirka commented Aug 1, 2024

One thing that's missing from my PR is the automated tests, perhaps it's worth copying those over?

You're right, I brought over your tests and reformatted them a bit to mesh with our existing type tests (ad20004). I tried to pare it down so we have a good cost/benefit balance between safety and maintenance burden — hope that's reasonable.

@mirka mirka marked this pull request as ready for review August 1, 2024 19:26
@mirka mirka requested a review from ajitbohra as a code owner August 1, 2024 19:26
Copy link

github-actions bot commented Aug 1, 2024

The following accounts have interacted with this PR and/or linked issues. I will continue to update these lists as activity occurs. You can also manually ask me to refresh this list by adding the props-bot label.

If you're merging code through a pull request on GitHub, copy and paste the following into the bottom of the merge commit message.

Co-authored-by: mirka <0mirka00@git.wordpress.org>
Co-authored-by: mikeybinns <mikeybinns@git.wordpress.org>
Co-authored-by: ciampo <mciampini@git.wordpress.org>
Co-authored-by: jsnajdr <jsnajdr@git.wordpress.org>
Co-authored-by: tyxla <tyxla@git.wordpress.org>
Co-authored-by: DaniGuardiola <daniguardiola@git.wordpress.org>

To understand the WordPress project's expectations around crediting contributors, please review the Contributor Attribution page in the Core Handbook.

@mirka
Copy link
Member Author

mirka commented Aug 1, 2024

I discussed this with a few other folks and nobody is outright opposed, so I think we can move forward to an official review so this can be merged 🙏

@tyxla tyxla requested a review from a team August 2, 2024 16:22
Copy link
Contributor

@DaniGuardiola DaniGuardiola left a comment

Choose a reason for hiding this comment

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

Good stuff! Straightforward implementation. Nice touch with the type tests, though it makes me wonder if we should add standards for type testing and/or support for tooling like https://github.com/MichiganTypeScript/type-testing - the eslint ignore comments for jest feel a bit like a hack.

As a reference, here's an example from a personal project where I do extensive type tests: https://github.com/DaniGuardiola/rpc-anywhere/blob/main/src/tests/types-test.ts

Other than that, I have no notes (other than a minor nit)!

'select',
false
> & { ref?: React.Ref< HTMLSelectElement > }
) => ReturnType< typeof UnforwardedSelectControl >;
Copy link
Contributor

Choose a reason for hiding this comment

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

I see what you were going for here, but I think I'd simplify the return type to just React.JSX.Element | null. That's the only thing that a component can ever return anyway, and it looks clearer to me this way. I don't really mind though, this is a nit.

Copy link
Member Author

Choose a reason for hiding this comment

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

That makes sense, thanks 🙏 2f195c2

mirka added 2 commits August 7, 2024 06:16
# Conflicts:
#	packages/components/CHANGELOG.md
#	packages/components/src/select-control/index.tsx
#	packages/components/src/select-control/test/select-control.tsx
#	packages/components/src/select-control/types.ts
@mirka
Copy link
Member Author

mirka commented Aug 6, 2024

Nice touch with the type tests, though it makes me wonder if we should add standards for type testing and/or support for tooling like https://github.com/MichiganTypeScript/type-testing - the eslint ignore comments for jest feel a bit like a hack.

@DaniGuardiola It has come up, most recently in #62400 (comment). I don't think we're quite at the point where we need dedicated type test tooling, as the downsides are still outweighing the upsides in my opinion (e.g. having to manage the tooling package updates, forcing contributors to learn yet another testing library, etc).

But I do think we should start to document some standards/conventions like you suggest, as this is the fourth time we're doing static type tests here!

@mirka mirka enabled auto-merge (squash) August 6, 2024 21:46
@mirka mirka merged commit de33f91 into trunk Aug 6, 2024
61 checks passed
@mirka mirka deleted the infer-select-control-types branch August 6, 2024 22:08
@github-actions github-actions bot added this to the Gutenberg 19.0 milestone Aug 6, 2024
@mikeybinns
Copy link
Contributor

Thanks for all your work on this @mirka ☺️

@mirka
Copy link
Member Author

mirka commented Aug 7, 2024

@mikeybinns Thanks for all your work as well 🙌

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
has dev note when dev note is done (for upcoming WordPress release) [Package] Components /packages/components [Type] Enhancement A suggestion for improvement.
Projects
None yet
Development

Successfully merging this pull request may close these issues.

7 participants