Skip to content
This repository has been archived by the owner on Apr 12, 2024. It is now read-only.

wip: feat(ngModel): run formatters / setModelValue #16237

Merged
merged 1 commit into from
Sep 29, 2017

Conversation

Narretz
Copy link
Contributor

@Narretz Narretz commented Sep 21, 2017

What kind of change does this PR introduce? (Bug fix, feature, docs update, ...)
feature

What is the current behavior? (You can also link to an open issue here)
There's currently no way to manually run the model -> view pipeline / formatters

What is the new behavior (if this is a feature change)?
An API to run the whole pipeline / the formatters is introduced.

Does this PR introduce a breaking change?
No

Please check if the PR fulfills these requirements

Other information:

The PR includes both a $format and a $setModelValue function. I think only one is needed, we just need to decide which one.

$format:

  • PRO: does the most needed thing, very straightforward: run the formatters and updates the viewValue
  • CON: developer must run $render manually
  • CON: empty classes are set even though the DOM is not updated. Could be moved to $render()
  • CON: no equivalent on the view -> model side (inconsistent API)

$setModelValue

  • PRO: equivalent to $setViewValue, runs the whole pipeline
  • CON: unintuitive that the function argument should / must be set to the current $modelValue, so the control does not get out of sync with the scope (same behavior as $setViewValue though)
  • CON: only runs $render if the viewValue has actually changed.
  • CON: always runs the validation, even though it might not be necessary

Both methods can handle the basic case, where an app developer wants to run the formatters the view -> model pipeline has been run, see #3407 or #5221

I personally tend to introduce $format as it has the smaller surface area and introduces fewer side effects. The full model -> view pipeline is not really needed for most cases.

@Narretz Narretz added this to the 1.6.x milestone Sep 21, 2017
@Narretz Narretz changed the title Feat model set wip: feat(ngModel): run formatters / setModelValue Sep 21, 2017

function ngModelWatchAction(modelValue) {
// if scope model value and ngModel value are out of sync
// TODO(perf): why not move this to the action fn?
Copy link
Contributor

Choose a reason for hiding this comment

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

Can delete this now


function ngModelWatchAction(modelValue) {
// if scope model value and ngModel value are out of sync
// TODO(perf): why not move this to the action fn?
if (modelValue !== ctrl.$modelValue &&
Copy link
Contributor

Choose a reason for hiding this comment

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

Can this be deleted now?

@@ -921,7 +924,8 @@ function setupModelWatcher(ctrl) {
}

return modelValue;
Copy link
Contributor

Choose a reason for hiding this comment

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

Can this be deleted now?

Copy link
Contributor

Choose a reason for hiding this comment

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

I guess we still need to return this because we need the digest to cycle again if the model value has actually changed...?

Copy link
Contributor

Choose a reason for hiding this comment

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

Yeah this comment is outdated now. This was from when this was moved to the listener fn...

@Narretz Narretz force-pushed the feat-model-set branch 2 times, most recently from 4eede61 to 09f501d Compare September 25, 2017 09:40
@Narretz
Copy link
Contributor Author

Narretz commented Sep 25, 2017

@jbedard I had to revert the code that was moved to the watchActionFn, because it caused test failures. Specifcally, it caused failures when the model was updated inside ngChange or the model setter - so I think the way it's done is necessary.

@petebacondarwin
Copy link
Contributor

petebacondarwin commented Sep 25, 2017

What about something like $processModelValue, which is effectively the same as your $setModelValue but doesn't take an argument, instead just using whatever is in th current $modelValue?

@Narretz
Copy link
Contributor Author

Narretz commented Sep 25, 2017

Hm, that sounds like a good in-between option. It' doesn't have a public equivalent for view -> model but $$parseAndValidate could be exposed if necessary.

@gkalpak
Copy link
Member

gkalpak commented Sep 25, 2017

Generally I prefer to keep the API consistent (it makes it much easier to wrap your head around), but in this case $processModelValue() sounds like a good idea.

BTW, I agree it makes sense to tie updateEmptyClasses() with the rendering.

@Narretz
Copy link
Contributor Author

Narretz commented Sep 26, 2017

Î'll update the PR then. Regarding $$updateEmptyClasses(), I thought about it again, and think it's okay that this is set before $render() is called. It's also more of an issue if we had a $format() function, because that would not have called render.
But $processModelValue calls everything in the pipeline, including $render(), so it's not a big deal that $$updateEmptyClasses() is called right before it.
On the view -> model pipeline it's also only called when the viewValue is committed, so it's also not directly tied to the actual DOM / render value.

@Narretz
Copy link
Contributor Author

Narretz commented Sep 26, 2017

Updated.
I left $$format and $$setModelValue as internal functions, and added $processModelValue with a PoC autocomplete example to illustrate the use case.

Copy link
Member

@gkalpak gkalpak left a comment

Choose a reason for hiding this comment

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

Left some docs-related comments.
Otherwise LGTM 👍

(ctrl.$modelValue === ctrl.$modelValue || modelValue === modelValue)
// checks for NaN is needed to allow setting the model to NaN when there's an asyncValidator
// eslint-disable-next-line no-self-compare
(ctrl.$modelValue === ctrl.$modelValue || modelValue === modelValue)
Copy link
Member

Choose a reason for hiding this comment

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

Nice try! But the indentation i still off 😛

* @description
*
* Runs the model -> view pipeline on the current
* {@link ngModel.NgModelController#$modelValue $modelValue}
Copy link
Member

Choose a reason for hiding this comment

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

Missing . at the end.

* Application developers usually do not have to call this function themselves.
*
* This function can be used when the $viewValue or the rendered DOM value of the control should
* be updated after a user input.
Copy link
Member

Choose a reason for hiding this comment

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

It is not necessarily about user input. The way I understand it, it is for when app state has changed in a way that may affect formatting. (OK, I admit most of the time the state changes as a result of user input, but still... 😛)

Copy link
Contributor Author

Choose a reason for hiding this comment

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

That's right. This is is just the most common case. Another would be that the formatters have changed but the model hasn't. If the model changes, the normal model -> view code should be enough. Unless someone changes the $modelValue directly, which is not recommended. I'll make this a it more generic.

* be updated after a user input.
*
* For example, consider a text input with an autocomplete list (for fruit, where the items are
* objects with a name and an id.
Copy link
Member

Choose a reason for hiding this comment

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

Missing closing ).

* For example, consider a text input with an autocomplete list (for fruit, where the items are
* objects with a name and an id.
* A user enters `ap` and then selects `Apricot` from the list.
* Based on this, the autocomplete widget will call $setViewValue({name: 'Apricot', id: 443}),
Copy link
Member

@gkalpak gkalpak Sep 27, 2017

Choose a reason for hiding this comment

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

It is better to wrap code in `. (E.g. $setViewValue(...) here or ctrl.$processModelValue() and $viewValue below.)

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Do you wrap in inline code blocks? It looks like your formatting is a bit off

Copy link
Member

Choose a reason for hiding this comment

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

Apparently if you want to escape `, you need to place the \ before, not after the ` 😇

* but the rendered value will still be `ap`.
* The widget can then call ctrl.$processModelValue() to run the model -> view
* pipeline again, which includes a formatter that converts the object into the string `Apricot`
* which is set to the $viewValue and rendered in the DOM.
Copy link
Member

Choose a reason for hiding this comment

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

Consider avoiding two which clauses. E.g. something line:

The widget can then call ctrl.$processModelValue() to run the model -> view pipeline, which will use a formatter to convert the object to the string Apricot, update the $viewValue and finally rendered it in the DOM.


ngModel.$parsers.push(function(value) {
if (angular.isString(value)) {
return scope.items.find(function(item) {
Copy link
Member

Choose a reason for hiding this comment

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

This is not ES5-compatible 😁

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Do we care? This only affects IE for core support, but I don't think we make guarantees for the docs

Copy link
Member

Choose a reason for hiding this comment

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

I don't feel too strongly, but it would be good for the docs to show code that works in all supportd browsers. It would also be a first (afaik).

</file>
<file name="index.html">
<div style="display: flex;">
<div margin-right: 30px;">
Copy link
Member

Choose a reason for hiding this comment

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

Forgot about style 😃

<input ng-model="val" process-model />

<ul>
<li ng-repeat="item in items | filter:val"><button ng-click="select(item)">{{item.name}}</li>
Copy link
Member

Choose a reason for hiding this comment

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

Ehem...relying on a method set on the scope by a directive on a different element doesn't sound like a good practice 😉
Maybe move all relevant code into the directive's template.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

It's just a POC that doesn't show correct encapsulation.
I thought about putting it in the directive's template but then it gets more complex too. I'd have to create another directive for the ngModel formatters / parsers, or search for the input directive in the autocomplete directive ... no child selectors in AngularJS unfortunately. Something like this: https://plnkr.co/edit/OOSjPSQag2huO6gtLFgg?p=preview
Idk if it's worth the complexity

Copy link
Member

Choose a reason for hiding this comment

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

I was having something like this in mind: https://plnkr.co/edit/AeOA4eMGGMONSTrLinH2?p=preview
(TBH, I find this whole usecase weird. It sounds like an unnecessary complicated way to create an autocomplete 😃)

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Well, that's less complex and shows better encapsulation. I'll take it. But for the record, I don't think a real world autocomplete would be successful with a hard-coded input element. At least not for me ;)

You know the customer is always right, and apparently many devs implement autocomplete like this.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Updated with a small tweak

Copy link
Contributor Author

@Narretz Narretz left a comment

Choose a reason for hiding this comment

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

I've fixed the typos and commented on the more tricky stuff

* Application developers usually do not have to call this function themselves.
*
* This function can be used when the $viewValue or the rendered DOM value of the control should
* be updated after a user input.
Copy link
Contributor Author

Choose a reason for hiding this comment

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

That's right. This is is just the most common case. Another would be that the formatters have changed but the model hasn't. If the model changes, the normal model -> view code should be enough. Unless someone changes the $modelValue directly, which is not recommended. I'll make this a it more generic.

* For example, consider a text input with an autocomplete list (for fruit, where the items are
* objects with a name and an id.
* A user enters `ap` and then selects `Apricot` from the list.
* Based on this, the autocomplete widget will call $setViewValue({name: 'Apricot', id: 443}),
Copy link
Contributor Author

Choose a reason for hiding this comment

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

Do you wrap in inline code blocks? It looks like your formatting is a bit off


ngModel.$parsers.push(function(value) {
if (angular.isString(value)) {
return scope.items.find(function(item) {
Copy link
Contributor Author

Choose a reason for hiding this comment

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

Do we care? This only affects IE for core support, but I don't think we make guarantees for the docs

<input ng-model="val" process-model />

<ul>
<li ng-repeat="item in items | filter:val"><button ng-click="select(item)">{{item.name}}</li>
Copy link
Contributor Author

Choose a reason for hiding this comment

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

It's just a POC that doesn't show correct encapsulation.
I thought about putting it in the directive's template but then it gets more complex too. I'd have to create another directive for the ngModel formatters / parsers, or search for the input directive in the autocomplete directive ... no child selectors in AngularJS unfortunately. Something like this: https://plnkr.co/edit/OOSjPSQag2huO6gtLFgg?p=preview
Idk if it's worth the complexity

@Narretz Narretz force-pushed the feat-model-set branch 3 times, most recently from 1ee7a7f to 6f08ae8 Compare September 27, 2017 19:06
@@ -921,7 +924,8 @@ function setupModelWatcher(ctrl) {
}

return modelValue;
Copy link
Contributor

Choose a reason for hiding this comment

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

I guess we still need to return this because we need the digest to cycle again if the model value has actually changed...?

@Narretz Narretz merged commit 05fdf91 into angular:master Sep 29, 2017
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Projects
None yet
Development

Successfully merging this pull request may close these issues.

5 participants