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

Dynamic tags in glimmer templates #389

Merged
merged 4 commits into from
Apr 12, 2019
Merged

Conversation

cibernox
Copy link
Contributor

@cibernox cibernox commented Oct 14, 2018


I think that this repetition makes templates harder to read, so I advocate that only simple bindings
can be passed in. If the user wants to use something more complex than a binding, they could use the `let`/`with` helpers
for that.
Copy link
Contributor

@buschtoens buschtoens Oct 14, 2018

Choose a reason for hiding this comment

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

Thinking of how I used to use tagName, I believe disallowing expressions would cause a lot of boilerplate. The old tagName always defaulted to div (or the tagName that was set on the component class), which was quite handy when you only sometimes needed to explicitly set a tagName from the invoking template.

Keeping this ability would be really nice. One can either easily do this by allowing an expression like you've shown.

The alternatives would either entail the {{#with}} construct or a computed property on the component class. Both feel boilerplatey to me.

However, I do agree that, if the same expression would be repeated multiple times in the template, or if it is ridiculously complex, users should not inline it.
Instead of outright banning this, we could also add such a linting rule to ember-template-lint.

Copy link
Contributor

Choose a reason for hiding this comment

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

if the same expression would be repeated multiple times in the template

Oh, I forgot about the closing tag. Maybe we could allow some sort of "infering" for the closing tag?

<{{or @tagName "div"}}>
</>

Copy link
Contributor Author

Choose a reason for hiding this comment

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

I considered it, but it looks ugly to me.
I know that at this point glimmer templates don't have to respect html rules, but the expectation that opening and closing tag are identical is very engrained in my brain.

@Baukereg
Copy link

There's a potential risk of dynamic tag names. Changing the tag name from the outside might break HTML conventions or cause conflicts between default styles of the browser and styles defined by the component. I feel like the component being solely responsible for its semantics is good practice.

Maybe these use cases can be solved by lifting the container element from the component's template?

<section>
  <JustContent />
</section>

<aside>
  <JustContent />
</aside>

@cibernox
Copy link
Contributor Author

cibernox commented Oct 14, 2018

I'm not sure I follow. Do you mean that changing from a span to a button can break styles? If so, I'd say that it's totally expected and not an issue with dynamic tag names, but with how browsers behave. But it's up to the user.

Tbh in my use cases the tagName is never changed in runtime once set. Maybe that is acceptable too. I'll add that to the open questions

@tomdale
Copy link
Member

tomdale commented Oct 14, 2018

We had actually considered a very similar syntax for component invocation in the past (as an alternative to the current angle bracket syntax). I agree that if we want to support something like this, this is the most immediately intuitive syntax.

My biggest concern is how much complexity supporting dynamic tag names pushes into Glimmer VM. Unlike almost everything else, the DOM really really really does not want you to treat tag name as mutable. Most things that look dynamic in a template—like binding an attribute, for example—are relatively simple to implement because we can call setAttribute or removeAttribute.

Making tagName dynamic means that we'd need to create an entirely new element, which means we have to think about how this interacts with lifecycle hooks like didInsertElement, whether child nodes/components get moved or re-rendered, etc. In general, I've come to be very, very nervous about syntax that looks simple and cheap, but comes with surprising behavior or performance implications (e.g. @each).

How would you feel about the constraint that tag name could be dynamic, but immutable? That is, you could set tag name based on a runtime value, but if that property changed after the component was rendered, it would produce a runtime error. This would make the implementation much simpler, and would eliminate many of the edge cases I'm worried about.

I also agree with @buschtoens on the concerns about arbitrary expressions and closing tags. For me, the rule is:

  • Things that look like expressions should act like expressions.

Unfortunately, this rule is difficult to apply in this scenario. The proposed syntax looks like an expression, so it would be very confusing why <{{@tagName}} would work but not <{{or @tagName "div"}}.

On the other hand, </{{or @tagName "div"}}> also looks like an expression, and we don't really have any precedent for "no-op expressions" in the syntax, as far as I can recall. Maybe we can rely on the ugliness to push people away from using it unless they really need it. 😉

@cibernox
Copy link
Contributor Author

cibernox commented Oct 14, 2018

@tomdale That's exactly what I was thinking when I said:

Tbh in my use cases the tagName is never changed in runtime once set. Maybe that is acceptable too. I'll add that to the open questions

I was updating the RFC to propose that tagName is dynamic but unbound. I was aware that tagName is not a property you can change on a DOM node, and changing it would mean creating a whole new element.

It totally seems a good compromise. I'll update the RFC.

About allowing expressions in tagNames, I was concerned because of ugliness, nothing else. What do think is that the opening and closing tags must have the exact same expression, not just one that resolves to the same value.

For instance:

<{{or @tagName "div"}}
  The content
</{{if @tagName @tagName "div}}>

Both expression are semantically equivalent, but they are not the ideantical. I propose that this would not even compile instead of waiting for runtime to see if expression result on the same value.

@cibernox cibernox force-pushed the dynamic-tag-names branch 4 times, most recently from a3b5824 to f87795d Compare October 14, 2018 23:13
@chancancode
Copy link
Member

chancancode commented Oct 15, 2018

Have you considered modeling this as a dynamic component invocation instead?

{{#let (element "li") as |Tag|}}
  <Tag ...>...</Tag>
{{/let}}

The element helper will produce a contextual component where, when invoked, will produce an element with the specified tag name. Personally I would rather see it modeled like this, as the concepts an infrastructure mostly already exists today. In fact, I am pretty sure you can already implement that in user land using Ember.Component.

@cibernox
Copy link
Contributor Author

cibernox commented Oct 15, 2018

@chancancode That is not super different to what I proposed in the alternatives:

<DynamicElement @tagName={{@tagName}} class="some-class" ...attributes>
</DynamicElement>

I still prefer to interpolate inside the just after the open markers. I believe it reads better and it's what I'd have naively tried if I was a newcomer to Ember.
Also, IDK what the long term plan for Ember.Component, but at some point outer-html components should be the default, right?
Splitting hairs your proposal also seems more expensive, as it involves let helper + the new element helper to create a contextual component + using that component (with the overhead of instantiating it, running the lifecycle hooks, etc...)

What I care more about is about having the feature. If there is technical reasons at the parser/VM level that make that syntax problematic and we should have a built-in component or helpers I can live with it.

@NullVoxPopuli
Copy link
Sponsor Contributor

I don't think interpolation of expressions should be allowed in tagName.

Sure, it could lead to boilerplate, but I'd hope with glimmer templates, changing the tagName in more the exception than the rule, so, using computeds should be ok

@chancancode
Copy link
Member

chancancode commented Oct 17, 2018

I think there are several reasons that I think the “interpolation” syntax is problematic:

  1. <{{this.tag}} ... looks to similar to <this.tag ... etc
  2. It would be unexpected for a {{...}} to be unbound
  3. Having to repeat a long or complicated expression in the closing tag position is noisy and error-prone

I published ember-element-helper based on my comment above. I encourage everyone to try it out and see if this works well for the use cases we have in mind in the real world. If it works out, we can implement this natively which would eliminate any performance concerns.

I am open to adding an <Element @tagName=... /> version too, but to me, it would be unexpected that it tearsdown the block when @tagName changes, whereas the sub expression form makes it pretty clear in my head (the input to the sub expression helper changes, which produced a different component, so the old component gets torn down).

@cibernox
Copy link
Contributor Author

@chadhietala for what I can see, this approach requires relies on the tagged nature of @ember/component. Will that approach continue to work if the default component manager for Ember eventually becomes like the one for glimmer-components?

  1. I see you the similarity
  2. My idea wasn't not so much to have a {{...}}} be unbound but to make it error if mutated
  3. Agree

You've convinced me that having a dedicated helper or component is probably a better solution. What I'm not convinced yet is if this let + element + <Tag> combination is too weird and the convenience of having a special built-in component (<DynamicTag> for instance) can beat explicitness here.

All things said is that this not a feature that is used all the time, so I could live with an inconvenient syntax if that makes things clearer or the implementation easier.

@chriskrycho
Copy link
Contributor

Having just come to this specific proposal but having thought on this a lot b/c we do occasionally use dynamic tags… I'm inclined to say the explicitness of let + element + <Tag> is a win overall. i think it's one less thing to teach – the semantics are pretty obvious if you understand each of those pieces separately – and the use cases are pretty narrow. Most of the time I find myself reaching for dynamic tag name, when I step back it's not the right solution. There are specific cases for addons in particular where it is, for sure. In our app code, though, I find that either specializing the components or using {{yield}} and contextual/higher-order components makes for a much more maintainable and much clearer to others consuming the code—especially when taking a11y into account.

@chadhietala
Copy link
Contributor

I really like @chancancode's suggestion here as it allows for addon authors to explore the usefulness of this pattern. I think that if this ends up being a very popular, perhaps we can consider adding syntax that basically does an expansion to what @chancancode's addon does but elides the component creation.

@chancancode
Copy link
Member

chancancode commented Oct 17, 2018

@cibernox Correct. The "user land" implementation relies on Ember.Component (and so one undesirable side-effect is you get the "ember-view" CSS class). I'm not sure if there is a plan for changing the behavior of that, or removing it from core, etc. For the foreseeable future it will probably co-exist with "Glimmer Components", at least that would be my personal guess. However, that shouldn't matter anyway, as I am not proposing that we abandon this RFC. Rather, once we have some real world experience to draw from, we can amend this RFC with what we settled on and implement that in core. With that I am quite certain that it can be as efficient as any of the current proposals.

I agree with you an others that this is sort of a power-user feature that should be used rarely, so at the end of the day ergonomics probably doesn't matter so much.

To elaborate on my (weakly-held) position that the teardown behavior/expectation is more expected with this setup:

{{#let (element this.tagName) as |Tag|}}
  <Tag>hello world</Tag>
   ~~~ since this thing is dynamic, it makes sense to me that when this changes, everything goes away
{{/let}}

{{!-- whereas... --}}

<Element @tagName={{this.tagName}} as |Tag|>
         ~~~~~~~~ it is less expected that changing this will make the whole outer thing go away
  hello world
</Element>

I am still open to changing my mind on that one, since it is more of a heuristic than a rule. You can easily write a component that violates that mental model today, I just personally think it's uncommon and probably a bad pattern, so I'm less inclined to adopt that kind of API in core.

@cibernox
Copy link
Contributor Author

I think you changed mine. I'll update my proposal.
Also, if you really want to enfonce bound tagNames the developer always can use the {{unbound}} (which was popular in the old times but I haven't used it in ages) helper for that:

{{#let (element (unbound this.tagName)) as |Tag|}}
  <Tag>hello world</Tag>
   ~~~ since this thing is dynamic, it makes sense to me that when this changes, everything goes away
{{/let}}

@JonForest
Copy link

JonForest commented Oct 18, 2018

We've written a fairly large-scale dynamic form implementation that simply would not be possible without dynamic tags.
Despite having 10s of different component types, logic between form elements, logic between forms, conditional display, etc. we really only use dynamic tags in two components (one for view and edit)

{{#each formElements as |formElement|}}
  {{#if (not (hide-in-edit formElement))}}
      {{component (concat 'dynamic-form/edit/' formElement.type)
        formElement=formElement
        state=state
        ....
  {{/if}}
{{/each}

And this could be comfortably handled with the let + element + <Tag> syntax I'd imagine.

@chancancode
Copy link
Member

@JonForest 👍 that's a pretty useful data-point! when you get a chance, I would love for you to try the add-on in your app to confirm that it works for your use cases.

@rwjblue
Copy link
Member

rwjblue commented Oct 19, 2018

I'm pretty 👍 on the current rendition of this, thanks for iterating towards such a nice solution!

tylerturdenpants added a commit to emberjs/website that referenced this pull request Oct 19, 2018
Ideas, feel free to add to list or claim:
- [ ] ```I've been getting a lot of questions about how tree-shaking is coming along. I would be willing to train anyone that wants to help on what's already done and what still needs to be done. Disclaimer: It's a lot of work! #emberjs #EmberFest``` https://twitter.com/kellyselden/status/1050717338595745792
- [ ] Include summary from pixelhandler (and cc him) #issue-triage team: https://discordapp.com/channels/480462759797063690/480523776845414412/500377829452546058
- [x] From jenweber on #support-ember-times: Another times topic: emberjs/ember.js#16910 I would summarize as, "ember.js has an addition to the test suite to help with API documentation! In the past, you might have noticed that a small number of methods  had missing docs. When code gets moved around in ember.js, such as putting functions in their own modules, it's easy to make mistakes that impact the documentation parser. This test builds the docs and checks them for unintentional changes. Thanks <link to ed> for the test and to everyone who helped fix missing docs over the years." Todd Jordan did the majority of the fixes when things got dropped, I think. Hard to say for sure 🔏 @kennethlarsen 
- [x] emberjs/rfcs#384 (:lock: @jessica-jordan)
- [x] emberjs/rfcs#386 (:lock: @jessica-jordan)
- [ ] emberjs/rfcs#387 (:lock: @jessica-jordan)
- [x] emberjs/rfcs#388 (:lock: @Alonski)
- [ ] emberjs/rfcs#389
- [x] EmberConf brainstorming session date, if confirmed (:lock: @amyrlam)
- [ ] Hacktoberfest roundup?
- [ ] Check https://github.com/jessica-jordan/whats-new-in-emberland
- [x] [ember-self-focused](https://engineering.linkedin.com/blog/2018/10/making-ember-applications--ui-transitions-screen-reader-friendly) (🔒 @chrisrng )
- [ ] ...

Ideas we are waiting on:
- [ ] EmberCamp talks, deep dive? http://embercamp.com/ and https://github.com/ember-chicago/ember-camp-2018-notes ... Maybe we wait for the talk videos to be published? Keep an eye on #ember-camp in Discord. not out yet
- [ ] GraphQL with Ember? https://emberfest.eu/schedule/#rocky-neurock Or maybe we can reach out to them for a post-EmberFest writeup? See also the blog post: https://medium.com/kloeckner-i/ember-and-graphql-8aa15f7a2554
@chancancode chancancode self-assigned this Oct 19, 2018
@rwjblue rwjblue added T-templates T-framework RFCs that impact the ember.js library labels Oct 24, 2018
@chancancode
Copy link
Member

OOPS! Clicked the wrong button by accident!


## How we teach this

The new helpers must be documented with examples in the API and Guides.
Copy link
Contributor

Choose a reason for hiding this comment

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

One thing we are trying hard to do better is make sure that RFCs include enough information to the "how we teach this" section to be an MVP documentation for the guides. In order for this RFC to move forward, some more content would need to be added here.

Additionally, this RFC seems like it will give folks a lot more surface area to do detrimental things re: a11y, and it would be important to address this in this section, even if it's more of a warning to authors to make sure they keep that in mind. 👍

@chancancode
Copy link
Member

@hakubo today on the core team call we confirmed the modifier restriction for components is unintentional. I'll write up a quick amendment RFC to clarify that today and lift that restriction in Ember soon.

@chancancode
Copy link
Member

Today we discussed this on the core team call, we felt that it was pretty close, but as @MelSumner mentioned we felt that we can flesh out the teaching section more. In addition to what @MelSumner said (examples, accessibility), @wycats and @krisselden also felt that there are probably some small XSS risk with this and we should at minimum add some warning to the teaching section.

@wycats
Copy link
Member

wycats commented Jan 18, 2019

Thinking about it a little, I think that there is some correctness risk with accepting arbitrary tag names (for example xlink: namespaces, etc.) and possible security risks.

I'd like to restrict the values to element names that look like simple HTML tags: lowercase tag names (including dashes); no colons, no sigils, etc.

Given these risks, it is also probably good to find something to say in the documentation about the risks, but that doesn't need to hold up the RFC.

@cibernox
Copy link
Contributor Author

@MelSumner @chancancode I've updated the RFC implementing the feedback!

Example:

```hbs
{{#let (element "li") as |Tag|}}
Copy link
Member

Choose a reason for hiding this comment

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

maybe (element this.tagName) – using a literal string here is a little pointless in this example

```
Since the element name can be defined in runtime, in order to protect the user from XSS vulnerabilities,
the passed tag name cannot contain anything but lowercase letters and dashes. Tag names containing
other characters (p.e `xlink:`) will throw an error.
Copy link
Member

Choose a reason for hiding this comment

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

e.g. ?

## How we teach this

If this RFC is accepted, Ember would get a new built-in helper to be added to `component`, `array`, `each` and in
[Ember.Templates.helpers](https://emberjs.com/api/ember/release/classes/Ember.Templates.helpers).
Copy link
Member

Choose a reason for hiding this comment

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

The documentation will need a at least somewhat meaningful and self-contained example to illustrate the point of the helper (and how to use it). Could you provide one? I assume you have some examples from your addons that could be distilled down.

[Ember.Templates.helpers](https://emberjs.com/api/ember/release/classes/Ember.Templates.helpers).

Since this feature is not very commonly used I think that updating the section of the API docs above is enough,
no need to mention it in the guides.
Copy link
Contributor

@MelSumner MelSumner Feb 10, 2019

Choose a reason for hiding this comment

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

Since using something like dynamic tags could have a serious impact on accessibility, I think it's very important to mention this in the documentation.

I can see dynamic tags being used by developers who believe they are being very clever, but in fact are being quite careless due to their ignorance. Especially considering places that don't encourage developer education, such as enterprise environments - where the end product has an even higher chance of being used by a large number of users, and as such, a larger percentage of folks who need things to be accessible.

The more I think about this RFC the more I feel very uneasy about it- I think the very least we can do is ensure that we inform developers of the potential foot guns associated with dynamic tag use.

Copy link
Member

Choose a reason for hiding this comment

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

@MelSumner do you have some examples of the potential foot guns you have in mind that @cibernox can incorporate them into the text here?

Copy link
Contributor

Choose a reason for hiding this comment

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

@chancancode good idea. I can help.

Copy link
Contributor

@MelSumner MelSumner Feb 10, 2019

Choose a reason for hiding this comment

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

@chancancode @cibernox here's the general idea of what I think should be included, at a minimum- whether or not this is in the guides or the API could perhaps be discussed via the learning core team (although I would probably recommend a reference in both). This is a pretty high risk area, imo.

How We Teach This

It's vitally important that developers have awareness the negative impact that misuse of dynamic tags can have on the accessibility of an Ember application.

While dynamic tags enable a great deal of flexibility in components, it's essential to recognize that these can also negatively impact accessibility. Developers should be keenly aware of the appropriate role that should be applied per HTML element (as specified in the WAI-ARIA specification), and ensure that the role is also updated, if necessary, as the tag name is changed. In some cases, the use of native, semantic HTML elements may eliminate the need to apply a role at all, so developers should consult the WAI-ARIA specification until they are certain of the correct use for their specific use cases.

If this RFC is accepted, Ember would get a new built-in helper to be added to `component`, `array`, `each` and in
[Ember.Templates.helpers](https://emberjs.com/api/ember/release/classes/Ember.Templates.helpers).

Since this feature is not very commonly used I think that updating the section of the API docs above is enough,
Copy link
Member

Choose a reason for hiding this comment

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

Assuming the current structure of the guide, this should fit into the Components > Customizing a Component's Element section. @jenweber can weigh in on where it will fit in the new guides (though I don't think it's necessary this RFC's job to hypothesis that).

While I agree that 1) the current coverage of "what goes into API docs vs guides" for template helpers is a bit ad-hoc and inconsistent, 2) it is not good to just basically duplicate the same information twice in both places, I think there are some good guiding principles of what goes where:

  • API docs is narrowly focused on a single helper. The objective is "so, you've figured out you want to use this helper, let me teach you how". It can cover things like the arguments it takes, return value, examples, (to an extent) use cases it solves, caveats/cautions and so on.

  • Guides should be more topical, focusing on the general use case as well as cross-cutting concerns. The objective is "so, you have a problem, you may not know how to solve it yet so let me present to you some options". It serves more of a discovery role, and can showcase different ways of doing the same thing, the pros and cons of each, examples, but can defer to the API docs for the specific details.

So, I think the point of including this in the guides is that, it is quite likely that you are aware of the problem but didn't doesn't yet know about the element helper as the solution you are looking for.

Copy link
Member

Choose a reason for hiding this comment

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

(That's only my personal take)

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 @chancancode. My own recommendation would be to briefly describe it in the guides, and link to the API docs from within the Components > Customizing a Component's Element section. The bulk of the explanation and use case with example would then go in the API docs, to avoid duplication. We are trying to move towards using the Guides to show how things fit together in real apps, rather than covering the API surface.


For accessibility and semantic reasons, sometimes a `<div>` may not be the best kind of tag.
We might want the panel to be a `<section>` element, or an `<aside>` when it's content is somewhat unrelated
with the rest of the page. Or maybe a `<legend>` if it's and the end of a form.
Copy link
Member

@chancancode chancancode Feb 10, 2019

Choose a reason for hiding this comment

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

Let's add something like

With the element helper proposed in this RFC, this can be accomplished with something like this:

{{#let (element @tagName) as |Tag|}}
  <Tag class="pt-10 pb-10 ps-20 box-shadow" ...attributes>
    {{yield}}
  </Tag>
{{/let}}

We might want the panel to be a `<section>` element, or an `<aside>` when it's content is somewhat unrelated
with the rest of the page. Or maybe a `<legend>` if it's and the end of a form.

## Detailed design
Copy link
Member

Choose a reason for hiding this comment

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

I suggest a reordering of the paragraphs.

We propose to add a new element helper that takes a tag name and generates a contextual component that, when invoked, renders the selected element.

Example:

...

Unlike ids, classes or other attributes, the tag name of DOM element cannot be changed in runtime. ...

To help the user understand that changing the tag of an element in runtime is an expensive operation, the syntax is chosen such that...

Since the element name can be defined in runtime, in order to protect the user from XSS vulnerabilities, ...

...


```hbs
{{!-- sidebar.hbs --}}
{{#let (element this.htmlTag) as |Tag|}}
Copy link
Member

Choose a reason for hiding this comment

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

{{!-- sidebar.hbs, a template-only component --}}
{{#let (element @htmlTag) as |Tag|}}

This way, we can eliminate the classic component class, both to simplify the example, and to avoid the awkwardness of "couldn't you just use tagName"?

I'm aware that we will loose the default aside tag name here, but maybe that's okay for the purpose of this example. Otherwise, something like {{#let (element (if @htmlTag @htmlTag "aside")) as |Tag|}} would be fine also.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

If ember-truth-helpers where part of core, the pattern I use the most is:

{{#let (element (or @htmlTag "aside") as |Tag|}}
  <Tag ...attributes>...</Tag>
{{/let}}

That's why I resorted to having a JS file.

Copy link
Member

@chancancode chancancode Feb 10, 2019

Choose a reason for hiding this comment

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

I agree. I think for the sake of the example, either not caring about having a default, using the if dance, or using a hypothetical or is all fine. However, using tagName: '' and essentially re-implementing the feature in user land is.... causing distraction at minimum. Classic component classes doesn't really have the problem that motivates the feature imo – my understanding is that we are designing this feature to backfill something we used to be able to do there, so using classic components to illustrate it doesn't quite work imo.

@cibernox
Copy link
Contributor Author

Another update.

@chancancode
Copy link
Member

@cibernox sorry for the delay (due to EmberConf, Octane, etc). I'll nominate this for FCP as soon as we resume the regular meetings (next Friday, probably).

@jenweber
Copy link
Contributor

jenweber commented Apr 5, 2019

Having another read through, I would like if this RFC had a full example of what it looks like for someone to update the ARIA roles, as the tag changes. We will have to write those docs and this goes best if the content is more detailed, so that people who are unfamiliar with the problem space can help out.

@chancancode
Copy link
Member

We discussed this on the framework team meeting today and believe this is ready for FCP!

@chancancode
Copy link
Member

chancancode commented Apr 5, 2019

@jenweber it kind of depends on the use case.

1. If, a component is internally doing the switch based on some argument, it would probably be something like this:

{{#let (element this.tagName) as |Tag|}}
  <Element role={{this.role}} ...attributes>...</Element>
{{/let}}
class MyComponent extends GlimmerComponent {
  get tagName() {
    return this.args.simulated ? 'div' : 'a';
  }

  get role() {
    // unsure if there are problems with just always adding the "link" role,
    // even when using a native "a" element
    return this.args.simulated ? 'link' : undefined;
  }
}

It's kind of contrived, but the intent of this example is to switch between using a native "a" tag or using making a different element and make "simulated link" instead.

Generally, I think there are very few (if any) good use cases for this where using the element helper is appropriate. If you know the variants ahead of time, you should just enumerate them in the template. A better example:

{{#if @actAsButton}}
  <button (extra button specific things here)>...</button>
{{else if @actAsSwitch}}
  <button (extra switch specific things here)>...</button>
{{else}}
  <a (extra link specific things here)>...</button>
{{/if}}

To share the content implementation (the "...") here, extract this wrapper part into a separate component instead.

2. If, a component is intending to allow the caller to customize the element, it should be the caller's responsibility to provide the tag name as long as the aria roles and other appropriate aria attributes:

{{#if @customTag}}
  {{#let (element @customTag) as |Tag|}}
    <Tag ...attributes>...</Tag>
  {{/let}}
{{else}}
  <aside ...attributes>...</aside>
{{/if}}

Here the ...attributes will forward any attributes (aria or otherwise) from the caller. I think this should be the main use case for the element helper.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Final Comment Period T-framework RFCs that impact the ember.js library T-templates
Projects
None yet
Development

Successfully merging this pull request may close these issues.