fix(autocomplete, select, virtualrepeat, input): correctly validate attribute values

[devversion]

Fixes the validation of the boolean attributes.

Autocomplete

Attributes:

required

disabled

• required ignored an empty attribute value (<md-autocomplete required>)
• disabled ignored an empty attribute value (<md-autocomplete disabled>)

---

Input

Attributes:

md-no-asterisk

- `md-no-asterisk` is ignoring false attribute value (``)

Extra Changes:

• required is not updating the md-required class (now observing for ng-required)
• Wrong test-case for md-no-asterisk

---

Select

Attributes:

multiple

- `multiple` is ignoring false attribute value (``)

---

Virtual Repeat

Attributes

md-on-demand

- `md-on-demand` is ignoring false attribute value (``)

Extra Changes:

• Missing attribute test for md-on-demand

---

Notes

• Validating negative attribute values for required, readonly, disabled will be ignored.
• Validating 0 as falsy value is also a useful feature and should be implemented too, because sometimes your backend will return a boolean in numeric style (See here).

Fixes #5831 Fixes #7063 Fixes #6415 Fixes #6393 Fixes #5896 Fixes #6698 Fixes #7105

[ThomasBurleson]

Could the value not interpolate to '0' also ? This solutions seems limited.

[devversion]

@ThomasBurleson - It now supports interpolated 0 values. Added a util method for attribute parsing, this will be helpful for #6714 too (if a util method isn't desired let me know :) ) + accompanying tests.

[devversion]

-- Now fixing #6698 too, after merging my old PR into this by using the new util method.

[ThomasBurleson]

@jelbourn - I think we should regex match on the value string. Thoughts or code suggestions ?

[devversion]

Won't do a negated match for /(false|0)/g?

[jelbourn]

I wouldn't consider the "zero" string to be falsy. Without that the check is fine as-is.

[robertmesserle]

@devversion Can you remove the check for '0' and update the PR? Other than that, it looks good to me.

[devversion]

@robertmesserle - Yes will do, just thought I need that because Thomas asked about the interpolated zero as a falsy value.

[devversion]

@ThomasBurleson @robertmesserle @jelbourn - Okay removed the zero check (but kept the initial check structure). But in my opinion I think zero should be used as a falsy value too.

[robertmesserle]

Just to clarify, the use-case would be something like:

<md-autocomplete ... required="0"></md-autocomplete>

Is that right?

I don't have a strong opinion one way or the other, I'll defer to @ThomasBurleson and @jelbourn for this.

[devversion]

Okay, just wan't to answer your question 😄 .

• Yes exactly this will allow using zero as false value too. This would be really useful when using something like a backend which is returning booleans in numeric style.

[devversion]

@ThomasBurleson At the moment in this PR, in most cases we support required="false" required="true", so the required attribute is accepting a boolean as value. But this is not a default behavior in HTML from W3C.

The users should better use ng-disabled, ng-readonly and ng-required.
So IMO we should not parse the attribute values from these type of attributes.

• disabled
• readonly
• required

Just tell me, then I will remove the new checks for these types. But please think about the issues too (#5896, #6415)

[jelbourn]

While treating these as boolean attributes is correct, it would be a breaking API change.

[devversion]

@jelbourn Sorry, I didn't get it. Should we continue this.. or should I remove that?

Edit: Ah, you answered to @robertmesserle's message.

[EladBezalel]

@jelbourn according what I've read this PR doesn't seems like a breaking API, it just enables wider range of values, more than that the old behavior won't break.

[jelbourn]

@EladBezalel We would be changing the behavior of, for example, <md-select disabled="false">. Someone might be depending on this already, so changing it would be a breaking API change.

[devversion]

@jelbourn I've changed that a few days ago, it's currently only a issue of the autocomplete directive (required, disabled). Should I add a parameter to the util function which excludes the negative checks (this can be used for attributes like readonly, disabled, required), because in general I need the negative checks for attributes like md-no-asterisk or multiple. I think this will be a good solution. Thoughts?

BTW: md-chips also retrieves a boolean (like <md-chips readonly="true"> - <md-chips readonly> isn't working)

[wembernard]

@devversion May I suggest to add required attribute in select section?

[devversion]

@wembernard I would like to have this PR merge, before looking at the actual select issue, because this may not relate to the actual boolean validation.

BTW: As discussed with @jelbourn I removed the breaking attribute changes. PTAL

[leocaseiro]

Will this PR fix the issue with multiple="false"?

If yes, I vote +1

UPDATE

I found the solution for multiple="false", we should bind using ng-multiple. As @rschmukler mentioned at #5459

[devversion]

@leocaseiro Yes, that will be a good workaround for that, and yes this PR adds a basic boolean validation for multiple, this should be added, because the only attributes which should not validate for negated boolean values are readonly, required, disabled...

[leocaseiro]

how about ng-required, ng-multiple, ng-disabled and ng-readonly?

[devversion]

@leocaseiro Yes exactly that's the right usage for these attributes. By default readonly, required, disabled (which are native attributes - without the md- prefix) are not interpolating a boolean value. That's why you should use the ng- prefixed attributes for these three native attributes, when you wan't to interpolate the boolean.