Skip to content

CHANGELOG.md

Latest commit

 

History

History
431 lines (262 loc) · 21.1 KB

CHANGELOG.md

File metadata and controls

431 lines (262 loc) · 21.1 KB

dom-accessibility-api changelog

0.6.3

Patch Changes

0.6.2

Patch Changes

  • #978 0128d16 Thanks @hiebj! - Add explicit typings location to support use in TypeScript ESM modules

0.6.1

Patch Changes

0.6.0

Minor Changes

0.5.16

Patch Changes

  • #902 df97b80 Thanks @eps1lon! - Fix a case of maximum call stack size exceeded when a node referenced itself in aria-labelledby.

0.5.15

Patch Changes

  • #893 d5af41d Thanks @eps1lon! - Don't consider title in 2E

    Effectively ensures that title will not have precedence over name from content. For example, the option in <option title="Title">Content</option> will now have "Content" as its accessible name instead of "Title".

0.5.14

Patch Changes

  • #827 a1daca5 Thanks @nolanlawson! - Follow aria-labelledby and aria-describedby if they point to elements in the same shadow root.

0.5.13

Patch Changes

  • #811 5b0f48e Thanks @eps1lon! - Prefer button subtree over title attribute.

     const name = computeAccessibleName(<button title="from-title">from-content</button>);
-'from-title' === name
+'from-content' === name

    <button title="from-title">from-content</button> would previously compute the accessible name "from-title". This is correct in ACCNAME 1.2 but is changed in the latest editors draft. The latest editors draft specifically refers to HTML-AAM which says that the subtree should take precedent over the title attribute. computeAccessibleName now calculates "from-content" as the accessible name.

0.5.12

Patch Changes

  • #800 de554b0 Thanks @pablo-abc! - Remover circular dependency, which fixes warnings thrown in certain environments.

0.5.11

Patch Changes

  • #796 cb38778 Thanks @calebeby! - <input type="number" /> now maps to role spinbutton (was textbox before).

0.5.10

Patch Changes

  • #770 7066180 Thanks @eps1lon! - Allow computing name for inaccessible elements

    This is mainly targetted at integration with @testing-library/dom. But it can also be used as a general performance boost when used in a JSDOM environment. The rationale is that most elements are part of the a11y tree. In those cases computing a11y tree exclusion is wasted. Since it's expensive, we can disable it. The recommendation is to only ignore a11y tree inclusion locally and specifically enable it for the tests where you do know that a11y tree inclusion will play a role.

0.5.9

Patch Changes

  • #762 b3e4a17 Thanks @eps1lon! - Compute name from author for menu role.

    Previously we wouldn't compute any name for menu to pass some web-platform-tests that covered an exotic use case. Now we correctly respect name from author (e.g. aria-label or aria-labelledby).

0.5.8

Patch Changes

0.5.7

Patch Changes

0.5.6

Patch Changes

  • #666 26ee73d Thanks @eps1lon! - Consider <label /> when computing the accessible name of <output />

    Given

    <label for="outputid">Output Label</label> <output id="outputid"></output>

    Previously the accessible name of the <output /> would ignore the <label />. However, an <output /> is labelable and therefore the accessible name is now computed using <label /> elements if they exists. In this example the accessible name is "Output Label".

0.5.5

Patch Changes

  • #627 0485441 Thanks @eps1lon! - Ensure certain babel helpers aren't required

    Source:

    -const [item] = list;
+const item = list[0];

    Transpiled:

    -var _trim$split = list.trim().split(" "),
-_trim$split2 = _slicedToArray(_trim$split, 1),
-item = _trim$split2[0]
+var item = list[0];

  • #629 383bdb6 Thanks @eps1lon! - Use label attribute for naming of <optgroup> elements.

    Given

    <select>
	<optgroup label="foo">
		<option value="1">bar</option>
	</optgroup>
</select>

    Previously the <optgroup /> would not have an accessible name. Though 2D in accname 1.2 could be interpreted to use the label attribute:

    Otherwise, if the current node's native markup provides an attribute (e.g. title) or element (e.g. HTML label) that defines a text alternative, return that alternative [...]

    This was confirmed in NVDA + FireFox.

0.5.4

Patch Changes

  • 3866289 #442 Thanks @geoffrich! - Correctly determine accessible name when element contains a slot.

    Previously, computing the accessible name would only examine child nodes. However, content placed in a slot is is an assigned node, not a child node.

    If you have a custom element custom-button with a slot:

    <button><slot></slot></button>

<!-- accname of inner <button> is 'Custom name' (previously '') -->
<custom-button>Custom name</custom-button>

    If you have a custom element custom-button-default with default content in the slot:

    <button><slot>Default name</slot></button>

<!-- accname of inner <button> is 'Custom name' (previously 'Default name') -->
<custom-button-default>Custom name</custom-button-default>

<!-- accname of inner <button> is 'Default name' (previously 'Default name') -->
<custom-button-default></custom-button-default>

    This is not currently defined in the accname spec but reflects current browser behavior.

0.5.3

Patch Changes

  • 76e8f93 #430 Thanks @ckundo! - Maintain img role for img with missing alt attribute.

    Previously <img /> would be treated the same as <img alt />. <img /> is now treated as role="img" as specified.

  • 96d4438 #436 Thanks @eps1lon! - Resolve presentational role conflicts when global WAI-ARIA states or properties (ARIA attributes) are used.

    <img alt="" /> used to have no role. By spec it should have role="presentation" with no ARIA attributes or role="img" otherwise.

0.5.2

Patch Changes

  • 03273b7 #406 Thanks @eps1lon! - Fix various issues for input types submit, reset and image

    Prefer input value when type is reset or submit:

    <input type="submit" value="Submit values">
-// accessible name: "Submit"
+// accessible name: "Submit values"
<input type="reset" value="Reset form">
-// accessible name: "Reset"
+// accessible name: "Reset form"

    For input type image consider alt attribute or fall back to "Submit query".

0.5.1

Patch Changes

  • fcc66ae #394 Thanks @marcosvega91! - Ignore title attribute if it is empty.

    Previously <button title="">Hello, Dave!</button> would wrongly compute an empty name.

0.5.0

Minor Changes

  • 9e46c51 #380 Thanks @eps1lon! - BREAKING CHANGE

    Ignore ::before and ::after by default.

    This was necessary to prevent excessive warnings in jsdom@^16.4.0. If you use this package in a browser that supports the second argument of window.getComputedStyle you can set the computedStyleSupportsPseudoElements option to true:

    computeAccessibleName(element, {
	computedStyleSupportsPseudoElements: true,
});

computeAccessibleDescription(element, {
	computedStyleSupportsPseudoElements: true,
});

    If you pass a custom implementation of getComputedStyle then this option defaults to true. The following two calls are equivalent:

    computeAccessibleName(element, {
	computedStyleSupportsPseudoElements: true,
});

computeAccessibleName(element, {
	getComputedStyle: (element, pseudoElement) => {
		// custom implementation
	},
});

Patch Changes

0.4.7

Patch Changes

  • d6c4455 #352 Thanks @eps1lon! - Support native labels in IE 11

    Also affects Edge < 18 and Firefox < 56.

0.4.6

Patch Changes

  • f7c1981 #288 Thanks @eps1lon! - Drop node 13 support

    We only stopped testing. Probability of breakage should be very low.

    New policy:

    Only active node versions are supported. Inactive node versions can stop working in a SemVer MINOR release.

  • fa53c51 #210 Thanks @eps1lon! - Implement accessbile description computation

    import { computeAccessibleDescription } from "dom-accessibility-api";

const description = computeAccessibleDescription(element);

    Warning: It always considers title attributes if the description is empty. Even if the title attribute was already used for the accessible name. This is fails a web-platform-test. The other failing test is due to aria-label being ignored for the description which is correct by spec. It's likely an issue with wpt. The other tests are passing (13/15).

0.4.5

Patch Changes

  • d668f72 #273 Thanks @eps1lon! - fix: Concatenate text nodes without space

    Fixes <h1>Hello {name}!</h1> in react computing "Hello name !" instead of Hello name!.

0.4.4

Patch Changes

  • 969da7d #240 Thanks @eps1lon! - Reduce over-transpilation

    Switched from

    • for-of to .forEach or a basic for loop
    • array.push(...otherArray) to push.apply(array, otherArray)

    This removed a bunch of babel junk that wasn't needed.

0.4.3

Patch Changes

  • b421d9e #168 Thanks @eps1lon! - fix: Use relative paths in exports field

    Fixes a crash when using ES modules in Node.

0.4.2

Minor Changes

  • 0897630 #155 - Publish version using ES6 modules allongside current CommonJS modules

0.4.1

Patch Changes

0.4.0

Minor Changes

Patch Changes

0.3.0

Minor Changes

  • 7f1ada0: Internal polish

0.2.0

Minor Changes

  • eb86842: Add option to mock window.getComputedStyle

    This option has two use cases in mind:

    1. fake the style and assume everything is visible. This increases performance (window.getComputedStyle) is expensive) by not distinguishing between various levels of visual impairments. If one can't see the name with a screen reader then neither will a sighted user
    2. Wrap a cache provider around window.getComputedStyle. We don't implement any because the returned CSSStyleDeclaration is only live in a browser. jsdom does not implement live declarations.

Bug Fixes

Features

  • name: Consider prohibited naming (#19) (6692d6b)
  • Consider all cases of "name from content" (#13) (835cb76)
  • Consider content from before and after pseudo elements (#5) (0987426)
  • Fork elementToRole from aria-query (#7) (fe4fab5)