[Rule RSRCH]: Request for new rule: table_caption_exists

[mbusch3]

Rule description

The current ruleset has a few existing checks to make sure that every <table> has an appropriate <caption> describing its content:

• table_caption_nested makes sure that any <caption> tags are nested properly inside the <table> tags.
• table_caption_empty makes sure that when a <caption> tag is present, it contains text.

However, when tables don't contain captions at all, there is no issue reported. Based on the definition of WCAG 1.3.1: Info and Relationships, and specifically Technique H39: Using caption elements to associate data table captions with data tables, every data table really, really should contain a caption.

I suggest we add a rule for table_caption_exists to verify that data tables have their required captions.

Relevant requirement(s)

1.3.1 Info and Relationships

High-level approach and rule scope

• Every <table> element that does not also have a role="presentation" or role="none" attribute should have a <caption> element nested as its first child element OR an aria-labelledby attribute with a valid ID. aria-label alone without an on-screen caption is insufficient to meet WCAG 1.3.1.
• Every element with the role="table" or role="treegrid" attribute should have either an aria-label attribute with text or an aria-labelledby attribute with a valid id (see MDN's ARIA: table role and MDN's ARIA: treegrid role.

Ruleset and failure level

WCAG 2.1

Failure level

Violation Recommendation

Different failure level

No response

Failure cases

Table Elements

1. Does the <table> element also have the role="presentation" or role="none" attribute? Yes: break... not relevant. No: proceed.
2. Does the <table> element have a child <caption> element positioned properly as its first child? Yes: pass. No: proceed.
3. Does the <table> element have an aria-labelledby attribute with a valid id? Yes: pass. No: fail.

Elements with Table Roles

1. Does an element have a role="table" or role="treegrid" attribute? No: break... not relevant. Yes: proceed.
2. Does this element have an aria-label attribute with a non-empty value OR an aria-labelledby attribute with a valid id? Yes: pass. No: fail.

[shunguoy]

@mbusch3 This used to be covered, more generically, in the rule aria_accessiblename_exists for the existence of an accessible name, which is an aria-label, aria-labelledby, or a caption for a table. However, Aria 1.3 has updated to remove the requirements of accessible names for the roles like table and treegrid to be better in line with WCAG 2.2, and the equal access rules were updated accordingly.

[tombrunet]

@mbusch3 WCAG 1.3.1 requires that a caption is programmatically determinable if a caption exists. H39 is how you can programmatically specify that caption. 1.3.1 does not say that all tables should have captions.

1.3.1 says that if you use provide a caption that you must use standard mechanisms to indicate that it's a caption (i.e., don't just throw plain text above the table that looks like a caption without the appropriate markup to indicate that it's a caption). The intent here is to ensure that things that are shown visually are also determinable programmatically.

[mbusch3]

@shunguoy, @tombrunet Thank you so much for the quick responses!

I checked the link that @shunguoy sent (Aria 1.3), and after reading the document, it looks like an author-provided accessible name IS required for elements with both table and treegrid roles. Am I reading it wrong?

[shunguoy]

@mbusch3 Please review the PR for this. It was supposed to be merged, but somehow not yet.

[tombrunet]

Okay, so this is reflected in the latest working draft (but still required on treegrid)
https://w3c.github.io/aria/#grid

[mbusch3]

@shunguoy I've gone over the PR, but I don't know if it needs my review, since it looks like it was merged back in February. I'm not sure how the internal processes work for w3c/aria or how releases get published there. It looks like the entire reason I submitted this issue comes down to the draft of the ARIA spec I was looking at when I was working on vetting the Equal Access scanner reports.

In the Editor's Draft, the "SHOULD" language is present and "Accessible Name Required" is not visible.
In the Working Draft, it still shows "Accessible Name Required" as True.

[shunguoy]

thanks @tombrunet That's the version I tried to find.
@mbusch3 w3c has a early review/editor's draft, working draft, and public version of a spec (ARIA, ARIA in HTML etc). Those versions are not always the same, causing confusions.

[mbusch3]

@shunguoy Thank you for that clarification! I am currently very much in the weeds of testing out the IBM Equal Access scanning tool as a primary component of an accessibility scanning tool. In the future, should I assume that the IBM code will be following the Editor's Draft, the Working Draft, or the public version of ARIA 1.3?

[mbusch3]

@tombrunet, @shunguoy

I feel that the issue still stands. Even with the ARIA spec change from "Accessible Name Required" to "Authors SHOULD provide an accessible name for a table", the check for a <caption> being present should still be done, just at a lower priority.

I've updated the original description's Failure Level from "Violation" to "Recommendation."

[MHoov]

@tombrunet I believe we don't adopt changes until they are published, correct?

[shunguoy]

@mbusch3 We basically treat the specs as 'live' documents, meaning they will be updated as needed without specific versioning (or just internal versioning). The update of IBM accessibility tools is priority-based, and the priority is determined based on a change request (e.g., a user issue, a spec change) and its corresponding requirement from WCAG, Section 506 etc., and the practical implication as well. The addition of a table caption or accessible name is very useful, and of a best practice, but not a requirement.

[mbusch3]

@shunguoy That's fine! If adding a caption is a best practice, recommended, and a strong SHOULD in even the yet-to-be-released spec, can we please add the test, just with a "RECOMMENDATION" level instead of a "VIOLATION" or "POTENTIAL VIOLATION"?