RFR: JDK-8276800 Fix table headers in NumericShaper.html

Alexey Ivanov aivanov at openjdk.java.net
Tue Nov 9 12:06:33 UTC 2021


On Tue, 9 Nov 2021 11:42:26 GMT, Alexey Ivanov <aivanov at openjdk.org> wrote:

>> This change introduces no visual difference, but improves a11y.
>
>> /integrate
> 
> I believe this is preliminary as our discussion isn't over yet.

> @aivanov-jdk Gladly. [Doccheck](https://openjdk.java.net/projects/code-tools/doccheck/) flagged it, my understanding is that every row needs to have a unique row header to aid screen readers. The previous code had one cell work as header for two rows.
> 
> One could perhaps look at [the standards](https://html.spec.whatwg.org/multipage/tables.html#attr-tdth-headers) and implement something even better, but this is what I came up with.

This information should be added both to the JBS bug description and to this PR description, otherwise this comes out of nowhere.

I think this approach is wrong. However, to satisfy this requirement of Doccheck dummy column with numbers was added to many tables which _logically_ have no row headers.

You change the semantics of the table by moving the header to another column. The rendering should rather change too, otherwise visual representation becomes misleading.

To me, the table looks correct currently: *Arabic* Unicode range serves as the title for both rows. I wonder how it's read by a screen reader: the original version and the updated version. What we should strive for is for clarity of the read table rather than blindly following the rule where each row should have a title, provided that the title is available in this particular case.

@jonathan-gibbons, Can this error be marked as false positive in the Doccheck report? Can the tool be smarter?

-------------

PR: https://git.openjdk.java.net/jdk/pull/6291



More information about the client-libs-dev mailing list