RFR: JDK-8274625: Search field placeholder behavior [v2]

[Pavel Rappo]

On Tue, 5 Oct 2021 17:20:25 GMT, Hannes Wallnöfer <hannesw at openjdk.org> wrote:

>> Please review a change to use the HTML `placeholder` attribute in the javadoc search input instead of an actual input value. This revives the fix I originally created for  JDK-8221366 but forwent back then because of insufficient support by the Microsoft IE and Edge browsers. MS Edge has since been updated and supports the placeholder attribute well now. 
>> 
>> Example pages generated with this change can be viewed here (top level pages only):
>> http://cr.openjdk.java.net/~hannesw/8274625/api.00/
>> 
>> I changed the color of the placeholder to a lighter grey than we used for our own solution because the placeholder attribute works in a slightly different way. While we removed the placeholder text when the user clicked on the input field, the standard placeholder is still visible right until the user starts entering text. Because of this, using a darker color for the placeholder is confusing because it can easily be mistaken for actual input. The lighter grey is also closer to the default value for most browsers and web frameworks such as bootstrap.
>> 
>> I also changed the HTML for the search input to not have an initial value of "search" which had then to be cleared by the search script. Furthermore, the `disabled` attributes now come in HTML5 value-less boolean format, i.e. `disabled` instead of `disabled="disabled"`. 
>> 
>> In addition to most significant desktop browsers I tested the generated docs on various mobile browsers on Android and iOS devices.
>
> Hannes Wallnöfer has updated the pull request with a new target base due to a merge or a rebase. The pull request now contains three commits:
> 
>  - Merge branch 'master' into JDK-8274625
>  - JDK-8274625: Do not set initial value attribute on search input
>  - JDK-8274625: Search field placeholder behavior

We should use standard tools for standard tasks where possible. In this case using a "placeholder" instead of the custom "watermark" workaround fixes the problem, simplifies the code and looks better.
 
 Update the copyright years before integrating.

src/jdk.javadoc/share/classes/jdk/javadoc/internal/doclets/formats/html/Navigation.java line 604:

> 602:                 .put(HtmlAttr.PLACEHOLDER, searchPlaceholder);
> 603:         HtmlTree inputReset = HtmlTree.INPUT(reset, HtmlIds.RESET_BUTTON)
> 604:                 .put(HtmlAttr.VALUE, reset);

Isn't it a mere coincidence that "reset" is both an input type and the value of the "value" attribute: 

<input type="reset" id="reset-button" disabled value="reset">

Why do we need this "value" attribute?

src/jdk.javadoc/share/classes/jdk/javadoc/internal/doclets/formats/html/markup/HtmlTree.java line 534:

> 532:                 .put(HtmlAttr.TYPE, type)
> 533:                 .setId(id)
> 534:                 .put(HtmlAttr.DISABLED, "");

I'm curious: if such a call to `put` translates to a boolean HTML attribute, what call to `put` translates to an HTML attribute with an empty value?

src/jdk.javadoc/share/classes/jdk/javadoc/internal/doclets/toolkit/resources/doclets.properties line 226:

> 224: doclet.Implementation=Implementation(s):
> 225: doclet.search=SEARCH:
> 226: doclet.search_placeholder=Search

Later we should consider a better placeholder, such as an example input. We already have the adjacent "SEARCH" text label and the magnifying glass icon. Something tells me that the user has a pretty good idea what this field is for and that they don't need yet another "Search".

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

Marked as reviewed by prappo (Reviewer).

PR: https://git.openjdk.java.net/jdk/pull/5825