Libdoc: argument, return and raises documentation as tables - #5742

• src/web/package-lock.json: Generated file

• src/web/package-lock.json: Generated file

src/web/libdoc/view.ts:642

• The tag click handler reads the tag text from e.target. When the tag text is highlighted (mark.js wraps matches in ), clicks can target the element and only pass the substring instead of the full tag, causing incorrect tag filtering. Use e.currentTarget (the .tag-link element) instead.

    document.querySelectorAll(".tag-link").forEach((elem) => {
      elem.addEventListener("click", (e) => {
        this.tagSearch((e.target! as HTMLSpanElement).innerText);
      });

src/web/libdoc/view.ts:547

• The clamped doc click handler runs for clicks on links inside the documentation. That can cause conflicting behavior (e.g., clicking a type link can both open the type modal and also open the arg-doc modal, resulting in multiple modal contents being appended). Skip opening the arg-doc modal when the click originates from an element.

This issue also appears on line 639 of the same file.

      const showDetails = (event: Event) => {
        event.stopPropagation();
        this.showArgDocModal(wrap);
      };
      doc.onclick = showDetails;

• src/web/package-lock.json: Generated file

src/web/libdoc/view.ts:218

• When switching the UI language, render() is called again (via initLanguageMenu), but the modal is not recreated if #modal-background already exists. That means the close button’s aria-label stays in the original language, which is an accessibility regression for screen readers after language changes. Update the existing close button label on every render (or recreate the modal) so it always matches the current translation.

    if (!document.getElementById("modal-background")) {
      createModal(this.translations.translate("closeDialog"));
    }

• src/web/package-lock.json: Generated file

src/web/libdoc/view.ts:642

• The tag click handler reads the tag name from e.target.innerText, which can be only the highlighted substring when mark.js wraps part of the tag (e.g. searching "foo" highlights foobar). Clicking then filters by a partial tag and produces the wrong results. Read from e.currentTarget (the .tag-link element) and trim it instead.

    document.querySelectorAll(".tag-link").forEach((elem) => {
      elem.addEventListener("click", (e) => {
        this.tagSearch((e.target! as HTMLSpanElement).innerText);
      });

src/web/libdoc/libdoc.html:367

• Tags are interactive (they register a click handler in view.ts) but are rendered as <span> elements, which are not focusable or keyboard-activatable by default. Using a <button type="button"> here provides correct semantics and keyboard support (Enter/Space) without extra JS.

                  {{#each tags}}<span
                      class="tag-link"
                      title="{{t 'showTagKeywords'}}"
                    >{{this}}</span>{{/each}}

src/web/libdoc/modal.ts:32

• The close action is wired both on the button and on its container, so clicking the close button triggers hideModal() twice (including scheduling duplicate timeouts to clear the content). Remove one of the listeners to avoid redundant work and potential double side effects if hideModal changes later.

  modalCloseButtonContainer.addEventListener("click", () => {
    hideModal();
  });

src/web/libdoc/view.ts:524

• updateDocClamping() skips docs whose clientHeight is 0 ("not laid out yet"), leaving them permanently clamped and non-clickable. With content-visibility: auto on .keyword-container, most offscreen keyword docs start out unlaid-out and may never get a more... button when they later enter the viewport.

    // Hidden documentation -- filtered out by the search, not laid out yet --
    // measures zero and would look like it fits. It stays clamped, which is
    // how the template renders it anyway.
    const measured = docs.map((doc) => ({
      known: doc.clientHeight > 0,
      overflowing: doc.scrollHeight > doc.clientHeight + 1,
    }));