Forward search does not scroll the page

[LasseBlaauwbroek]

When doing a forward search from a TeX file, pdf-tools correctly finds the location of the search and jumps to the correct page. However, if the location on the page is below the viewport, the page itself is not scrolled. If you manually scroll down, the arrow does point to the correct location though. Strangely, if the location is above the viewport the scrolling does happen. Testcase:

\documentclass{article}
\usepackage{lipsum}
\begin{document}
This is test number one.

\lipsum[3-7]

This is test number two.

\lipsum[7]

This is test number three.
\end{document}

Try doing a forward search on all three test sentences. The first and last always work. The second only if you manually scroll it into view.

Presumably the error is here:

pdf-tools/lisp/pdf-sync.el

Lines 661 to 677 in d971298

	(defun pdf-sync-forward-search (&optional line column)
	"Display the PDF location corresponding to LINE, COLUMN."
	(interactive)
	(cl-destructuring-bind (pdf page _x1 y1 _x2 _y2)
	(pdf-sync-forward-correlate line column)
	(let ((buffer (or (find-buffer-visiting pdf)
	(find-file-noselect pdf))))
	(with-selected-window (display-buffer
	buffer pdf-sync-forward-display-action)
	(pdf-util-assert-pdf-window)
	(when page
	(pdf-view-goto-page page)
	(when y1
	(let ((top (* y1 (cdr (pdf-view-image-size)))))
	(pdf-util-tooltip-arrow (round top))))))
	(with-current-buffer buffer
	(run-hooks 'pdf-sync-forward-hook)))))

I do indeed not see any explicit scrolling.

On a slightly unrelated note: It would be great if instead of an arrow, a rectangle could be shown around the exact searched area. Especially with a two-column layout it can be difficult to see what the arrow points to.

[antoine-levitt]

I think this is the same as #648

[ghost]

I played around a bit with this today. We just aren't scrolling to the location in the pdf (just going to the page and hoping it's visible. I think I have a fix. I'll create a PR and get it finalized, but you can try this if you want:
(Edit: this fix is 'wrong' in a way. I'm seeing now that pdf-util-tooltip-arrow is supposed to account for the image being in view.)

(defun pdf-sync-forward-search (&optional line column)
  "Display the PDF location corresponding to LINE, COLUMN."
  (interactive)
  (cl-destructuring-bind (pdf page _x1 y1 _x2 _y2)
      (pdf-sync-forward-correlate line column)
    (let ((buffer (or (find-buffer-visiting pdf)
                      (find-file-noselect pdf))))
      (with-selected-window (display-buffer
                             buffer pdf-sync-forward-display-action)
        (pdf-util-assert-pdf-window)
        (when page
          (pdf-view-goto-page page)
          (when y1
            (let ((top (* y1 (cdr (pdf-view-image-size)))))
              (pdf-view-show-vertical-location top)
              (pdf-util-tooltip-arrow (round top))))))
      (with-current-buffer buffer
        (run-hooks 'pdf-sync-forward-hook)))))

(defun pdf-view-show-page-location (page-location)
  "Make sure PAGE-LOCATION is visible in window.

If the PAGE-LOCATION is not clearly visible, then we scroll
the PAGE-LOCATION to 1/3 the current view window."
  (let* ((body-height (window-body-height nil t))
         (visible-top (window-vscroll nil t))
         (visible-bottom (+ visible-top body-height))
         (padding (frame-char-height)))

    ;; If page location is out or 'close' to being out of
    ;; the viewport we scroll the spot to the 1/3 spot
    ;; of the viewport
    (when (or (< page-location (+ visible-top padding))
              (> page-location (- visible-bottom padding)))
      (image-set-window-vscroll
       (max 0
            (round (- page-location
                      (/ body-height 3))))))))

[vedang]

Hi @LasseBlaauwbroek ,

I have merged @tylerware's change into vedang/pdf-tools@2f6be19 , and it fixes this issue.

Re: showing a rectangle around the searched text, I'll track that in a separate todo for myself. Thanks!

[LasseBlaauwbroek]

Thats great! Thanks for your efforts.

[vedang]

@LasseBlaauwbroek: Can you please close this issue? This helps me better track the open issues. Thank you!

[LasseBlaauwbroek]

@vedang It seems a bit strange to track the issues on your fork using this repository. This issue has been fixed there, but not here. If this repo is ever resurrected, the state of the issues would be completely inconsistent.

May I suggest that you import the issues of this repo to your own? A quick google suggests this script may be able to help: https://github.com/IQAndreas/github-issues-import

[vedang]

Okay @LasseBlaauwbroek!

My idea was that if @politza revives this repo, I'll submit a PR to him from my master branch, with all the patches I've applied. I understand your point of view as well :)