Search matches characters across new lines

[xavier114fch]

Just observing this with the default tracemonkey paper.

1. Use the Find tool and enter "na" in the search box. Check the "Highlight all" box.
2. Scroll down to page 2 of the paper, and see the 3rd paragraph. It highlights the "na" in "In a", which is across a new line.
3. Change the search query to "andt" and you can see it highlights "andt" in "and the".

Seems it needs some sort of checks to ensure that search will not return these results?

[Quicksaver]

That is not the only case where the search returns wrong matches or fails to find the correct matches.

The PDFFindController.pageContents object, which PDFJS's find method uses to return search matches, often doesn't reflect exactly what a page's text content actually is.

In my add-on, FindBar Tweak, I use the same object in the same way to retrieve find results from pdf documents. Through its Find in All Tabs feature, the limitations of this object are very apparent:

• In every pdf page there are no newline characters, ever!
• Often, space characters are omitted in this object, greatly influencing search results. http://www.selab.isti.cnr.it/ws-mate/example.pdf is especially problematic, there are absolutely no spaces in this specific document!

At least the space characters should more closely reflect the actual document; newline characters aren't as critical since they are never in a search query anyway, a single space character in place of newlines would be a close enough solution in my opinion.

I would love to try and help fix this myself, but I looked into the code that builds this object and I don't think I have enough knowledge of PDF.JS's rendering mechanism to be able to help with this... However, I hope my add-on can somehow help with debugging this, since it can show the contents of this object in an easy and direct way without having to manually search all over the pdf pages to know if the object's text contents are accurate.

[piotrex]

Changing block <div> to inline <span> in text layer as per #2989 will probably improve this. (At least it should cause that copied text across lines will contain spaces)

[jazzy-em]

+1
Problem is actual.

Search results are incorrect due to this problem.
Would be considered this issue?

[jazzy-em]

A similar problem is solved in #5783.

[murilo-ramos]

I know this is an old thread, but anyone knows if this is issue was solved?
I'm having the same situation where the search finds 'words' across new lines, and I tested with versions 1.9.426 (stable) and 2.0.550 (press release) and the issue persists.

[calixteman]

It's partially fixed thanks to af4dc55 which helps to detect EOL.
And this PR: #13261 should definitely fix the problem.