Search highlighting is broken

[cmb]

Hello Community,

I just found out that the search highlighting is broken, when one searches for multiple words, where one of the words is contained in another.

Consider a page with the word "word". Searching for "word or" will list this page; after clicking on this search result, the page will be shown, with "word" being highlighted. Fine -- well, look at the generated HTML source code:

Code: Select all

<span class="xh_find">w<span class="xh_find">or</span>d</span>

Um, do we really want nested classes? Not a problem per se, but looks a bit clumsy.

Anyway, let's search for "or word":

Code: Select all

w<span class="xh_find">or</span>d

Only "or" is highlighted. Bug.

Now consider a page with the text "उद्धरेदात्मनात्मानं नात्मानमवसादयेत्" (that's Sanskrit, BTW). We search for "रेदात्मनात्मानं नात्मान":

Code: Select all

उद्ध<span class="xh_find">रेदात्म<span class="xh_find">नात्मान</span>ं</span> <span class="xh_find">नात्मान</span>मवसादयेत्

Bug, because the diacritic mark " ं" is separated from "न", which should be "नं", actually.

I'm not sure if it is possible to solve the combining character problem efficiently, and I tend to classify that as "won't fix". Solving the problem that only a partial word is highlighted (second case) should be easy by simply ordering the search words by length.

Thoughts?

Christoph

[svasti]

An interesting problem. Seems to me not so terrible and could be fixed in 1.6.4?
Very cute, using the Sanskrit example, and that you realized that there was a doubling of glyphs.

Solving the problem that only a partial word is highlighted (second case) should be easy by simply ordering the search words by length.

+1

BTW the Sanskrit says: "Lift yourself up by yourself, don't dispirit yourself." (The word "yourself" makes for the doubling of the glyphs)

[cmb]

I've fixed the "or word" problem, refactored out XH_highlightSearchWords() and added unit-tests for it which serve as documentation of the mentioned issues (r1448).