Unexpected Mathjax error in SERP

GaurangTandon commented 6 years ago

Entry number 8 has an error in two mathjax displays ("Missing close brace" and "Extra close brace or missing open brace" respectively).
this is a screenshot
the original post on MSE is free from any mathjax error though
A similar problem exists for entry 7 as well
I could not come up with any possible reason for this

Broken Mathjax of entry 8 copy-pasted for reference:

...Put it this way \int {\frac{x}{{\sqrt { ... t {\frac{{2ax + b}}{{\sqrt {a{x^2} + bx + c} }}dx}  - \frac{b}{{2a}}\int {\frac{{dx}}{{\sqrt {a{x^2} + bx + c} }}} 
\displaystyle \frac{c}{a} - \frac{{{b^2}}}{{4{a^2}}} < 0 =  -  ... 2}.

w32zhong commented 6 years ago

@GaurangTandon Thank you for reporting, will investigate later when I get some time.

w32zhong commented 6 years ago

@GaurangTandon Hi, the reason is actually quite simple, since the search result snippet is trying to summarize a document in a short paragraph, it has to skip some content and show as many highlighted words as possible. This leads to the problem you have seen: In all the cases that you find this problem, the content that is skipped (those will be replaced by a ... string) is in the middle of a LaTeX expression, and that is very likely to invalidate a LaTeX expression.

The current content skipping strategy is simple: Given a number of keywords in the document (within a threshold limit MAX_HIGHLIGHT_OCCURS), pad the left and right side of each keywords, those content that are not padded will be skipped, the keywords along with their "padding" will be displayed. The related logic is here: https://github.com/approach0/search-engine/blob/4780e499519677433543cb92ba8baa04b56f959a/search/snippet.c#L124-L125

One way to fix this issue is not skipping any LaTeX content, but some LaTeX content are very long and this strategy will make some snippet unacceptable long. So a more smart algorithm is needed to either include complete LaTeX clip or do not include any part of that clip if it is too long.

We can leave this issue open before a better skipping strategy algorithm is implemented.

approach0 / search-engine

Unexpected Mathjax error in SERP #22