Not digging the 3.2 search feature

[cthree]

I'm not sure if this is a beta thing but I find the 3.2 search (on this site in particular) to be very unfriendly. Here are a couple of examples of what I mean:

"logging in"

Says invalid query because on of the words is < 3 chars. I don't think the user should have to know that, can the software not simply ignore or remove short words? Not allowing the use of short words, even if they aren't used as search keywords, makes it impossible to form natural questions:

What is a ...
How do I ...
Where is ...
How do ...

Not good IMHO.

"IE javascript error"

Says javascript is invalid term AND that there is a short word. How do you search for something related to javascript without saying "javascript"? Thank god I don't run a web design forum. There needs to be a better way to prevent xss issues that don't involve omitting "javascript" from the range of possible keywords.

I hope this feedback is helpful.

[Matt]

Actually, I want to revisit those unsearchable words.

Javascript isn't added to prevent XSS. It's a remanent from when we used to store posts fully parsed. Some BBCode would generate inline javascript, so any search for javascript would return those 'false' matches.

Similarly, we do not allow img, quote, color, span, div, style, etc to be searched for similar reasons. They would return HTML tags and not true matched text.

Now (and since 3.1) we store posts as BBCode and parse then cache on demand. So we can certainly remove 'javascript'. However, we still wouldn't be able to remove the BBCode tag names because it would still return 'false' matches (for example, if you used a color tag and searched for color). There is no easy way to tell MySQL's full text engine or Sphinx to ignore matches wrapped in square brackets.

[.Ian]

On the other hand there are times when one wishes to search for code examples that are in code boxes - so the ability to search say for $data would be useful.