Cannot reproduce Odd character prohibits search

[Wildcat Media]

We had an issue where we could not search for a thread due to a member using an odd character in the thread title:

•

I believe in HTML that may be the · character, or something similar (I haven't coded HTML in awhile now to remember). Whatever it is, it is outside the standard range of ASCII characters. With the dot, we could not find the thread with a thread title search. We edited the thread title to replace the character with a hyphen, and the thread now appears in search.

I guess on a Mac, the Command-8 combination will insert that character. In Windows ALT+0149 on the keypad will insert it. This could be more a bug in ElasticSearch than in XF, though (and regular readers know how I feel about ES ).

 

[Mike]

No problem with this here: http://xenforo.com/community/search/3292507/?q=uniqueword&o=date&c[node]=6

It may be something down to tokenization, so the exact title would be important. If it is that, you can change the tokenizer in your ES install if you wish.

 

[Wildcat Media]

Thank you for checking this out, Mike. Unfortunately someone on the staff renamed the thread, but I may press them to see if they can remember what it was exactly, so I can attempt to troubleshoot it on my end. If it is the tokenizer as you say, that will give me a chance to adjust the settings to make it work.

 

[Wildcat Media]

No problem with this here: http://xenforo.com/community/search/3292507/?q=uniqueword&o=date&c[node]=6

It may be something down to tokenization, so the exact title would be important. If it is that, you can change the tokenizer in your ES install if you wish.

I found the thread, finally, and renamed it again to include the • character, and now it works.

Last night, though, I upgraded to ES to the most recent version, as opposed to 0.19.16 which is what we were using previously.

Another thought: ES was crashing a lot, and it is possible the member caught us during one of the times before we had regenerated the search indexes...the whole thing could have been a coincidence.