Page MenuHomePhabricator

Searching for strings without letters or numbers returns error without description ("An error has occurred while searching: The search backend returned an error:")
Closed, DeclinedPublic

Details

Reference
bz47770

Event Timeline

bzimport raised the priority of this task from to High.Nov 22 2014, 1:15 AM
bzimport set Reference to bz47770.
bzimport added a subscriber: Unknown Object (MLST).

Rephrasing summary and raising priority according to observations on https://de.wikipedia.org/wiki/Wikipedia:Fragen_zur_Wikipedia#Hat_die_.28Volltext-.29Suche_ein_Unicode-Problem.3F:

It seems like all characters that are neither letters nor numbers (according to Unicode General Category) are removed from the search string. This includes private characters as well as surrogates (which removes all characters outside BMP, even if they are letters). If the resulting string is empty, the said error is thrown. Some examples:

https://en.wikipedia.org/w/index.php?title=Special%3ASearch&profile=default&search=a%EE%80%80&fulltext=Search (search from first comment, with an "a" prepended, is treated like a search for "a" alone)

https://en.wikipedia.org/w/index.php?title=Special%3ASearch&profile=default&search=%24&fulltext=Search (search for "$", throws error)

https://en.wikipedia.org/w/index.php?title=Special%3ASearch&profile=default&search=!%22%C2%A7%24%25%26%2F%28%29%3D%3F%3C%3E*%2B%27%23%2C.%3B%3A-_&fulltext=Search (search for a string of punctuation chars, throws error)

Probably namespaces are removed from the search, too, which causes bug 49675 with the same error thrown.

Thanks for investigating. This might also explain several other bug reports. :-/

I opened bug 51661 for the characters outside the BMP. If the user enters a search that doesn't contain any indexed character, one can't expect a sensible result. But of course there should be a descriptive error message, like "Sorry, there are no results because no part of your search is in the search index. Please add some letters or numbers to your search to get a result."

  • Bug 51790 has been marked as a duplicate of this bug. ***
  • Bug 51887 has been marked as a duplicate of this bug. ***

Marking WONTFIX as MWSearch has reached its end of life and won't be improved further.

We have a similar issue in CirrusSearch which is being tracked at bug 53013.