Re: Searching for %SYS

Hi Thomas:

In message <[email protected]>,
Thomas Arendsen Hein writes:
>* John P. Rouillard <[email protected]> [20190827 03:52]:
>> In message
>> <DM6PR08MB477859405D66CDA22C6174A9CAA10@DM6PR08MB4778.namprd08.prod.
>> outlook.com>, Tom Ekberg writes:
>> >[...] What is the preferred way to search for issues or messages containing
>> >something like %SYS?
>> 
>> At first guess, I suspect the full text search index strips
>> punctuation. I see reference to the pattern \b\w{%d,%d}\b in add_text
>> in a few places.  IIRC \w matches alphanumerics and _. So I read the
>> regexp as matching just after the % in %SYS. So %SYS isn't in the
>> index but SYS is.
>> 
>> I think we could change that regexp to: \b[\w%s]{%d,%d}\b and make %s
>> substitute a list of characters specified in config.ini. Maybe call
>> the setting indexer_wordchars?
>
>This may confuse the indexer, not sure.

I am sure any stemmer would have issues, are you concerned about
non-alphanumerics being represented in the indexer? I am not sure how
they are set up but if they can handle foreign locales I suspect the %
etc won't be a problem. 

>What about adding the possibility to do a (slow!) full text search?

Yeah, that's possible. However slow doesn't begin to describe it if
you have thousands of messages/files etc.

>> In an emergency "grep -lr '%SYS' db/" then looking at the
>> messages/files to see what issues it's associated with could work. But
>> this search does ignore the title string...
>
>Title search on the search page already bypasses the indexer, so
>searching for %SYS in the title of an issue already works.

I didn't know that. Does that got straight to the db with a substring
search?

Have a great day.

--
				-- rouilj
John Rouillard
===========================================================================
My employers don't acknowledge my existence much less my opinions.