According to Olivier Korn: > At 16:00 02/10/2002 -0500, Randall Fish wrote : > >Our users are searching for 'breast' and 'breasts' and no search results are > >being returned. A search for 'mammogram' returns documents containing > >'breast' and 'breasts'. The db.wordlist file contains the references to both > >words. The words do not exist in the bad words file. > > It looks very similar with the problem I had with ht://Dig version 3.1.5 > because my locale-settings was influencing the way the sort program was > working. > > The solution was to add LC_COLLATE=C before the call (as it is done in the > version 3.1.6 of the rundig script). I know you are using this version of > ht://Dig, but maybe Cygwin is somehow different in its way of managing > locales... Couldn't it be ?
This was my gut reaction yesterday too, and I was about to e-mail asking whether db.wordlist appeared to be sorted correctly under Cygwin. But then I remembered that I had fixed htmerge in 3.1.6 to build the word database correctly even if db.worlist entries aren't in the right order. It's just less efficient at it, and results in a bigger word database, than if the words are sorted correctly, but the db shouldn't be losing entries. Yet, lost entries in db.words.db seems to be the most reasonable explanation for this problem. About all I can think to suggest right now is to try compiling and running db/db_dump/db_dump.c to look at the entries in db.words.db and see if anything is missing. If entries are missing, then htmerge isn't building this db correctly under Cygwin. If the db is complete, then htsearch doesn't seem to be finding all the entries under Cygwin, or is rejecting them for some unknown reason. Oh, have you tried reindexing from scratch to see if that makes the problem go away? It may be that the database has become corrupted somehow. (Although htmerge rebuilds db.words.db from scratch each time you run it, it may be that htsearch is dropping results because of a corrupted db.docdb.) Does the number of matches reported by htsearch agree with how many results are actually shown? -- Gilles R. Detillieux E-mail: <[EMAIL PROTECTED]> Spinal Cord Research Centre WWW: http://www.scrc.umanitoba.ca/ Dept. Physiology, U. of Manitoba Winnipeg, MB R3E 3J7 (Canada) ------------------------------------------------------- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf _______________________________________________ htdig-general mailing list <[EMAIL PROTECTED]> To unsubscribe, send a message to <[EMAIL PROTECTED]> with a subject of unsubscribe FAQ: http://htdig.sourceforge.net/FAQ.html
