Now that the new GIN/tsearch2 based search engine has been running on search.postrgresql.org it's time for a bit of statistics of the searches being performed. The following statistics are pulled from the most recent 99,733 searches performed. Note that these statistics are taken pre-lexing, so searches for different variants of the same word count as different words.
A total of 32,810 unique search terms were submitted, meaning that each search term was used on an average only 3 times.
4930 of the searches resulted in stepping pas the first page of results, meaning that 95%25 of all searchers only looked at the first page of results.
17 searches stepped all the way to the 50th page (the maximum number of pages shown). My guess is that all of these were people testing and not actually looking for the hits there.
The 25 most common search terms, and the number of searches for them, were:
foobar 3873
search 1796
alter table 732
create table 588
jdbc 517
select 508
update 507
insert 475
copy 463
pg_dump 458
odbc 386
sequence 381
vacuum 368
date 361
cast 350
case 349
replication 330
grant 320
psql 318
join 296
create user 275
between 266
tsearch2 264
timestamp 260
cluster 250
The first two are certainly very interesting, with foobar representing almost 4%25 of the searches. You may wonder why this is, and it has a very logical explanation - that's the nagios system that monitors the servers functionality...
The second one is explained by that being what you get if you hit "search" on the frontpage without typing anything in the search box first.
The good news is that system load has remained vastly below what it was with the old search engine. Most of the time, you can't see anything on the server at all, and the load avg is 0.02/0.01/0.00 right now (my stats processing being what brought it so far up).
New comments can no longer be posted on this entry.