The index is a B-Tree, not hashed. The order of the segments of the key
makes a big difference to queries, as pointed out earlier. It doesn't
make any significant difference to the time it takes to create of
maintain the index.
JS
Taka wrote:
Ah, maybe I wasn't quite clear enough.
What I meant was, is there any performance difference between:
CREATE INDEX ON my_table ( a , b , c )
and
CREATE INDEX ON my_table ( c , b , a )
I'm guessing not, presumably because the index is using some kind of hashing
but I thought I'd ask, just to be sure :-)
Taka <sqlite-T/[EMAIL PROTECTED]> wrote:
Does it make a difference what order the columns are specified when
creating an index?
Yes. An index on colums (a, b, c, d) can also speed up search on column
a, on a pair of columns (a, b) and on a triple (a, b, c). If you have a
query with a where clause looking like "where a=1 and b=2" the matching
records will be found using the index. But if you have "where a=1 and
c=2" then all records with a=1 will be found using the index, then a
sequential scan through those records will be needed to find records
matching c=2
Igor Tandetnik
