>Parse the entry string into words (aka tokens) and assemble with the and
>operator. E.g. 'red cat' becomes 'red & cat'. >Then add vector; more info
>in articles I provide links to later in this note.
>WHERE to_tsvector ( productname || ' ' || productdescription ) @@
>to_tsquery ( 'red & cat' )
Since there were no responces for a while, I went with another solution.
Splitted search string to words like you but converted query to
select
+case when productname ilike '%red%' then 2 else 0 end
+case when productdescription ilike '%red%' then 1 else 0 end
+case when productname ilike '%cat%' then 1.7 else 0 end
+case when productdescription ilike '%cat%' then 0.7 else 0 end
from products
order by 1 desc
limit 100
This allows to define relevance.
Is my solution reasonable ?
Andrus
