phoenix-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From James Taylor <jamestay...@apache.org>
Subject Re: Phoenix "LIKE 'query%' " performance
Date Sun, 18 Sep 2016 15:19:55 GMT
On Sunday, September 18, 2016, Cheyenne Forbes <
cheyenne.osanu.forbes@gmail.com> wrote:

>
>    - Can it be fast?
>    - does it use the Hbase regex feature?
>
> It's a similar implementation, but with some optimizations around the
prefix patterns mentioned below.

>
>    - how can I make it case insensitive? so when I do "LIKE 'query%' "
>    the results include "Query"
>
>  Use ILIKE for case insensitive search. Until PHOENIX-1291 gets
implemented, prefix ILIKE patterns as described below will be full table
scans (i.e. slow).

>
>    - Can I get millisecond results using "WHERE column LIKE" on a large
>    table? couple terabytes of data
>
> If you have have a prefix LIKE pattern (i.e. pkCol LIKE 'XAVIER%') where
the data is significantly filtered based based on leading with the prefix,
then it can be reasonably fast (but still not like a point lookup).

If you have a few particular LIKE patterns you need to be very fast, you
could add functional indexes for them:
http://phoenix.apache.org/secondary_indexing.html#Functional_Index

Otherwise, without other filters in your WHERE clause, your query will be a
full table scan.

>
>    - is it recommended
>
>

Mime
View raw message