Re: Trim performance on 9.5

Vincent Elschot <vinny@xxxxxxxxx> · Sun, 20 Nov 2016 09:12:47 +0100



    Op 18/11/2016 om 16:58 schreef William
      Ivanski:

    
      I just ran EXPLAIN ANALYZE, please see images
        attached. Field doesn't have a index.

        
          Em sex, 18 de nov de 2016 às 12:16, vinny <vinny@xxxxxxxxx>
            escreveu:

          
          On
            2016-11-18 15:06, William Ivanski wrote:

            > Hi,

            >

            > I recently did major improvements on perfomance on our
            routines by

            > simply removing the call for trim functions on specific
            bottlenecks.

            > Please see images attached for a simple example.

            >

            > I'm using PostgreSQL version 9.5.5-1.pgdg80+1 on Debian
            8.6. Someone

            > knows if it's a bug on trim function? Thanks in
            advance.

            >

            > --

            >

            > William Ivanski

            
            Did you run EXPLAIN on these queries?

            
            I'm guessing that you have an index on the field, but not on

            TRIM(field),

            which would mean that the database is forced to seqscan to
            fetch every

            row value, trim it and then compare it.

          
      -- 

      
        William Ivanski
      
    
    Neither exeution times are really "fast", I'd suggest creating an
    index on the TRIM() version of the field.