Re: Indexed PF vs. LF performance

[Vernon Hamberg <vhamberg@xxxxxxxxxxxxxxx>]

Hi Tim

I can't imagine it being much different - an index in a PF is in a 
separate space object on the system from the data itself - and an LF is 
just a different space object. Both will have entries in a tree that 
point to RRNs in the PF.

Sometimes we don't remember that PFs are compound objects - they are 
made up of several different kinds of low-level object types - *MBR, 
*FMT, etc. You can do a DMPOBJ of a PF and get and idea of all these 
things. BTW, DMPOBJ does not show you the actual raw data as stored - 
you have to dig into SST to do that.

I guess one thing is to make a test - I mean, if it's 2 billion 
operations, it might matter - 100? Naw!

HTH
Vern

On 7/30/2012 3:19 PM, Tim Adair wrote:
> A colleague asked me about this and I haven't been able to find any
> definitive information.
>
> For RPG (IV, but not ILE) doing CHAINs, SETLLs, and READs - which is faster,
> an indexed PF or a flat PF with an LF defined with the key?  For the sake of
> this question, let's assume we're talking about a single key field.  The
> file is DDS-defined, and must remain that way for now.  And for this example
> we're not interested in rewriting the RPG using SQL.
>
> We're on 7.1.
>
>
> Thoughts please?
>
> TIA.
>
>
> ~TA~