Dynamically built selection sets

Ray Scheel ray.scheel at wsdtx.org
Thu Jul 1 14:45:22 PDT 2004


The mention of languishing bug reports and another project I'm working on
jogged my memory about my recent but unacknowledged bug report about the
incompatibility with @pw- at pz with index scan in selection sets (at least
under Linux).  I've sent another query to fpsupport about the status of that
report, but in the meantime I am wondering about how to create a more
effective workaround that would build a selection set that would be
compatible with the index scan feature using information entered/passed into
the calling shell script, presuming a working sel.* table can be built by a
*nix script.

This is what at least works when index scan is on and there is an index
built on fields 3 and 4, as it bypasses whatever triggers index scan and
steps through all records to do the comparison:
::::::::::::::
my_table/sel.works
::::::::::::::
:@px:eqf:4
:@py:eqf:3

where this does not work (does not select anything) on the same values
passed:
::::::::::::::
my_table/sel.dontwork
::::::::::::::
:3:eqf:@py
:4:eqf:@px

I was wondering what it would take in a *nix shell or PHP script to manually
build a selection table using the passed/entered values such that a file
would be created like the following I could then use as a -s in a call to
rreport and immediately delete: 

::::::::::::::
my_table/sel.temp_rand#
::::::::::::::
:3:eqf:[contents of @py]
:4:eqf:[contents of @px]

Actually, I'm hoping someone already has such a beast they would share that
would be a better (functionally / faster for the end user) workaround that
bypassing index scan on 500K+ record files where index scan on a manually
built file finishes selections in a flash while the dynamic method takes
several minutes.

Ray

Ray Scheel               ray.scheel at wsdtx.org 
Windham School District     Computer Services
Programmer                       936-291-5206 


More information about the Filepro-list mailing list