Hi,

I am with Craig and Tamar here, with one possible exeption.

Doing time series, it sometimes pays off to have denormalized data,
since record movement on the disk is minimized.

If you have such denormalization and access the fields
via loops and name expressions like

for lnMonth = 1 to 120
  replace ("nPrz" + Padl(lnMonth, 3,"0")) with ;
      lnNewPrice/eval("TableUsedPr.nPrc" + Padl(lnMonth,3,"0"))
endfor

it is nearly always faster to at least scatter each record into a single array for processing, if you have arranged your record structure(s) accordingly.

HTH

thomas