[reportlab-users] RE: Formatting long tables

[Robin Becker]

I'm looking at your original algorithm and it seems quite workable, but
the sample size of rows used to compute the widths is determined by the
available height (ie you consider rows in the first fit tranche). That
makes things a little hard in terms of description.

If I understand the intent of your second proposal we take a fixed
sample of rows to do the estimation. That seems better in that we can
give the user control of that number so they can  use the first n1 rows
plus some n2 other random rows to estimate the width. This method also
makes things independent of the available height for the first tranche.

That's a feasible way to do table sizing provided we have a backup when
things go wrong.
-- 
Robin Becker