Re: New COUP*() test files and more about Gnumeric performance

[Nick Lamb <njl98r ecs soton ac uk>]

By the time you read this a new 97coup.zip file should be available on
my web site http://www.ecs.soton.ac.uk/~njl98r/excel/ with the "correct"
values from Excel 97. They are, of course, different from those in the
Excel XP files I provided earlier.

On Mon, Jan 21, 2002 at 01:07:24PM +1000, Phillip Shelton wrote:
> My copy of Excel 97 opened them all and past all the tests.

Nope. Your copy of Excel 97 opened them, printed out the values stored in
the file (which include "OK") and did no calculations of any kind. Even
if you pressed "F9 (Recalculate)" it will not recalculate in this case.

In my experience Excel 97 can be best made to recalculate formulae by
manually re-entering them, but in this case you can get away with simply
selecting the first row of formula entries (columns D to R I think) and
using the lower-right square marker on the selection to extend this
formula back over the 4999 other rows.

Then wait for a bit, Excel now has to (shock!) read the spreadsheet that
you loaded and actually do some calculations. There is feedback in the
status bar IIRC.

Excel isn't smart enough to optimise this out in Excel 97 and it will
therefore recalculate the formulae. For most of the sheets (e.g. the
xpcoupncd.xls and xpcouppcd.xls files) saved from Excel XP this will
cause the triggers to read "Fail" instead of "OK".

The Excel 2k and Excel XP trick of using search & replace to recalculate
a spreadsheet doesn't work in Excel 97 because it can't handle replace
operations on function names without filling the sheet with #NAME?

<rant>
Please do not ever provide a mechanism in Gnumeric which sidesteps the
full recalculation when requested by a user. By all means provide a
heuristic-driven "fast" recalc, but it must be possible for a user to
say "I don't trust you. Work this out again from scratch".
</rant>

Nick.