Don't read the following!  ;-)

On 18/12/2019 06:39, Bernie van't Hof wrote:
> Well I'm not much better!
> 
> Just about sorted crosstab (at least conceptually) and made an awkward 
> discovery in the process.
> 
> CREATE TABLE `Answer` AS SELECT ... seems innocuous enough.
> 
> But I hadn't noticed before that it brings NULL/NOTNULL settings from 
> SELECT columns into the new table. The benefits are obvious, but is 
> different to pdox which generates the Answer table unkeyed. Ho Hum :(

"Unkeyed" sounds correct to me - we're emulating a PDoxDOS app! Surely 
(?) the Answer table will contain numerous data and index fields, from 
some/all the contributing tables, and maybe a few calculated fields, 
summary fields, counts, etc, but presumably there's no way of knowing 
what would be a suitable primary index on that answer table, nor what 
sec-indices would be useful?

I presume (but I forget) that unfilled data/index fields in PDoxDOS will 
contain spaces (for strings), 0.00 (IEEE) for floats, binary-zeros 
(NULL) for binary/integer fields, probably 0000 (binary or IEEE?) also 
for dates, 0.000 (IEEE) for [Dates+]Times, etc, and I expect unfilled 
SQL fields would contain the same values? I wonder would empty 
Dates/Times be different?

If I can do a bit of digging for you, let me know.

   - Michael