Hello Mark, >>> Any objections if I just remove this caution in the new version? >> >> No, please correct this. I sometimes use expressions for ordering >> levels, e.g. with IIF or DECODE constructs that return, say, the >> numbers 1, 2 and 3, and they do the expected thing in Fb 2.5 and 3, >> i.e. they sort by the numerical value returned, not by the contents >> of column 1, 2 or 3. > > Did you mean "No, I have no objections", or "No, don't remove it, but > correct it"?
I meant: no objections, please remove the incorrect caution! (I.e., correct the section by removing what's incorrect.) > The caution is entirely wrong. I could write another note > to explain the actual behaviour, but that sounds more like an > improvement for later than removing factually incorrect information. I guess that's up to you. The most important thing is to remove incorrect information. Then, while you're at it, *if* you have time and *if* you feel like it, you could write a useful and correct note. But that can also wait till later. > > It would be silly anyway, because the value of X+Y may vary wildly > > across rows, so which one would determine the column number > > to sort by? > > > > Maybe what they're (trying to) say is true for constant expressions > > and/or expressions involving only constants and local variables in > > PSQL. > > Constant expressions other than `number` or `(number)` sort by the value > as well (so `1+1` does not sort by column 2). You're right. > Sorting by a PSQL variable > (is that possible?) will likely also not sort by column position, as the > sort needs to be known at prepare time. > > Interestingly enough, Firebird does perform some sort of sort when using > constant expressions. Without thorough testing it looks like it performs > an equivalent of `ORDER BY <all columns of the select clause>`, but that > might just be the limited set of data I used when testing. I just played around a bit and indeed it seems that adding an ordering item like "3+2" - with or without parentheses - causes some ordering pattens in (parts of) the output, but I'm not able to tell what it does exactly - if anything! Cheers, Paul _______________________________________________ Firebird-docs mailing list Firebird-docs@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/firebird-docs
