Re: Should to_number return a sequence?
- Posted by petelomax 5 days ago
- 293 views
Why shouldn't it return a number or NaN on failure?
Or does NaN not exist in Eu?
I can't speak definitively to whether or not NaN exists in Euphoria or Phix. But, I can say, I haven't seen that yet... sort of.
-- inf = 1e300*1e300 #ilASM{ fld1 fldz fdivp [32] lea edi,[inf] [64] lea rdi,[inf] [] call :%pStoreFlt } -- Erm, this one is a bit bizarre... -- On the one hand it seems RDS Eu does not support nan properly, but then it somehow does... -- If you try testing for nan, it seems to go all pear-shaped, but avoiding the tests -- seems to make it happy again, and yet print "nan" and "inf" like a good little boy... -- Of course, you shouldn't be using this code on RDS Eu anyway. -- --/**/ nan = -(inf/inf) --/* Phix nan = 3.245673689e243 -- RDS --*/
So that's how I've been creating nan for the past 15 years or more, but I cannot remember why I replaced 1e300*1e300 and
I'm sure I still use that method in a few other places (including 64-bit) - it might very well have been a multithreading issue.
I suspect there's a fairly strong probability that specific value for nan only works on 32-bit RDS Eu.
-- Phix to_number("x") -- this returns a sequence. Which is Not a Number.
-- Phix to_number("x",0) -- this returns 0. Which is a Number.
PS: I would have loved to have been a fly on the wall and see how you found to_number in the Euphoria docs after failing to find it in Phix.chm...
PPS: At least in my book, to_number() [optionally/sometimes] delivering a typecheck is [often] a perfectly good debugging technique.
