t/31.schema.t fails tests 238 and 3838 when nvtype is IBM DoubleDouble about yaml-pp-p5

[perlpunk]

Thanks!

I actually want to compare via == and not eq ($def{data} could contain 300.00 for example).
I just didn't think of floating point arithmetic problems, although I should have known it ;-)

I fixed the test simply by avoiding rounding errors with sprintf in 9e907d1

Could you test it?

[sisyphus]

It seems that the first change where cmp_ok() is changed from performing an 'eq' test to an '==' might be the wrong thing.
It produces thousands of non-numeric warnings of the form:
Argument "*" isn't numeric in numeric eq (==) at (eval in cmp_ok) t/31.schema.t line 120.
where "*" is such things as "true", "y", "yes", "no", "null", "off", "on", "~", to name a few.

That's not causing any tests to fail, but for me the second change, where $data is replaced by sprintf("%0.2f",$data), is alone sufficient to allow all tests to pass.

However, I don't think I'm using the exact same version of 31.schema.t as is presented at 9e907d1 .
What seems to be line 128 of that file is actually line 120 on my copy of the file.

The interesting thing about that second change is that it fixes the failures by coercing the LHS of the '==' comparison to the same incorrect representation of 685230.15 as held by the RHS ($def{data}).
That approach is a little unusual and threw me initially - though it's a totally sane, valid and most practical solution.

Cheers,
Rob

[perlpunk]

Huh, weird.
It looks like the change got accidentally applied to the $type eq 'str' condition on your side. Then I get the same warnings as you.
Here's the direct link to the correct file (branch testfloat):
https://github.com/perlpunk/YAML-PP-p5/blob/testfloat/t/31.schema.t

Thanks!

[sisyphus]

Yep - https://github.com/perlpunk/YAML-PP-p5/blob/testfloat/t/31.schema.t is fine.

Cheers,
Rob