Negative integer literals (w/unary minus sign) incorrectly interpreted as floats

[jquartel]

If you have an integer-only expression you expect the result to remain integer, but here are a couple of instances where using a negative integer can result in a floating point result (note that the only way I know how to test for the resultant type is to use toString):

1. If the first (or only) value of the expression is a negative integer, as in
   toString(-1+2,"%d")

2. If a variable in a with() or forEach() function is a negative integer, as in
   with(-1,a,toString(2+a,"%d"))

In both of these cases the result is
Error: java.util.IllegalFormatConversionException: d != java.lang.Double
but if you use "%f" in the string conversion you get 1.000000 .

I am using OpenRefine Version 3.6.2 in Safari Version 16.1 on MacOS 12.6.1 .

[tfmorris]

Very weird! I can confirm the bug, but it's in code that's worked for over a decade. I strongly suspect a compiler bug, but I've reorganized things to work around the problem.

[tfmorris]

Actually, this is probably due to the binary numeric promotion which is called out in the JLS, so perhaps it's been broken for a while due to the specification being tightened up, but we didn't have test coverage for this case, so no telling how long it's been broken.