Float Specification

[mzuenni]

The CTD spec currently says that float := -?[0-9]+(\.[0-9]+)?([eE][+-]?[0-9]+)?
but test_28_prog.in has this SET(a=1/2.). So I guess CTD accepts 2. as float (at least in the code).

What is the intended behaviour here?
(I am currently trying to write a python parser for CTD files, so I might have more of these annoying questions in the future)

[mzuenni]

Related: test_31_data.err10 is 0e00. The name of the test indicates that this should be rejected, but according to the definition above, its a valid float?

[eldering]

Indeed the implementation allows 2. in the checktestdata program, see https://github.com/DOMjudge/checktestdata/blob/main/checktestdata.l#L51-L58, but not in the testdata, see https://github.com/DOMjudge/checktestdata/blob/main/libchecktestdata.cc#L1051-L1052.

Although that's a bit of an inconsistency, I think that it makes sense to leave it this way. Otherwise existing programs might break. Also, it's not really important to enforce strictness in checktestdata programs.

[eldering]

Related: test_31_data.err10 is 0e00. The name of the test indicates that this should be rejected, but according to the definition above, its a valid float?

Indeed, we enforce that integers and floats (both in the exponent and digits before the decimal separator) can't have a leading zero followed by other digits. That's not correctly represented in the grammar specification.

[mzuenni]

Although that's a bit of an inconsistency, I think that it makes sense to leave it this way. [...]
Also, it's not really important to enforce strictness in checktestdata programs.

Yes I agree with that. Maybe the doc should mention this?