[Issue 949] New: Wrong spec/compiler behaviour for DecimalFloat and HexFloat

[d-bugmail at puremagic.com]

http://d.puremagic.com/issues/show_bug.cgi?id=949

           Summary: Wrong spec/compiler behaviour for DecimalFloat and
                    HexFloat
           Product: D
           Version: 1.006
          Platform: PC
               URL: http://digitalmars.com/d/lex.html
        OS/Version: Windows
            Status: NEW
          Severity: normal
          Priority: P2
         Component: www.digitalmars.com
        AssignedTo: bugzilla at digitalmars.com
        ReportedBy: hennrich.bloebaum at gmx.de


Decimal:
        0
        NonZeroDigit
        NonZeroDigit DecimalDigits

DecimalFloat:
        DecimalDigits .
        DecimalDigits . DecimalDigits
        DecimalDigits . DecimalDigits DecimalExponent
        . Decimal
        . Decimal DecimalExponent
        DecimalDigits DecimalExponent


Current DecimalFloat spec does allow floats like:

writefln(.0);  // 0
writefln(.10); // NonZeroDigit DecimalDigits

but not:

writefln(.01); // 0 DecimalDigits (Zero followed by digits)

There's no rule in the spec for this, but it's lexed by the compiler.
Underscores in exponents aren't handled correct, too. Change DecimalFloat to
somthing similar:


DecimalFloat:
        DecimalDigits2 .
        DecimalDigits2 DecimalDigits .
        DecimalDigits2 . DecimalDigits
        DecimalDigits2 DecimalDigits . DecimalDigits
        DecimalDigits2 . DecimalDigits DecimalExponent
        DecimalDigits2 DecimalDigits . DecimalDigits DecimalExponent
        . DecimalDigits2
        . DecimalDigits2 DecimalDigits
        . DecimalDigits2 DecimalExponent
        . DecimalDigits2 DecimalDigits DecimalExponent
        DecimalDigits2 DecimalExponent
        DecimalDigits2 DecimalDigits DecimalExponent

DecimalDigits2
        0
        NonZeroDigits

DecimalExponentStart
        e
        E
        e+
        E+
        e-
        E-

DecimalExponent
        DecimalExponentStart DecimalDigits2
        DecimalExponentStart DecimalDigits2 DecimalDigits
        DecimalExponentStart DecimalDigits DecimalDigits2


Not sure if it's correct, but the above shouldn't allow:

writefln(._);
writefln(_.);

but:

writefln(.01);




HexFloat:
        HexPrefix HexDigits . HexDigits HexExponent
        HexPrefix . HexDigits HexExponent
        HexPrefix HexDigits HexExponent


Against the spec, this generates no errors by the compiler:

writefln(0x.p1);  // HexPrefix . HexExponent
writefln(0x1.p1); // HexPrefix HexDigits . HexExponent

The first rule should give an error, while the second should be added to the
spec.


--