string = const char $\ast$ (Given)

nstrt = int $\ast$ (Given and Returned)

dreslt = double $\ast$ (Returned)

jflag = int $\ast$ (Returned)

• Uses the strtod() system call to do the parsing. This may lead to subtle differences when compared to the SLA/F parsing.

• All " D" characters are converted to " E" to handle fortran exponents.

• Commas are recognized as a special case and are skipped if one happens to be the next character when updating nstrt. Additionally the output nstrt position will skip past any trailing space.

• If no number can be found flag will be set to 1.

• If the number overflows or underflows jflag will be set to 2. For overflow the returned result will have the value HUGE_VAL, for underflow it will have the value 0.0.

• For compatiblity with SLA/F -0 will be returned as " 0" with jflag == -1.

• Unlike slaDfltin a standalone " E" will return status 1 (could not find a number) rather than 2 (bad number).

• The code is more robust if the C99 copysign() function is available. This can recognize the -0.0 values returned by strtod. If copysign() is missing we try to scan the string looking for minus signs.