Convert free-format input into double precision floating point.
*string char string containing field to be decoded *nstrt int where to start decode (1st = 1)
*nstrt int advanced to next field *dreslt double result *jflag int -1 = -OK, 0 = +OK, 1 = null field, 2 = error
1 A tab character is interpreted as a space, and lower case d,e are interpreted as upper case. 2 The basic format is ^.^@^ where # means + or -, ^ means a decimal subfield and @ means D or E.
Leading spaces are ignored. Embedded spaces are allowed only after # and D or E, and after . where the first ^ is absent. Trailing spaces are ignored; the first signifies end of decoding and subsequent ones are skipped.
Any character other than +,-,0-9,.,D,E or space may be used to end a field. Comma is recognised by slaDfltin as a special case; it is skipped, leaving the pointer on the next character. See 12, below. 5 Both signs are optional. The default is +. 6 The mantissa defaults to 1. 7 The exponent defaults to e0. 8 The decimal subfields may be of any length. 9 The decimal point is optional for whole numbers. 10 A null field is one that does not begin with +,-,0-9,.,D or E, or consists entirely of spaces. If the field is null, jflag is set to 1 and dreslt is left untouched. 11 nstrt = 1 for the first character in the string. 12 On return from slaDfltin, nstrt is set ready for the next decode - following trailing blanks and (if used) the comma separator. If a separator other than comma is being used, nstrt must be incremented before the next call to slaDfltin.
a) A +, -, D or E is left unsatisfied. b) The decimal point is present without at least one decimal subfield. c) An exponent more than 100 has been presented. 14 When an error has been detected, nstrt is left pointing to the character following the last one used before the error came to light. This may be after the point at which a more sophisticated program could have detected the error. For example, slaDfltin does not detect that '1e999' is unacceptable until the whole field has been read. 15 Certain highly unlikely combinations of mantissa & exponent can cause arithmetic faults during the decode, in some cases despite the fact that they together could be construed as a valid number. 16 Decoding is left to right, one pass.
a) As dictated by the string length. b) Detected during the decode. (b overrides a.) P.T.Wallace Starlink 17 November 1993