|
The char_to_decimal functions parse a numeric token from at most nmax characters in a string **pc or file *pf or function ((*pget)) into a decimal record *pd, classifying the form of the string in *pform
and *pechar. The accepted syntax is intended to be sufficiently flexible to accommodate many languages: whitespace value or whitespace
sign value,where whitespace is any number of characters defined by isspace in <ctype.h>, sign
is either of [+-], and value can be number, nan, or inf. inf
can be INF (inf_form) or INFINITY (infinity_form) without regard to case. nan can be NAN (nan_form) or NAN(nstring) (nanstring_form) without regard to case; nstring is any string of characters not containing ')' or NULL; nstring is copied to pd->ds and, currently, not used subsequently. number consists of significand or significand efield where significand must contain one or more digits and may contain one point; possible forms are
| |
digits (int_form)
digits. (intdot_form)
.digits (dotfrac_form)
digits.digits (intdotfrac_form)
|
efield consists of echar digits or echar sign digits,.where echar is one of [Ee], and digits contains one or more digits.
When fortran_conventions is nonzero, additional input forms are accepted according to various Fortran conventions:
- 0
- no Fortran conventions
- 1
- Fortran list-directed input conventions
- 2
- Fortran formatted input conventions, ignore blanks (BN)
- 3
- Fortran formatted input conventions, blanks are zeros (BZ)
When fortran_conventions is nonzero, echar may also be one of [DdQq], and efield may also have the form
sign digits.
When fortran_conventions>= 2, blanks may appear in the digits strings for the integer, fraction, and exponent fields and may appear between echar and the exponent sign and after the infinity and NaN forms. If fortran_conventions== 2, the blanks are ignored. When fortran_conventions==
3, the blanks that appear in digits strings are interpreted as zeros, and other blanks are ignored.
When fortran_conventions is zero, the current locale's decimal point character is used as the decimal point; when fortran_conventions is nonzero,
the period is used as the decimal point.
The form of the accepted decimal string is placed in *pform. If an efield is recognized, *pechar is set to point to
the echar.
On input, *pc points to the beginning of a character string buffer of length >= nmax. On output, *pc points to a character
in that buffer, one past the last accepted character. string_to_decimal() gets its characters from the buffer; file_to_decimal() gets its characters from *pf and records them in the buffer, and places a null after the last character read. func_to_decimal() gets its characters from an int function ((*pget)) .
The scan continues until no more characters could possibly fit the acceptable syntax or until nmax characters have been scanned. If the nmax
limit is not reached then at least one extra character will usually be scanned that is not part of the accepted syntax. file_to_decimal() and func_to_decimal() set *pnread to the number of characters read from the file; if greater than nmax, some characters were lost. If no characters were lost, file_to_decimal()
and func_to_decimal() attempt to push back, with ungetc(3C)
or ((*punget)) , as many as possible of the excess characters read, adjusting *pnread accordingly. If all unget calls are successful,
then **pc will be NULL. No push back will be attempted if ((*punget)) is NULL.
Typical declarations for *(pget) and *(punget) are:
| |
int xget(void)
{ ... }
int (*pget)(void) = xget;
int xunget(int c)
{ ... }
int (*punget)(int) = xunget;
|
If no valid number was detected, pd->fpclass is set to fp_signaling, *pc is unchanged, and *pform is set to invalid_form.
atof(3C) and strtod(3C) use string_to_decimal(). scanf(3C) uses file_to_decimal().
|
