string_to_decimal

, file_to_decimal

, func_to_decimal

- parse characters into decimal record

Synopsis

#include <floatingpoint.h>

void string_to_decimal(char **pc, int nmax, 
     int fortran_conventions, decimal_record *pd,
     enum decimal_string_form *pform, char **pechar);

void func_to_decimal(char **pc, int nmax, 
     int fortran_conventions, decimal_record *pd,
     enum decimal_string_form *pform, char **pechar,
     int (*pget)(void), int *pnread, int (*punget)(int c));

#include <stdio.h>

void file_to_decimal(char **pc, int nmax,
     int fortran_conventions, decimal_record *pd,
     enum decimal_string_form *pform, char **pechar,
     FILE *pf, int *pnread);

Description

These functions attempt to parse a numeric token from at most nmax characters read from a string **pc, a file *pf, or function (*pget). They set the decimal record *pd to reflect the value of the numeric token recognized and set *pform and *pechar to indicate its form.

The accepted forms for the numeric token consist of an initial, possibly empty, sequence of white-space characters, as defined by isspace(3C), followed by a subject sequence representing a numeric value, infinity, or NaN. The subject sequence consists of an optional plus or minus sign followed by one of the following:

• a non-empty sequence of decimal digits optionally containing a decimal point character, then an optional exponent part

• one of INF or INFINITY, ignoring case

• one of NAN or NAN(string), ignoring case in the NAN part; string can be any sequence of characters not containing ')' (right parenthesis) or '\0' (null).

The fortran_conventions argument provides additional control over the set of accepted forms. It must be one of the following values:

0

no Fortran conventions

1

Fortran list-directed input conventions

2

Fortran formatted input conventions, blanks are ignored

3

Fortran formatted input conventions, blanks are interpreted as zeroes

When fortran_conventions is zero, the decimal point character is the current locale's decimal point character, and the exponent part consists of the letter E or e followed by an optional sign and a non-empty string of decimal digits.

When fortran_conventions is non-zero, the decimal point character is “.” (period), and the exponent part consists of either a sign or one of the letters E, e, D, d, Q, or q followed by an optional sign, then a non-empty string of decimal digits.

When fortran_conventions is 2 or 3, blanks can appear in the digit strings for the integer, fraction, and exponent parts, between the exponent delimiter and optional exponent sign, and after an INF, INFINITY, NAN, or NAN(string). When fortran_conventions is 2, all blanks are ignored. When fortran_conventions is 3, blanks in digit strings are interpreted as zeros and other blanks are ignored.

The following table summarizes the accepted forms and shows the corresponding values to which *pform and pd->fpclass are set. Here digits represents any string of decimal digits, “.” (period) stands for the decimal point character, and exponent represents the exponent part as defined above. Numbers in brackets refer to the notes following the table.

Notes:

1. The whitespace_form is accepted only when fortran_conventions is 2 or 3 and is interpreted as zero.

2. For all numeric forms, pd->fpclass is set to fp_normal if any non-zero digits appear in the integer or fraction parts, and otherwise pd->fpclass is set to fp_zero.

If the accepted token has one of the numeric forms and represents a non-zero number x, its significant digits are stored in pd->ds. Leading and trailing zeroes and the radix point are omitted. pd->sign and pd->exponent are set so that if m is the integer represented by pd->ds,

-1**(pd->sign) * m * 10**(pd->exponent)

approximates x to at least 511 significant digits. pd->more is set to 1 if this approximation is not exact (that is, the accepted token contains additional non-zero digits beyond those copied to pd->ds) and to 0 otherwise.

If the accepted token has the NAN(string) form, up to 511 characters from the string part are copied to pd->ds.

pd->ds is always terminated by a null byte, and pd->ndigits is set to the length of the string stored in pd->ds.

On entry, *pc points to the beginning of a character string buffer. The string_to_decimal() function reads characters from this buffer until either enough characters are read to delimit the accepted token (for example, a null character marking the end of the string is found) or the limit of nmax characters is reached. The file_to_decimal() function reads characters from the file *pf and stores them in the buffer. The func_to_decimal() function reads characters one at a time by calling the function (*pget)() and stores them in the buffer; (*pget)() must return integer values in the range -1 to 255, where -1 is interpreted as EOF and 0, ..., 255 are interpreted as unsigned char values. Both file_to_decimal() and func_to_decimal() read characters until either enough characters are read to delimit the accepted token, EOF is encountered, or the limit of nmax characters is reached. These functions, therefore, typically read one or more additional characters beyond the end of the accepted token and attempt to push back any excess characters read. Provided that the punget argument is not NULL, func_to_decimal() pushes back characters one at a time by calling (*punget)(c), where c is an integer in the range 0 to 255 corresponding to a value previously read via (*pget)(). After pushing back as many excess characters as possible, file_to_decimal() and func_to_decimal() store a null byte in the buffer following the last character read and not pushed back and set *pnread to the number of characters stored in the buffer prior to this null byte. Since these functions can read up to nmax characters, the buffer must be large enough to hold nmax + 1.

On exit, *pc points to the next character in the buffer past the last one that was accepted as part of the numeric token. If no valid token is found, *pc is unchanged. If file_to_decimal() and func_to_decimal() successfully push back all unused characters, *pc points to the null byte stored in the buffer following the last character read and not pushed back.

If the accepted token contains an exponent part, *pechar is set to point to the position in the buffer where the first character of the exponent field is stored. If the accepted token does not contain an exponent part, *pechar is set to NULL.

Usage

If the _IOWRT flag is set in *pf, file_to_decimal() reads characters directly from the file buffer until a null character is found. (The _IOWRT flag should only be set when file_to_decimal() is called from sscanf(3C).) Otherwise, file_to_decimal() uses getc_unlocked(3C), so it is not MT-safe unless the caller holds the stream lock.

Attributes

See attributes(5) for descriptions of the following attributes:

See Also

ctype(3C), decimal_to_floating(3C), getc_unlocked(3C), isspace(3C), localeconv(3C), scanf(3C), setlocale(3C), strtod(3C), ungetc(3C), attributes(5)