|
|
[Contents]
[Previous] [Next]
[Last]
Chapter 26
Floating Point Number to String Conversion
NSPR provides functions that convert double-precision floating point numbers
to and from their character string representations. The header file
prdtoa.h declares these functions.
These conversion functions were originally written by David M. Gay of
AT&T. They use IEEE double-precision (not IEEE double-extended)
arithmetic.
PR_strtod
Syntax
#include <prdtoa.h>
PRFloat64 PR_strtod(const char *s00, char **se);
Parameters
The function has the following parameters:
| s00 |
The input string to be scanned. |
| se |
A pointer that, if not NULL, will be assigned the address
of the last character scanned in the input string. |
Returns
The result of the conversion is a PRFloat64 value representative
of the input string. If the parameter se is not NULL
the location it references will also be set.
Description
PR_strtod converts the prefix of the input decimal string pointed
to by s00 to a nearest double-precision floating point number.
Ties are broken by the IEEE round-even rule. The string is scanned
up to the first unrecognized character. If the value of se
is not (char **) NULL, PR_strtod stores a pointer to
the character terminating the scan in *endptr. If the answer
would overflow, a properly signed HUGE_VAL (infinity) is returned.
If the answer would underflow, a properly signed 0 is returned. In
both cases, PR_GetError() returns the error code PR_RANGE_ERROR.
If no number can be formed, se is set to s00, and 0 is
returned.
PR_dtoa
Syntax
#include <prdtoa.h>
PRStatus PR_dtoa(
PRFloat64 d,
PRIntn mode,
PRIntn ndigits,
PRIntn *decpt,
PRIntn *sign,
char **rve,
char *buf,
PRSize bufsz);
Parameters
PR_dtoa takes the following parameters:
| d |
The floating point number to be converted to ASCII. |
| mode |
The type of conversion to employ.
| 0 |
Shortest string that yields d when read in and rounded to
nearest. |
| 1 |
Like 0, but with Steele & White stopping rule [2];
e.g., with IEEE P754 arithmetic, mode 0 gives 1e23 whereas mode 1 gives
9.999999999999999e22. |
| 2 |
max(1, ndigits) significant digits. This gives a return
value similar to that of ecvt, except that trailing zeros are
suppressed. |
| 3 |
Through ndights past the decimal point. This gives a
return value similar to that from fcvt, except that trailing zeros
are suppressed, and ndigits can be negative. |
| 4,5,8,9 |
Same as modes [2..3] but using left to right digit generation. |
| 6-9 |
Same as modes [2..3] but don't try fast floating-point estimate (if
applicable). |
| all others |
Treated as mode 2. |
|
| ndigits |
The number of digits desired in the output string. |
| decpt |
A pointer to a memory location where the runtime will store the offset,
relative to the beginning of the output string, of the conversion's decimal
point. |
| sign |
A location where the runtime can store an indication that the conversion
was of a negative value. |
| rve |
If not NULL this location is set to the address of the end
of the result. |
| buf |
The address of the buffer in which to store the result. |
| bufsz |
The size of the buffer provided to hold the result. |
Results
The principle output is the null-terminated string stored in buf.
Trailing zeros are suppressed.
If rve is not NULL, *rve is set to point to the end
of the returned value.
If the input parameter, d, is +infinity, -infinity
or NAN, *decpt is set to 9999.
Description
Sufficient space is allocated to the return value to hold the suppressed
trailing zeros.
PR_cnvtf
Syntax
#include <prdtoa.h>
void PR_cnvtf(char *buf, PRIntn bufsz, PRIntn prcsn, PRFloat64 fval);
Description
PR_cnvtf is a simpler interface to convert a floating point number
to a string. In fact, it is written to conform to the ECMA standard
of Javascript (ECMAScript).
The input argument fval is the double-precision floating point
number to be converted, and prcsn is the number of digits of precision
to generate floating point value. On return, the result is written
to the buffer pointed to by buf of size bufsz.
References
Gay's implementation is inspired by these two papers.
[1] William D. Clinger, "How to Read
Floating Point Numbers Accurately," Proc. ACM SIGPLAN '90, pp. 92-101.
[2] Guy L. Steele, Jr. and
Jon L. White, "How to Print Floating-Point Numbers Accurately," Proc.
ACM SIGPLAN '90, pp. 92-101.
[Contents] [Previous]
[Next] [Last]
Last updated: Wed Jul 15 09:53:03 PDT
1998
Copyright © 1998 Netscape
Communications Corporation
|
