• Choose a category
• All
• Classic-Fiction

function join(array, n, fs, k, s)

{

# Recombine array[1]...array[n] into a string, with elements

# separated by fs

if (n >= 1)

{

s = array[1]

for (k = 2; k <= n; k++)

s = s fs array[k]

}

return (s)

}

String Formatting

The last string functions that we present format numbers and strings under user control: sprintf( format,expression 1,expression 2 ,...) returns the formatted string as its function value. printf( ) works the same way, except that it prints the formatted string on standard output or redirected to a file, instead of returning it as a function value. Newer programming languages replace format control strings with potentially more powerful formatting functions, but at a significant increase in code verbosity. For typical text processing applications, sprintf( ) and printf( ) are nearly always sufficient.

printf( ) and sprintf( ) format strings are similar to those of the shell printf command that we described in Section 7.4. We summarize the awk format items in Table 9-4Table 9-4. These items can each be augmented by the same field width, precision, and flag modifiers discussed in Chapter 7.

The %i, %u, and %X items were not part of the 1987 language redesign, but modern implementations support them. Despite the similarity with the shell printf command, awk's handling of the %c format item differs for integer arguments, and output with %u for negative arguments may disagree because of differences in shell and awk arithmetic.

Table 9-5. printf and sprintf format specifiers

Item

Description

%c

ASCII character. Print the first character of the corresponding string argument, or the character whose number in the host character set is the corresponding integer argument, usually taken modulo 256.

%d, %i

Decimal integer.

%e

Floating-point format ([-]d.precision e[+-]dd).

%f

Floating-point format ([-]ddd.precision).

%g

%e or %f conversion, whichever is shorter, with trailing zeros removed.

%o

Unsigned octal value.

%s

String.

%u

Unsigned value. awk numbers are floating-point values: small negative integer values are output as large positive ones because the sign bit is interpreted as a data bit.

%x

Unsigned hexadecimal number. Letters a-f represent 10 to 15.

%X

Unsigned hexadecimal number. Letters A-F represent 10 to 15.

%%

Literal %.

Most of the format items are straightforward. However, we caution that accurate conversion of binary floating-point values to decimal strings, and the reverse, is a surprisingly difficult problem whose proper solution was only found in about 1990, and can require very high intermediate precision. awk implementations generally use the underlying C library for the conversions required by sprintf( ) format items, and although library quality continues to improve, there are still platforms in which the accuracy of floating-point conversions is deficient. In addition, differences in floating-point hardware and instruction evaluation order mean that floating-point results from almost any programming language vary slightly across different architectures.

When floating-point numbers appear in print statements, awk formats them according to the value of the built-in variable OFMT, which defaults to "%.6g". You can redefine OFMT as needed.

Similarly, when floating-point numbers are converted to strings by concatenation, awk formats them according to the value of another built-in variable, CONVFMT.[4] Its default value is also "%.6g".

The test program in Example 9-8 produces output like this with a recent nawk version on a Sun Solaris SPARC system:

$ nawk -f ofmt.awk

[ 1] OFMT = "%.6g" 123.457

[ 2] OFMT = "%d" 123

[ 3] OFMT = "%e" 1.234568e+02

[ 4] OFMT = "%f" 123.456789

[ 5] OFMT = "%g" 123.457

[ 6] OFMT = "%25.16e" 1.2345678901234568e+02

[ 7] OFMT = "%25.16f" 123.4567890123456806

[ 8] OFMT = "%25.16g" 123.4567890123457

[ 9] OFMT = "%25d" 123

[10] OFMT = "%.25d" 0000000000000000000000123

[11] OFMT = "%25d" 2147483647

[12] OFMT = "%25d" 2147483647