randline(1)

NAME

SYNOPSIS

DESCRIPTION

randline selects a random line or lines from its input, either standard input or the named files. The algorithm used is the classic one-pass one described in Knuth, with a couple of extensions.

There are a few options:

-kh

(``keep header'') Always pass through the first line of the input (which is assumed to be a header of some sort); perform random selection from among the remaining lines.

-n n

(or just -n) Select n lines.

-s

(``stable'') When selecting more than one line, print them out in the same order they were in in the input, rather than shuffling them randomly.

-seed s

Use initial random seed s.

-seedfile f

Save random seed state in file f for repeatable behavior between runs.

-version

Print program's version number.

-w

Use weighted probabilities when selecting lines. The first (or last, whichever looks numeric) column on each input line gives its ``weight''. Lines with higher weights are selected with higher probability: specifically, if two lines a and b have weights in the ratio m:n, line a will be selected m/n times as often as line b. (The weights of the selected lines are not printed on output.)

Another way of thinking about this is that if an input line has weight w, the behavior is the same as if that line appeared in the input w times.

(This feature is still somewhat experimental.)

-?, -help

Print a brief help message.

BUGS

There is a fixed limit (currently 1000) on the longest input line that can be read.

There is no provision for fractional or floating-point weights under the -w option; all weights must be integral.

SEE ALSO

HISTORY

The program is based on the algorithm in Knuth, section xxx.

This documentation corresponds to version 3.01 of the program.

See http://www.eskimo.com/~scs/src/#randline for possible updates.

AUTHOR

Steve Summit, scs@eskimo.com