Description

In the following example, we want to extract all strings matched [0-9]+(a sequence of digits), and print them one per line.

Note that Script₃ is the most efficient one.

Raw Input		Desired Output
number 123 number 45 number number 6 another 7890 more 13579 no numbers in this line 2468086 last 1248163264 no more		123 45 6 7890 13579 2468086 1248163264

Script and Comments

Script₁ [ 1] s/^[^0-9]*([0-9]+)/\1\n/ [ 2] /\n/!d [ 3] P [ 4] D
Comments The `-r' option of GNU sed must be used or you have to escape parentheses and the `+' quantifier. Each line may contain zero, one or more substrings matched [0-9]+. A line is treated as repeating sequences of unmatched string (may be empty) followed by a matched string. The strategy used by this script is as follows: Take the first unmatched-matched sequence, insert a newline character after the matched string and remove the unmatched one (this is accomplished by Step [1]). After Step [1], the Pattern Space may contain Two lines, where the first one contains only the matched string and the other is the part of the original line which has not been processed yet. Step [3] will print the matched string, and Step [4] will delete it and make sed jump to Step [1]. One line, (Step [1] failed), implying that there are no more matched strings. In this case, Step [2] will delete it and start a new cycle.
Script₂ [ 1] s/^[^0-9]*([0-9]+)/\1\n/ [ 2] /\n/P [ 3] D	Comments A neat version.
Script₃ [ 1] s/[0-9]+/\n&\n/g [ 2] s/[^\n]\n([0-9]\n)/\1/g [ 3] s/\n[^\n]*$//p [ 4] d
Comments Step [1] will enclose every matched string with a pair of newline characters. After Step [1], we can treat a line as a interleaving sequence of unmatched strings and matched strings, where two strings are separated by a newline character. For example, if `X' and `Y' are used to denote an unmatched and a matched string, respectively, then a line can be treated as X \n Y \n X \n Y \n ... \n X Step [2] removes every X and the newline character following it, making the line be Y \n Y \n ...\n Y \n X. Step [3] removes the trailing X and the preceding newline character.