Pattern Matching

Shell globbing

Pattern matching in the shell against filenames has metacharacters defined differently from the rest of unix pattern matching prgorams. * is match any character except whitespace, ? is match one character except whitespace. so *.c is match any filename ending with the two characters .c (this will list out all c source files in the directory, assuming the directory's owner is sane).

grep, sed

Table of metacharacters:

1. ^ (caret) match beginning of line. Anchors match.
2. $ (dollar sign) match end of line. Anchors match.
3. . (dot) match any character. Beware, command line globbing uses ? instead.
4. * (star) matches zero or more of preceding chracters. Beware, command line uses * as in .*.
5. [] (square braces) set of characters inside braces, match any one of.
6. [^ ] (carat at first character inside braces), match any character except those inside braces
7. [a-z] (use of dash inside braces) match a range. If - is to be matched, must be first character, to avoid misinterpretation as range operator.
8. () {parenthesis, must be escaped with backslash), save match for later use with \n, where n is a number.
9. {m}, {m,} and {m,n} (braces, which must be escaped with a backslash), matched m, more than m, or between m and n repretitions of preceeding character.
10. & (ampersand) expands to the matched string, used in sed.

Grep, sed Flags for grep of note:

• -i, case insensitive
• -v, invert, select non-matching lines
• -c, give count of matching lines.

• -n, print the line only if forced to
• -f, commands from a file

• form is [address][,address][!]command [arguments] You tend to have to enclose this in single quotes of the shell will demolish it. Or double quotes if you want shell variables expanded inside the mess.
• No address: all lines; one address: lines matching address are processed; two addresses: first address starts processing, second address ends processiong.
• Addresses can be line numbers, the dollar sign or a reg. exp enclosed in //.
• example: s/a/b/g, substitute b for a, globally. Drop the g and you only substitute the first occurrance of a on a line. Add p with the g to print out the line, especially if you are using sed -n.
• example: /but/d, delete any line that says "but", not buts allowed!

Examples

Match three letter reversal patterns:

grep '\(.\)\(.\)\(.\)\3\2\1' web2

sed 's/^.*:\*:\([^:]*\).*$/\1/' /etc/passwd