One-Liners

sed -n '5p'            file   # print line 5
sed -n '2,4p'          file   # lines 2 through 4
sed -n '$p'            file   # last line, like tail -1
sed -n '5q;p'          file   # first 4 lines, like head -4, no SIGPIPE
sed -n '/re/p'         file   # like grep
sed -n '/re/!p'        file   # like grep -v
sed -n '/re/Ip'        file   # like grep -i
sed -n '/re/{p;q;}'    file   # first match only, then stop reading
sed -n '$='            file   # line count, like wc -l
sed -n '/re/='         file   # line numbers of matches
sed -n '/^.\{65\}/p'   file   # lines longer than 64 characters
sed -n 'n;n;p'         file   # every third line, portable
gsed -n '0~3p'         file   # every third line. GNU only

sed '5d'              file   # drop line 5
sed '2,4d'            file   # drop a range
sed '$d'              file   # drop the last line
sed '1,2d'            file   # skip a two-line header
sed '/^$/d'           file   # drop every blank line
sed '/^[[:space:]]*$/d' file # drop whitespace-only lines too
sed '/^#/d;/^$/d'     file   # strip comments and blanks from a config
sed '/re/,$d'         file   # from the first match to end of file
sed '/^begin$/,/^end$/d' file  # drop a marked block
sed '/^$/{N;/^\n$/D;}' file  # squeeze runs of blanks down to one

sed 's/old/new/'        file   # first match on each line
sed 's/old/new/g'       file   # every match
sed 's/x/Y/3'           file   # the third match only
gsed 's/x/Y/3g'         file   # third match onward. GNU only, BSD errors
sed 's/[[:space:]]\{1,\}/ /g'  file   # collapse runs of whitespace
sed 's/^/> /'           file   # quote a mail reply
sed 's/$/;/'            file   # append to every line
sed 's|/usr/local|/opt|g' file # alternate delimiter, no escaping needed
sed 's/<[^>]*>//g'      file   # crude tag stripper
sed 's/\(.*\)\//\1|/'   file   # change the LAST slash, thanks to greedy .*
sed '5!s/a/b/'          file   # everywhere except line 5
sed '/^#/!s/a/b/'       file   # skip comment lines

sed 's/^[[:space:]]*//'              file   # trim leading
sed 's/[[:space:]]*$//'              file   # trim trailing
sed 's/^[[:space:]]*//;s/[[:space:]]*$//' file   # trim both
sed 's/\r$//'                        file   # CRLF to LF, dos2unix
sed 's/$/\r/'                        file   # LF to CRLF, unix2dos
sed -n '/\r$/l'                      file   # find the CR-terminated lines
sed 's/\t/    /g'                    file   # tabs to four spaces
sed 's/  */\t/g'                     file   # spaces to tabs

sed "s/$(printf '\t')/    /g" file   # portable tabs to spaces
sed $'s/\t/    /g'            file   # bash and zsh, also portable to BSD sed

sed '1!G;h;$!d'    file   # reverse the file, like tac
sed -n '1!G;h;$p'  file   # same thing, shorter to read
sed '/\n/!G;s/\(.\)\(.*\n\)/&\2\1/;//D;s/.//'   # reverse each line, like rev
sed = file | sed 'N;s/\n/\t/'      # number lines, like cat -n
sed G             file   # double-space
sed 'n;d'         file   # undo double-spacing
sed -e :a -e '$q;N;3,$D;ba' file   # last 2 lines, like tail -2
sed -e :a -e 'N;$!ba' -e 's/\n/,/g' file   # join all lines with commas
sed '$!N;s/\n/ /'  file   # join lines in pairs
sed '$!N; /^\(.*\)\n\1$/!P; D'  file   # like uniq

sed -n '/BEGIN/,/END/p'          file   # the block, markers included
sed -n '/BEGIN/,/END/{//!p;}'    file   # the block, markers excluded
sed '/BEGIN/,/END/d'             file   # everything but the block
sed -n '/re/{=;p;}' file | sed 'N;s/\n/:/'   # like grep -n

# print every blank-line-separated paragraph that contains foo
sed -e '/./{H;$!d;}' -e 'x;/foo/!d' file

# reflow each paragraph onto one line
sed -e '/./{H;$!d;}' -e 'x;s/\n/ /g;s/^ //;p' -e d file

sed $'1i\\\nHEADER'  file   # prepend a line, bash or zsh
sed $'$a\\\nFOOTER'  file   # append a line
sed $'/^\\[main\\]$/a\\\nkey = value'  config   # add under a section
gsed '1i HEADER'     file   # GNU only one-line form
sed '1r header.txt'  file   # splice a file in after line 1
sed '$r footer.txt'  file   # splice a file in at the end
sed '10q'            file   # stop after 10 lines
sed '/^done$/q'      file   # stop at a sentinel
gsed '/error/q1'     file   # exit 1 if a match was seen. GNU only