#Sed Notes II##
-
Multiple-line pattern space
N, D, P: multiple line
n, d, p: single line
sed '
/operator/{
N
s/Owner and Operator\nGUIde /Installation GuIde #\n means to insert a new line,
}' #but new line cannot be put in the replacement
sed '
/operator/{
N
s/Owner and operator\nGuide/Installation Guide\
/
}'
sed '
/Owner/{
N
s/Owner *\n*and *\n*Operator *\n*Guide /Installation Guide # *\n* means the new line is optional
}'
sed '
s/Owner and Operator Guide/Installation Guide/
/Owner/{
N
s/ *\n/ /
s/Owner and Operator Guide */Installation Guide\
/
}'
-
$!N # It excludes the end of the line from N command
-
A script for extracting figures
sed '
/<para>/{ # when <para> occurs
N
C\ # Change the multiple lines to
.LP
}
/<Figure Begin>/,/<Figure End>/>{ # between <Figure Begin> and <Figure End>
w fig.interleaf # write to file fig.interleaf
/<Figure End>/i\ # insert the following
.FG\
<insert figure here>\
.FE
d # delete the original patterns
}
/^$/d' # delete blank lines
-
multi-line deletion
Cf.
sed '
/^$/{
N
/^\n$/d
}' # only removes the additional blank lines in odd lines
and
sed /^\n$/D # removes all additional blank lines (It deletes at least two blank lines that occur together)
-
Multi-line print
Cf:
sed '
/UNIX$/{
N # append a new input
s/\nSystem/Operating &/ # & means the regex 1
P # print the modified line
D # prevent the repeating
}
and
sed '
/UNIX$/{
N
/\nSystem/{
s// Operating &/
P
D
}
}'
-
sed 's/@f1(\(.*\))/\\fB\1\\fR/g'
# The original file use @f1 to bold the text, so the script use fB and fR to do so.
# \1\ matches the pattern to be kept \(.*\)
sed 's/@f1(\([^)]*\))/\\fB\1\\fR/g'
# use [^)] to replace .*
sed '
s/@f1(\([^)]*\))/\\fB\1\\fR/g
/@f1(.*/{
N
s/@f1(\(.*\n[^)]*\))/\\fB\1\\fR/g
}'
sed '
s/@f1(\([^)]*\))/\\fB\1\\fR/g
/@f1(.*/{
N
s/@f1(\(.*\n[^)]*\))/\\fB\1\\fR/g
P
D
}'
# Once making a substitution across two lines,
# print the first line and then delete it.
# With the second portion remining in the pattern space,
# control passes on the top of the script where we see
# if there is an @f1 remaining on the line.
-
Hold that line
h, H # Hold: copy or append contents of pattern space to hold space.
g, G # Get: copy or append contents of hold space to pattern space.
x # Exchange: Swap contents of hold space and pattern space.
the lowercase commands overwrite the contents of the target buffer,
the uppercase commands append to the existing contents.
Trial:
1
2
11
22
111
222
sed '
/1/{
h
d
}
/2/{
G
}'
Result:
2
1
22
11
222
111
# hold command with delete command is a very common pairing.
-
capital transformation
sed '
/the .* statement/{
h # hold the pattern
s/.*the \(.\) statement.*/\1/ # extract the pattern
y/abcdefghijklmnopqrstuvwxyz/ABCDEFGHIJKLMNOPQRSTUVWXYZ/ # transform
G # get the pattern from the hold space
s/\(.*\)\n(.*the \).*\( statement.*\)/\2\1\3/ # reorder
}'
-
Correcting index entries
grep "^\.XX" $* | sort -u |
sed '
h
s/[][\\*.]\\&/g
x
s/[\\&]/\\&/g
s/^\.XX //
s/$/\//
x
s/^\\\.XX \(.*\)$/\/^\\.XX \/s\/\1\//
G
s/\n//'
sed '/^\.XX /s/"asterisk (\*) metacharacter"/"asterisk (*) metacharacter"/'
-
Building block of text
sed '
/^$!{
H
d
}
/^$/{
x
s/^\n/<p>/
s/$/<\/p>/
G
}'0
sed '
${
/^$//{ # It deals with the last line.
H
s/^.*$//
}
}
/^$/!{ #this part does not reach the bottom
H # hold the contents, appending to the existing contents
d
}
/^$/{ # This part reaches the bottom
x # It swaps the contents of the hold space and the pattern space
s/^\n/<p>/ # It matches the terminal new line, not the new line in the original contents
G
}'
- Branch and test
b: branch, t: test
[address]b[label] # The label is optional.
label: up to seven characters, starts with a new line with a colon
:mylabel #no space after the label!
b mylabel
Three branch loops:
1.
:top
command 1
command 2
/pattern/b top # It returns to the top
command 3 # It only executes command 3 if the pattern doesn't match.
2.
command 1
/pattern/b end # If it matches /pattern/, command2 will be skipped.
command 2
:end
command 3
3.
command 1
/pattern/b dothree # wither command 2 or command 3 will be executed
command 2
b
:dothree
command 3
eg:
sed '
/^\.ES/,/^\.EE/b # It branches when it meets the regex
s/^"/\'\'/
s/"$/\'\'/
s/"? /\'\'? /g
.
.
.
s/\\(em\\^"/\\(''/g
s/"\\(em/''\\(em/g
s/\\(em"/\\(em''/g'
-
[address]+[label]
eg, an extractor of index title
sed '
/Rh 0/{
s/"\(.*\)" "\(.*)" "\(.*\)"/"\1" "\2" "\3"/ #1
t # Label can be added to t.
s/"\(.*\)" "\(.*)"/"\1" "\2"/ #2
t
s/"\(.*\)"/"\1"/ #3
}'
# If 1 is true, then it goes to 2;
# if 2 is true, then it goes to 3.
sed '
:begin
/@f1(\([^)]8\))/{
s//\\fb\/\\fR/g # If it in the same line ??? How to substitue nothing to something?
b begin
}
/@f1(.*/{
N
s/@f1(\([^)]*\n[^)]*\))/\\fB/g
t again
b again
} # If it is in multiple lines
:again
P
D'
-
Join a phrase
:
# phrase -- search for words across lines
# $1 = search string: remaining args = filenames
search =$1 # It assign search to $1
shift # shell built-in varible
for file
do
sed '
/'"$search"'/b # see below
N # It appends the contents to the next regex.
h
s/.*\n// # It remove the previous lines.
/'"$search"'/b # Without a label, it passes to the bottom, same as the last b command.5
g # It gets the original contents from the hold space.
s/ *\n/ / # It converts multi-line to a single line.
g
b
}
g # It retrieves the duplicates, that preserves the newline, from the hold space. `
D' $files
done
/'"$search"'/b #
The single quotation prevents shell interpreting it;
double quotation makes sure that shell reads it first.
