This question already has answers here:
"sed" special characters handling
(3 answers)
Is it possible to escape regex metacharacters reliably with sed
(4 answers)
Escape a string for a sed replace pattern
(17 answers)
Closed 5 years ago.
I have a file called ethernet containing multiple lines. I have saved one of these lines as a variable called old_line. The contents of this variable looks like this:
SUBSYSTEM=="net", ACTION=="add", DRIVERS=="?*", ATTR{address}=="2r:11:89:89:9g:ah", ATTR{dev_id}=="0x0", ATTR{type}=="1", KERNEL=="eth*", NAME="eth1"
I have created a second variable called new_line that is similar to old_line but with some modifications in the text.
I want to substitute the contents of old_line with the contents of new_line using sed. So far I have the following, but it doesn't work:
sed -i "s/${old_line}/${new_line}/g" ethernet
You need to escape your oldline so that it contains no regex special characters, luckily this can be done with sed.
old_line=$(echo "${old_line}" | sed -e 's/[]$.*[\^]/\\&/g' )
sed -i -e "s/${old_line}/${new_line}/g" ethernet
Since ${old_line} contains many regex special metacharacters like *, ? etc therefore your sed is failing.
Use this awk command instead that uses no regex:
awk -v old="$old_line" -v new="$new_line" 'p=index($0, old) {
print substr($0, 1, p-1) new substr($0, p+length(old)) }' ethernet
Related
This question already has an answer here:
Perl regex with pipes
(1 answer)
Closed 2 years ago.
Here's a snippet of my code:
code=$(</dev/stdin)
prefix="$2"
code=$(perl -p -e "s/\"/\"$prefix/" <<< "$code");
It takes something like #include "myLib.h" and turns it into #include "(some_prefix)myLib.h", however, if I input something with a forward slash, like ./script.h -d lib/ < code.txt, I get an error:
syntax error at -e line 1, at EOF
Execution of -e aborted due to compilation errors.
Is there a way of escaping the backslash in the variable $prefix?
you could use pipes in your regex like in stackoverflow.com/questions/31830613/perl-regex-with-pipes
code=$(</dev/stdin)
prefix="$2"
code=$(perl -p -e "s|$prefix|" <<< "$code");
Since you make a regular expression out of user input, you need to escape it for regular expression. Merely changing the delimiter may be not enough, since some characters with special value in regular expression can be legitimately used in a filepath.
Therefore:
code=$(</dev/stdin)
prefix=$(printf %s "$2" | sed 's/[][()\.^$?*+/]/\\&/g')
code=$(perl -p -e "s/\"/\"$prefix/" <<< "$code");
The second line prepends \ before ][()\.^$?*+/.
UPD per #Wiktor Stribiżew; however you form the replacement string with user input, not the regex (as I imagined), so you need to replace there only the delimiter; and could just as well change the delimiter.
This question already has answers here:
How do I use grep to extract a specific field value from lines
(2 answers)
Closed 3 years ago.
I am trying to match a pattern and set that as a variable.
I have a file with many "value=key". I want to find the value for key "fizz".
In the file I have this string
fizz="something_cool"
I try to parse it as:
cat file | grep fizz="(.*)"
I was thinking it would give me the group output, and then I would be able to use $1 to select it.
I also play with escaping characters and sed and awk. But I could not manage to get it working.
You need to enable extended regex for using unescaped ( and ) and quote pattern properly to make it:
grep -E 'fizz="(.*)"' file
However awk might be better choice here since it will do both search and filter in same command.
You may just use:
awk -F= '$1 == "fizz" {gsub(/"/, "", $2); print $2}' file
something_cool
This question already has answers here:
Replace All Lines That Do Not Contain Matched String
(4 answers)
Closed 3 years ago.
I have a problem with making sed command, which gonna change lines, where =sometext= occurs and change it to another pattern, but will not do it when https occcurs in that line. I have no idea how I should change this command:sed -i 's/=\([^=]*\)=/{{\1}}/g'
You'll want to read the sed manual about matching lines: https://www.gnu.org/software/sed/manual/sed.html chapter 4:
The following command replaces the word ‘hello’ with ‘world’ only in lines not containing the word ‘apple’:
sed '/apple/!s/hello/world/' input.txt > output.txt
Use multiple blocks, e.g.:
sed '/=sometext=/ { /https/b; s/.../.../; }'
This question already has answers here:
Whether to escape ( and ) in regex using GNU sed
(4 answers)
Closed 4 years ago.
I need to insert '--' at the beginning of the line if line contains word VARCHAR(1000)
Sample of my file is:
TRIM(CAST("AP_RQ_MSG_TYPE_ID" AS NVARCHAR(1000))) AP_RQ_MSG_TYPE_ID,
TRIM(CAST("AP_RQ_PROCESSING_CD" AS NVARCHAR(1000)))
AP_RQ_PROCESSING_CD, TRIM(CAST("AP_RQ_ACQ_INST_ID" AS NVARCHAR(11)))
AP_RQ_ACQ_INST_ID, TRIM(CAST("AP_RQ_LOCAL_TXN_TIME" AS NVARCHAR(10)))
AP_RQ_LOCAL_TXN_TIME, TRIM(CAST("AP_RQ_LOCAL_TXN_DATE" AS
NVARCHAR(10))) AP_RQ_LOCAL_TXN_DATE, TRIM(CAST("AP_RQ_RETAILER" AS
NVARCHAR(11))) AP_RQ_RETAILER,
I used this command
sed 's/\(^.*VARCHAR\(1000\).*$\)/--\1/I' *.sql
But the result is not as expected.
Does anyone have idea what am I doing wrong?
this should do:
sed 's/.*VARCHAR(1000).*/--&/' file
The problem in your sed command is at the regex part. By default sed uses BRE, which means, the ( and ) (wrapping the 1000) are just literal brackets, you should not escape them, or you gave them special meaning: regex grouping.
The first and last (..) you have escaped, there you did right, if you want to reference it later by \1. so your problem is escape or not escape. :)
Use the following sed command:
sed '/VARCHAR(1000)/ s/.*/--\0/' *.sql
The s command appplies to all lines containing VARCHAR(1000). It then replaces the whole line .* by itself \0 with -- in front.
Through awk,
awk '/VARCHAR\(1000\)/ {sub (/^/,"--")}1' infile > outfile
This question already has answers here:
Getting values between quotes
(2 answers)
Closed 9 years ago.
I have text lines like this
blahblah"word1"blahblah"word2"blahblah"word3"
I only want the text between the quotes and without the quotes. I could do an awk and us the " as a separator. And then get every second match. However, is there any way I can just use awk (or another command) to return words between sets of quotes? so I'd get back word1, word2, word3?
Thanks,
Here you go:
echo 'blahblah"word1"blahblah"word2"blahblah"word3"' | perl -ne 'print map("$_\n", m/"([^"]*)"/g)'
Depends which language you're using, but the regular expression to do this would be:
(?<=^(("[^"]*){2})*")[^"]+(?=")
That example will match everything between "s. if you want it to match only words between "s, use:
(?<=^(("[^"]*){2})*")\b+(?=")
The main difference is with the second example, spaces and most special characters will not be allowed. With the first example, all characters except for "s will be allowed between the "s. That includes new lines.
Non-robust, but fun:
sed -E 's/(^|")[^"]*("|$)/ /g'