Replace particular string at fixed position using sed - sed

I have a simple text file containing several lines of data where each line has got exactly 26 characters.
E.g.
10001340100491938001945591
10001340100491951002049591
10001340100462055002108507
10001340100492124002135591
10001340100492145002156507
10001340100472204002205591
Now, I want to replace 12th and 13th character if only these two characters are 49 with characters 58.
I tried it like this:
sed 's/^(.{12})49/\58/' data.txt
but am getting this error:
sed: -e expression #1, char 18: invalid reference \5 on `s' command's RHS
Thanks in advance

The captured group is \1, so you want to put the 11 (not 12) characters, then 58:
sed -E 's/^(.{11})49/\158/' data.txt
You also need -E or -r if you don't want to escape square and curly brackets.
With your input, the command changes 49 to 58 in lines 1, 2, 4 and 5.

If you like to try an awk solution:
awk 'substr($0,12,2)==49 {$0=substr($0,1,11)"58"substr($0,14)}1' file
10001340100581938001945591
10001340100581951002049591
10001340100462055002108507
10001340100582124002135591
10001340100582145002156507
10001340100472204002205591

In your expression you are replacing the first 14 characters (if you got it right). But you need to replace 11 plus the two (12th, 13th). More importantly, sed is fussy about escaping brackets, so you need backslashes in front of \( and { etc. Finally - the number of the capture group is 1. You omitted that number.
Putting it all together, you get
sed 's/^\(.\{11\}\)49/\158/'
There are flags you can use in sed to make the expressions more "regular" (changing the meaning of ( vs \(). What flag that is depends on your platform (version of sed), I believe.

Related

Replace variable num in double quotes - SED

I have a line like this:
"abc/x-y-z": "^1.4"
I need to replace ^1.4with * in the same file such that the output is "abc/x-y-z": "*"
The num inside the double quotes could be any variable number.
I tried this but it is highly specific to ^1.4 number:
sed -i '21s/^1.4/*/' abc.json
With your shown samples, please try following. You need to escape ^ here to make it literal character and you need to escape . dot as well to make it treat as literal character.
sed 's/\^1\.4/*/' Input_file
OR as per OP's comment to make it dynamic try:
sed 's/\^[0-9]+\.[0-9]+/*/' Input_file
Also if you are performing it on 21st line of your file then use 21s like you tried in your attempt. This code will substitute only very 1st occurrence of ^1.4 here in case you want to substitute all occurrences then use g(globally substitution) option for above code.
I have not used -i option(to do inplace update into Input_file itself) once you are happy with results then use sed -i option in above code.
When making changes to files, I prefer the file editor ed to the stream editor sed (ed is standard, sed -i isn't, and different versions have different quirks that bite people here on a regular basis).
ed -s input.txt <<EOF
21s/"^[[:digit:]]\{1,\}\(\.[[:digit:]]\{1,\}\)\{0,1\}"/"*"/
w
EOF
On line 21, matches a quote followed by a carat followed by 1 or more digits, optionally followed by a period and another sequence of 1 or more digits and finally the trailing quote character. All that is replaced by "*", and finally the changed file is written back to disk.
Posix BREs are a pain, no? GNU ed 1.17 and newer, and NetBSD ed can take EREs instead:
ed -Es input.txt <<EOF
21s/"\^[[:digit:]]+(\.[[:digit:]]+)?"/"*"/
w
EOF
which is a lot easier to read.

sed charachter to leave a match untouched

If I have
123456red100green
123456bee010yellow
123456usb110orange
123456sos011querty
123456let101bottle
and I want it to be
123456red111green
123456bee111yellow
123456usb111orange
123456sos111querty
123456let111bottle
notice: the first 6 characters don't change,,,,
the following 6 change,,,,
also these strings might be anywhere in a file (beginning, end, anywhere)
I want to specify sed to
1)find 123456
2)skip the next three characters
3)replace the next three with 111
The closest I've come to is:
sed '/s/123456....../123456...111/g'
I know dots mean anything but I don't know the equivalent on the other side. In short how to command sed to leave characters in a match untouched.
sorry for having been unclear of what I want please bear with me
Matching 123456 followed by three characters that are not to be modified, and then replacing the next three characters with 111:
sed 's/\(123456...\).../\1111/g' file
The \( ... \) captures the part of the string that we don't want to modify. These are re-inserted with \1. The whole matching bit of the line is replaced by "the bit in the \( ... \) (i.e. \1) followed by 111".
If you want to change each and every zero (as in your examples), then just sed 's/0/1/g' would do. Or sed -e '/^123456/ s/0/1/g' to do the same on lines starting with 123456.
But to count characters, as you ask, use ( .. ) to capture the varying parts and \1 to replace them (using sed -E). So:
echo 123456abcdefgh | sed -Ee 's/^(123456...).../\1111/'
outputs 123456abc111gh. The \1 puts back the part matched by 123456..., the next three ones are literal characters.
(Without -E, you'd need \( .. \) to group.)

SED command to remove words at the end of the string

I want to remove last 2 words in the string which is in a file.
I am using this command first to delete the last word. But I couldn't do it. can someone help me
sed 's/\w*$//' <file name>
my strings are like this
Input:
asbc/jahsf/jhdsflk/jsfh/ -0.001 (exam)
I want to remove both numerical value and the one in brackets.
Output:
asbc/jahsf/jhdsflk/jsfh/
Using GNU sed:
$ sed -r 's/([[:space:]]+[-+.()[:alnum:]]+){2}$//' file
asbc/jahsf/jhdsflk/jsfh/
How it works
[[:space:]]+ matches one or more spaces.
[-+.()[:alnum:]]+ matches the 'words' which are allowed to contain any number of plus or minus signs, periods, parens, or any alphanumeric characters.
Note that, when a period is inside square brackets, [.], it is just a period, not a wildcard: it does not need to be escaped.
([[:space:]]+[-+.()[:alnum:]]+) matches one or more spaces followed by a word.
([[:space:]]+[-+.()[:alnum:]]+){2}$ matches two words and the spaces which precede them.
Note the use of character classes like [:space:] and [:alnum:]. Unlike the old-fashioned classes like [a-zA-Z0-9], these classes are unicode safe.
OSX (BSD) sed
The above was tested on GNU sed. For BSD sed, try:
sed -E 's/([[:space:]][[:space:]]*[-+.()[:alnum:][:alnum:]]*){2}$//' file
To remove everything that follows a number with decimal places
This looks for a decimal number with optional sign and removes it, the spaces which precede it, and everything which follows it:
$ sed -r 's/[[:space:]]+[-+]?[[:digit:]]+[.][[:digit:]]+[[:space:]].*//' file
asbc/jahsf/jhdsflk/jsfh/
How it works:
[[:space:]]+ matches one or more spaces
[-+]? matches zero or one signs.
[[:digit:]]+ matches any number of digits.
[.] matches a decimal point (period).
[[:digit:]]+ matches one or more digits following the decimal point.
[[:space:]] matches a space following the number.
.* matches anything which follows.
It looks like there is a tab between what you want to keep and what you want to get rid of. I don't have linux in front of me but try this.
sed 's/\t.*//'
This is assuming your strings are always formatted similarily which is what I take from your comment.
This might work for you (GNU sed):
sed -r 's/\s+\S+\s+\S+\s*$//' file
or if you prefer:
sed -r 's/(\s+\S+){2}\s*$//' file
This matches and removes: one or more whitespaces followed by one or more non-whitespaces twice followed by zero or more whitespaces at the end of the line.

SED Command to remove first digits and spaces of each line

I have a simple text file in below format.
1 12658003Y
2 34345345N
3 34653785Y
4 36452342N
5 86747488Y
6 34634543Y
so on
10 37456338Y
11 33535555Y
12 37456378Y
so on
100 23432434Y
As you can see there are two white spaces after first number.
I'm trying to write SED command to remove the digits before whitespaces. Is there any SED command to remove spaces and number before spaces?
Output file should look like below.
12658003Y
34345345N
34653785Y
36452342N
so on..
Please assist. I'm very new to shell scripting.
sed 's/[0-9]\+\s\+//' infile > outfile
Explanation:
s: we want to use substitution
/: mark start and end of the expression we want to match
[0-9]: match any digit
+: match the previous one or more time
\s: space
+: match the previous one or more time
/: mark start of what we want to change our matches to (which is nothing)
/: some special operators goes after this (we use no such)
infile: the file we want to change
>: pipe stdout to
outfile: where we want to store output
Your sed command would be,
sed 's/.* //g' file
This would remove the first numbers along with the space followed.
Remove leading digits, then following spaces:
sed 's/^[0-9]* *//' file
sed 's/^[0-9]*[ ]*//g' input.txt

Matching strings even if they start with white spaces in SED

I'm having issues matching strings even if they start with any number of white spaces. It's been very little time since I started using regular expressions, so I need some help
Here is an example. I have a file (file.txt) that contains two lines
#String1='Test One'
String1='Test Two'
Im trying to change the value for the second line, without affecting line 1 so I used this
sed -i "s|String1=.*$|String1='Test Three'|g"
This changes the values for both lines. How can I make sed change only the value of the second string?
Thank you
With gnu sed, you match spaces using \s, while other sed implementations usually work with the [[:space:]] character class. So, pick one of these:
sed 's/^\s*AWord/AnotherWord/'
sed 's/^[[:space:]]*AWord/AnotherWord/'
Since you're using -i, I assume GNU sed. Either way, you probably shouldn't retype your word, as that introduces the chance of a typo. I'd go with:
sed -i "s/^\(\s*String1=\).*/\1'New Value'/" file
Move the \s* outside of the parens if you don't want to preserve the leading whitespace.
There are a couple of solutions you could use to go about your problem
If you want to ignore lines that begin with a comment character such as '#' you could use something like this:
sed -i "/^\s*#/! s|String1=.*$|String1='Test Three'|g" file.txt
which will only operate on lines that do not match the regular expression /.../! that begins ^ with optional whiltespace\s* followed by an octothorp #
The other option is to include the characters before 'String' as part of the substitution. Doing it this way means you'll need to capture \(...\) the group to include it in the output with \1
sed -i "s|^\(\s*\)String1=.*$|\1String1='Test Four'|g" file.txt
With GNU sed, try:
sed -i "s|^\s*String1=.*$|String1='Test Three'|" file
or
sed -i "/^\s*String1=/s/=.*/='Test Three'/" file
Using awk you could do:
awk '/String1/ && f++ {$2="Test Three"}1' FS=\' OFS=\' file
#String1='Test One'
String1='Test Three'
It will ignore first hits of string1 since f is not true.