[SOLVED] Replace string in several files at position

ceantuco · 05-01-2017, 11:33 AM

I tried it but it did not replace anything.

I tried the command below:

Code:

sed -i 's/\(  ACCT: \)[^ ]*/\1XXXXXXXXXXXX/' filename

and the string changed from:

Code:

  ACCT: 3478913473413536

to:

Code:

  ACCT: XXXXXXXXXXXX

Now, how could I modify the code so it preserves the last 4 digits/characters?

Thank you!

MadeInGermany · 05-01-2017, 11:55 AM

The following cuts ACCT: surrounded by space, and puts it back via a back-reference.
The following 12 digits are substituted by 12 X

Code:

sed 's/^\([[:space:]]*ACCT:[[:space:]]*\)[0-9]\{12\}/\1XXXXXXXXXXXX/' filename

ceantuco · 05-01-2017, 11:58 AM

After playing with the command above, I tried the command below:

Code:

sed -i 's/\(  ACCT: \)[^0 ]*/\1XXXXXXXXXXXX/' filename

and the string changed from:

Code:

  ACCT: 3478913473413536

to:

Code:

  ACCT: XXXXXXXXXXXX413536

I tried replacing the

Code:

^0

with numbers from 1-10; however, it does not produce the desired output of:

Code:

  ACCT: XXXXXXXXXXXX3536

I think I am close to the solution but I don't know what else to try/change.

Thank you

ceantuco · 05-01-2017, 12:00 PM

MadeinGermany,

I posted the response above before I saw your post. I will test it and provide update.

Thank you!

Turbocapitalist · 05-01-2017, 12:16 PM

It is also important that you get to the point where you can build or extend regular expressions on your own. One, it will save you effort. Two, you encounter them everywhere these days.

When you have the working solution, which I think you will in the next few minutes, try looking at the manual page already mentioned and working through this tutorial: http://www.regular-expressions.info/quickstart.html to analyze the components of the pattern used.

ceantuco · 05-01-2017, 01:04 PM

**CORRECTION**

MADEINGERMANY, Thank you so much! the solution you provided worked like a charm! yes, there are other expressions with less characters in the string that I might need to also modify so I will play with the solution you provided to change those expressions.

Since I have thousands of files I need to modify and they are located in multiple sub folders within a folder, should I run the code as shown below?

Code:

sed -i -r 's/^\([[:space:]]*ACCT:[[:space:]]*\)[0-9]\{12\}/\1XXXXXXXXXXXX/' /home/user/folder/*

Thank you

sundialsvcs · 05-01-2017, 01:13 PM

Quote:

Originally Posted by pan64

Code:

/^ACCT:/   look for lines beginning with ACCT:

... and, to illustrate just how subtle regular-expressions ("regexes") can be, it is the "^" character which says, "anchor to start-of-line." (Similarly, "$" anchors to end-of-line.)

There are many regex tester web-sites out there which are frankly worth their weight in gold. They let you type in a regex, and a test string, and they will actually show you the logic that is applied to obtain a "match" or "doesn't match" result. They will also point out any syntax-errors in your regex. "Don't leave home without it.™"

ceantuco · 05-01-2017, 01:17 PM

Thank you!

sundialsvcs · 05-01-2017, 01:20 PM

Quote:

Originally Posted by ceantuco

Since I have thousands of files I need to modify and they are located in multiple sub folders within a folder, should I run the code as shown below?

Personally, I use the find command to produce a list of the files that it considers to be "matches." Then, I e-y-e-b-a-l-l that list ... several times. (I'm only gonna get one chance at this ...)

Finally, I apply that list, e.g.:

cat listfile | xargs -I{} mycommand "{}"

(Notice the use of "quotes" in the substituted command, to handle the case of a path-string that contains spaces.)

Sometimes I'll start by executing [font=courier]echo commands that echo the command-strings I am about to execute. Sometimes, I'll capture that echo-output into a file, give it one... last... look..., then execute that file.

That degree of caution has saved my left foot many times.

Turbocapitalist · 05-01-2017, 01:39 PM

Also, with the -i option for GNU sed you can specify a backup file to be made, named with the given suffix. See

Code:

man sed

ceantuco · 05-01-2017, 01:55 PM

Will do! Thank you all for your responses!

ceantuco · 05-01-2017, 01:57 PM

MadeinGermany, Your solution is the one that worked.
Thank you!

Turbocapitalist · 05-01-2017, 02:05 PM

Be sure to look up what options -i and -r do. The parenthesis and braces have to be written normally if you use -r. And -i usually takes some string with it.

Code:

sed -r 's/^([[:space:]]*ACCT:[[:space:]]*)[0-9]{12}/\1XXXXXXXXXXXX/' /home/user/folder/*

Also check the references mentioned for explanations of all the components in the regular expression:

^, ( ), [ ] as in [0-9], [:space:], *, and {12}

In the replacement lookup \1

Same for the command s///

ceantuco · 05-01-2017, 03:53 PM

Thank you! I was able to modify the command to be able to change other strings in the same file which contain less/more characters and they all worked okay. I will def read up on 'sed' and the references mentioned above. I will also do more testing before changing all files. Thank you again!