[SOLVED] Replace text lines identified by leading text in line within multiple files

juergen852 · 09-12-2014, 04:35 PM

Hello,

I have some 100 files with different entries like

File1: text V2.0 at line 22
File2: text V2.20 at line 14
File3: text V2.042 at line 29

--> at different line positions within those files.

I want to replace those different "vers">V2.xxx numbers with a unique entry like "vers">V2.2

I am well aware of sed -i.bak 's/foo/bar/g' * but this does not work as I need to replace the line, not only some text.

Any Ideas?

keefaz · 09-12-2014, 05:05 PM

Quote:

Originally Posted by juergen852

I am well aware of sed -i.bak 's/foo/bar/g' * but this does not work as I need to replace the line, not only some text.

I don't get it, you want to replace the lines or the version number?

juergen852 · 09-12-2014, 05:22 PM

Problem: Some 100 .html-files which all contain some version information. This version information is not always the same number yet. So I want to replace former version information within those files with a new (same) one for all files.

Well, I cannot do a sed s/V2.0/V2.3/g' as the search argument is not always V2.0. Some of the files contain V2.1 some contain V2.03 or so.
And I want to replace that version number within all files with the same number.

So my idea was:
Search all files for the line that starts with , and replace the former content of the whole text line with some new content.

astrogeek · 09-12-2014, 05:27 PM

Quote:

Originally Posted by juergen852

Well, I cannot do a sed s/V2.0/V2.3/g' as the search argument is not always V2.0. Some of the files contain V2.1 some contain V2.03 or so.
And I want to replace that version number within all files with the same number.

So my idea was:
Search all files for the line that starts with , and replace the former content of the whole text line with some new content.

Why do you think you cannot do that with sed?

Try this (test on an example file first - this is untested!):

Code:

sed -i 's/"vers">V2.[0-9]*/"vers">V2.2/g' filename

You can adjust the amount of the line that you match and replace as actually needed, this example is based on your own.

syg00 · 09-12-2014, 05:38 PM

I've found using "*" like that can lead to unexpected matches (on zero repetitions) - better to use "+" if the data are known to always be composed of at least one digit.

juergen852 · 09-12-2014, 05:40 PM

.... Maybe I'm just too old
Great Idea. Thanks.

I will try tomorrow and update here.

juergen852 · 09-21-2014, 01:45 PM

OK, this helped me out, although I had to use the command 10 times with different Search-Terms.
There was "2.1.0" or "Beta 2.2" and so on who are not covered by a "V2.[0-9]*"

So back to my original Idea: Is there a way to replace a complete line of text with a new line of text if the search term ( <span class="vers") matches the text in the line at any position?

astrogeek · 09-21-2014, 01:56 PM

Quote:

Originally Posted by juergen852

OK, this helped me out, although I had to use the command 10 times with different Search-Terms.
There was "2.1.0" or "Beta 2.2" and so on who are not covered by a "V2.[0-9]*"

So back to my original Idea: Is there a way to replace a complete line of text with a new line of text if the search term ( <span class="vers") matches the text in the line at any position?

How about just adapting the sed expression to handle those cases...

Code:

sed -i 's/"vers">V2.[0-9.]*/"vers">V2.2/g' filename

OR

sed -i 's/"vers">V2[^<]*/"vers">V2.2/g' filename

OR

sed -i 's/"vers">[^<]*/"vers">V2.2/g' filename

Your original examples did not include any versions with multiple decimals so the sed we gave did not handle them. Now it will.

The last one will cover absolutely anything between the angle-brackets (your Beta case).

*** EDIT ***

By the way, I have not intended to ignore your question about replacing the whole line, but you are always way ahead by being as specific as possible in doing replacements across multiple files. Replacing whole lines can easily corrupt the files if it encounters any number of special cases not anticipated by the expression. So always best to identify the range of cases then craft an expression that handles them.

juergen852 · 09-21-2014, 04:21 PM

sed -i 's/"vers">[^<]*/"vers">V2.2/g' filename solved my problem.

(I still wonder whether sed or awk may to the complete line replacement. Just in case I hit a problem which cannot be solved by expressions.)

Big Thanks to everybody who helped me.

astrogeek · 09-21-2014, 04:54 PM

Quote:

Originally Posted by juergen852

sed -i 's/"vers">[^<]*/"vers">V2.2/g' filename solved my problem.

(I still wonder whether sed or awk may to the complete line replacement. Just in case I hit a problem which cannot be solved by expressions.)

Big Thanks to everybody who helped me.

You are welcome!

The following one (untested, but should be OK) is one way to do the full line...

Code:

sed -i 's/^<span class="vers">.*$/<span class="vers">V2.2<\/span>/' filename

... but I can think of several surprises that you might hit such as if there were any cases of additional text on a line you would miss the line or lose the additional text, leading spaces would cause a line to be missed, line-breaking whitespace would break it, etc...

Anyway, glad to help, good luck!