Search multiple lines in sed or vi

vijay_babu1981 · 10-29-2009, 04:28 AM

Hi experts,

I have following code:

Code:

for (int i=0; i<DimSum ;i++)
{
   a[i] = new double [10];
}
for (int j=0; j<DimSum ;j++)
{
   b[j] = x+y;
}

I want to search all such for loops where a "new" is present. I want to replace DimSum from all such loops with SumDim.

To search, I tried in vi (or in sed):

Code:

/for.*DimSum.*\n.*\n.*new

This works for the given search. But if the code is

Code:

for (int i=0; i<DimSum ;i++)
{
   x = junk;//multiple lines before new
   a[i] = new double [10];
}

then I need to search for any number of new line characters.

basically, I want ".*\n.*" to be searched any number of times until a "new" is found (thus I want to have a * after it) or a "}" is encountered . But I dont know how to group "\n.*" for the following * to take effect and to stop search if "}" is found.

Can anyone help me with this mess?

Regards

Thanks in Advance!

ghostdog74 · 10-29-2009, 04:36 AM

Code:

$ awk 'BEGIN{RS="}"}/new/{ gsub("DimSum","SumSum")}{print $0 RT}' file
for (int i=0; i<SumSum ;i++)
{
   a[i] = new double [10];
}

for (int j=0; j<DimSum ;j++)
{
   b[j] = x+y;
}

vijay_babu1981 · 10-29-2009, 04:45 AM

Hi ghostdog74

Thanks for the quick response. I am trying this out.

BTW, Can this be done in sed or vi also? If not then I think I should start learning awk!

Regards

chrism01 · 10-29-2009, 08:53 PM

Why do you only want to replace DimSum with SumDim in those loops? Surely you need to replace it everywhere, or your code won't work?

vijay_babu1981 · 10-30-2009, 12:35 AM

Quote:

Originally Posted by chrism01

Why do you only want to replace DimSum with SumDim in those loops? Surely you need to replace it everywhere, or your code won't work?

I want to change only those for loops which allocate memory using "new".
Also which de-allocates using delete.

I want to search all for loops wherever new/delete is encountered and change the for loop.

Regards

David the H. · 10-30-2009, 12:55 AM

Multi-line actions can be done in sed, but in general it's not particularly easy to do so. It usually involves adding multiple lines to the search buffer first, and maybe using the hold buffer as well.

Some useful links for sed and awk:

http://www.grymoire.com/Unix/Sed.html
http://sed.sourceforge.net/sedfaq.html

http://www.grymoire.com/Unix/Awk.html
http://www.gnu.org/manual/gawk/html_node/index.html

vijay_babu1981 · 10-30-2009, 01:09 AM

Quote:

Originally Posted by David the H.

Multi-line actions can be done in sed, but in general it's not particularly easy to do so. It usually involves adding multiple lines to the search buffer first, and maybe using the hold buffer as well.

Some useful links for sed and awk:

http://www.grymoire.com/Unix/Sed.html
http://sed.sourceforge.net/sedfaq.html

http://www.grymoire.com/Unix/Awk.html
http://www.gnu.org/manual/gawk/html_node/index.html

I have read http://www.grymoire.com/Unix/Sed.html already.

I am able to read/search multiple lines using '\n' in my search pattern.

I am able to search and replace as well using back-referencing.

My only trouble is that sometimes there can be multiple and unknown number of '\n' in my search pattern.

Example1:

for.*DimSum.*\n.*new

will search
for(int i; i<DimSum; i++){
junk = new double;

Example2:
for.*DimSum.*\n.*\n.*new

will search
for(int i; i<DimSum; i++)
{
junk = new double;

But what to do when number of '\n' is not known?

ghostdog74 · 10-30-2009, 01:58 AM

Quote:

Originally Posted by vijay_babu1981

If not then I think I should start learning awk!

Regards

forget about sed or vi already.. start learning awk now...Or Python/Perl for more complicated tasks...!!

David the H. · 10-30-2009, 11:02 AM

Use the right tool for the right job.

Sed is a line-based editing program. It mostly works on one line at a time. And while it can be used for multi-line editing, it's not truly suited for it. It's best for simple substitutions and extracting strings from text when the matching needs are not too complicated.

Awk is a field-based editor. It defaults to one word-one field, but you can redefine the fields as necessary, so you can also use it to work on lines, paragraphs, individual characters or whatever. It's also includes a more complex language syntax that makes it more flexible than sed for many uses. But it's perhaps still not always the best; its not the most user-friendly program, and awk scripts sometimes have to be rather long and complex.

I haven't really used perl much yet, but my impression is that it excels in text editing in general, so it's probably the best suited of all for complex matching needs like this.