Search-replace in code

fuzzyBuzz · 10-11-2007, 06:30 AM

Hi,

I have code that currently uses the function:
void getMyTime(MyTimeStruct*);
and i would like to replace all calls to this function to calls to a different function with a prototype:
MyTimeStruct newGetMyTime(void);

For example, i would like the following line:
getMyTime(&startTime);
to be changed to:
startTime = newGetMyTime();

Is there any easy automated way to do this?

Centinul · 10-11-2007, 06:39 AM

You could try using something like SED on each of your files that has that function in it. Your request seems simple enough that beginner SED tutorials on Google should get you where you need to be. If I remembered the syntax I would write it out.

HTH,

Centinul

GrapefruiTgirl · 10-11-2007, 06:39 AM

You can do this in one statement using some combination of cat, grep & sed, with sed being the main ingredient (and sed probably can do it all by itself, without grep or cat).

Check out the man page for sed.

matthewg42 · 10-11-2007, 08:12 AM

If you're doing a large global replace in a lot of files, be careful - what if the string you are replacing is a sub-string of another function name which you do not want to modify? Check first.

Perl has some nice regular expression extensions which can help, e.g. the \b pattern (word boundary), and you can user Perl on the command like much like sed and friends., e.g.

Code:

perl -i.original -npe 's/\bmain\b/MAIN/g' *.c

...will replace all instances of the whole word "main" with "MAIN" in all files whose name ends in ".c", saving the original version of the file with the name suffix ".original" (e.g. test.c becomes test.c.original).

matthewg42 · 10-11-2007, 08:24 AM

Oh, for removing the stuff between the brackets, you need to add a little something. Don't forget to escape the brackets with a \ character since they are special characters in regular expressions, i.e. the Perl code is like this:

Code:

perl -i.original -npe 's/\bgetMyTime\s*\([^)]*\)/newGetMyTime()/g;' *.c

You can look in the perlrun manual page to see what the -i, -n, -p, and -e options do. The Perl program (the bit in the single quotes) consists of a single substitution operation. The syntax is

Code:

s/replacethis/withthis/flags;

Where:

replacethis is a Perl-style regular expression pattern to search for (see perlre manual page),
withthis is what to replace found-patterns with.
flags is a list of single character options which modify the behaviour of operation. "g" means global - replace all instances of the pattern on a given input line instead of just the first match, which is the default behaviour.

A more detailed description of the substitution operator can be found in the perlop manual page.

angrybanana · 10-11-2007, 10:56 PM

This command should work

Code:

$ cat a
getMyTime(&startTime);

$ sed 's/\(.*\)(&\(.*\))/\2 = new\1()/' a
startTime = newgetMyTime();

fuzzyBuzz · 10-12-2007, 02:14 AM

Quote:

Originally Posted by angrybanana

This command should work

Code:

$ cat a
getMyTime(&startTime);

$ sed 's/\(.*\)(&\(.*\))/\2 = new\1()/' a
startTime = newgetMyTime();

This exactly what i had in mind. Thanks.

fuzzyBuzz · 10-12-2007, 02:56 AM

Based on angrybanana's sed command i created a substitution that only affects getMyTime function.

Code:

$ cat a
getMyTime(&startTime);
setMyTime(&startTime);
$ cat s
s/getMyTime(&\(.*\))/\1 = newGetMyTime()/
$ sed -f s a
startTime = newGetMyTime();
setMyTime(&startTime);

I had to put the command in a script file because i am running under cygwin and the ampersand character (&) in the regular expression is confusing the shell. Any way to use ampersand in regular expressions in the shell?

angrybanana · 10-12-2007, 08:09 AM

Quote:

Originally Posted by zvirack

I had to put the command in a script file because i am running under cygwin and the ampersand character (&) in the regular expression is confusing the shell. Any way to use ampersand in regular expressions in the shell?

Dunno much about cygwin, but you could try double quotes (") or backslash (\).

fuzzyBuzz · 10-13-2007, 05:02 PM

Quote:

Originally Posted by angrybanana

Dunno much about cygwin, but you could try double quotes (") or backslash (\).

Tried that but didn't work for me.

ntubski · 10-14-2007, 01:15 PM

Use backslash and double quotes, or single quotes:

Code:

$ cat a.c
getMyTime(&startTime);
setMyTime(&startTime);

$ sed 's/getMyTime(&\(.*\))/\1 = newGetMyTime()/' a.c
startTime = newGetMyTime();
setMyTime(&startTime);

$ sed "s/getMyTime(\&\\(.*\\))/\\1 = newGetMyTime()/" a.c
startTime = newGetMyTime();
setMyTime(&startTime);

fuzzyBuzz · 10-15-2007, 01:12 AM

Quote:

Originally Posted by ntubski

Use backslash and double quotes, or single quotes:

Code:

$ cat a.c
getMyTime(&startTime);
setMyTime(&startTime);

$ sed 's/getMyTime(&\(.*\))/\1 = newGetMyTime()/' a.c
startTime = newGetMyTime();
setMyTime(&startTime);

This form doesn't work in cygwin. I get the following error message:

Code:

sed: -e expression #1, char 12: unterminated `s' command
'\' is not recognized as an internal or external command,
operable program or batch file.

Quote:

Originally Posted by ntubski

Code:

$ sed "s/getMyTime(\&\\(.*\\))/\\1 = newGetMyTime()/" a.c
startTime = newGetMyTime();
setMyTime(&startTime);

This form works in cygwin! Thanks for the tip.

I was wondering if there could be a more general regular expression that would substitute for all possible c forms of the statement. For example, include also the following cases:

getMyTime (&startTime); // Spaces between function name and parenthesis
getMyTime(
&startTime
)
; // Statement broken into several lines. This is a legal c statement.

ntubski · 10-15-2007, 10:37 AM

Quote:

Originally Posted by zvirack

This form doesn't work in cygwin. I get the following error message:

Maybe you miscopied the command? I actually tested both commands in cygwin before posting.

Quote:

Originally Posted by zvirack

I was wondering if there could be a more general regular expression that would substitute for all possible c forms of the statement. For example, include also the following cases:

1. getMyTime (&startTime); // Spaces between function name and parenthesis
2. getMyTime(
&startTime
)
; // Statement broken into several lines. This is a legal c statement.

sed is a line based tool, it might be possible to do this with pattern spaces or something but I'm not sure. Just spaces is easy though:

Code:

$ sed 's/getMyTime[ ]*([ ]*&[ ]*\(.*\)[ ]*)/\1 = newGetMyTime()/' a.c
startTime = newGetMyTime(); // Spaces between function name and parenthesis
getMyTime(
&startTime
)
; // Statement broken into several lines. This is a legal c statement. Not changed by the sed regexp though :(