extracting multiple substrings from string using sed

Question

I'm working on a cocos2dx project and recently we updated to the latest version of cocos2dx, this introduced a number of warnings which I'm cleaning up.

I have a lot of code like:

CCPoint somePoint = ccpAdd(this->getPosition(), _someRandomOffset);

The method ccpAdd is deprecated, in favour of the + operator, I want to replace such instances. I've tried searching on google, but I can't find out how to extract two strings using sed and build them back together.

CCPoint somePoint = this->getPosition() + _someRandomOffset;

My question is, how can I automate this replacement using some script against my source files?

Bonus points if the sed command can handle nested ccpAdd commands, like:

CCPoint somePoint = ccpAdd(this->getPosition(), ccpAdd(one, two));

Or maybe sed is the wrong tool for the job?

did you write your codes in some IDE? it would be safer to use its refactory feature. e.g your codes could be `...=ccpAdd(..);another statement without newline;` or `...=ccpAdd(func(func2(a,b),x),ccpAdd(foo,bar));` — Kent, Sep 12 '13 at 07:58

score 1 · Accepted Answer · edited May 23 '17 at 12:20

I think

sed 's/ccpAdd(\(.*\),\(.*\))/\1+\2/g'

does the trick.

This will however not work for nested occurrences, and will produce strange results for multiple occurrences of the pattern on a single line.

Unfortunately, sed doesn't have non-greedy operators, so the second problem must be solved by switching to another tool, say perl:

perl -pe 's|ccpAdd\((.*?),(.*?)\)|\1 + \2|g'

To get the nesting right, you can just re-run the same perl command several times until there are no more matches (this works because of the non-greedy operators).

score 1 · Answer 2 · answered Sep 12 '13 at 09:11

An altenative in python using the module pyparsing with recursivity:

Assuming content of infile as:

CCPoint somePoint = ccpAdd(this->getPosition(), _someRandomOffset);
CCPoint somePoint = ccpAdd(this->getPosition(), ccpAdd(one, two));

And script.py as:

from pyparsing import *

parser = Forward()
parser << Literal('ccpAdd').suppress() \
    + Literal('(').suppress() \
    + ( parser | OneOrMore(CharsNotIn(',')) ) \
    + Literal(',').suppress() \
    + ( parser | OneOrMore(CharsNotIn(',)')) ) \
    + Literal(')').suppress()

parser.setParseAction(lambda t: ' + '.join(t))

with open('infile', 'r') as f:
    for line in f:
        r = parser.transformString(line)
        print(r, end='')

Run it like:

python3 script.py

That yields:

CCPoint somePoint = this->getPosition() +  _someRandomOffset;
CCPoint somePoint = this->getPosition() + one +  two;

@jonynz: Glad to read that, and thank you for the confirmation. I'm not used to this kind of parsing and it's fine to see that can be useful for someone. — Birei, Sep 12 '13 at 10:55

Kokozaurus · Answer 3 · 2013-09-12T08:11:03.520

-1

What I would do is to use

sed 's/regex to match the old string/new string/'

and then write it to a file. so it would be something like this:

sed 's/regex to match the old string/new string/' > file

and ofcourse to provide sed with input just redirect the standard output of cat:

cat thefile.txt | sed ...

edited Sep 12 '13 at 08:11

answered Sep 12 '13 at 07:58

Kokozaurus

639
7
22

I would suggest you remove this answer before you get downvotes... it doesn't answer the question. – Kent Sep 12 '13 at 08:03
Yeah, many cases of ccpAdd use different variables, like ccpAdd(layer->getPosition(), offset), and ccpAdd(center, offVect), so it isn't generic enough – jonynz Sep 12 '13 at 08:04
i mean his answer isn't generic, but regex, yeah that will work, but I don't understand how to pull out the strings 1 and 2 (separated by the comma) – jonynz Sep 12 '13 at 08:05
1

The "Grouping and back-references in sed" section here: http://www.thegeekstuff.com/2009/10/unix-sed-tutorial-advanced-sed-substitution-examples/ would help you for that. – Kokozaurus Sep 12 '13 at 08:26

extracting multiple substrings from string using sed

3 Answers3