regex question (Asher?)

[Kuciwalker]

I have a piece of a sed script that is working correctly and I'm not sure why. This bothers me.

The replacement rule is supposed to change a function of the form:

log(foo,fmt,arg1,arg2)

into:

log(foo,fmt,(cast)arg1,(cast)arg2)

This is my rule:

s/\(foo.*,.*\),\(.*\),\(.*\)/\1,(cast)\2,(cast)\3/
              ^

If I execute:

echo 'log(foo,"omg,wtf",arg1,arg2)' | sed 's/\(foo.*,.*\),\(.*\),\(.*\)/\1,(cast)\2,(cast)\3/'
                       ^

I get:

log(foo,"omg,wtf",(cast)arg1,(cast)arg2)

I don't understand why the marked comma in the rule is matching the marked comma in the input, rather than the preceding comma. Does sed just try to find the longest possible match or something?

[onodera]

Yes, yes it does. Use lazy quantifiers, like *?

[Kuciwalker]

I'm perfectly happy with it behaving like this; I just wanted to make sure it would behave this way consistently before I ran the script on ~1MB of source code.

[VetLegion]

I don't understand why the marked comma in the rule is matching the marked comma in the input, rather than the preceding comma.

.* means "any character, one or more number of times"

Comma is a character, so foo.* matches foo,

[SlowwHand]

"I have a piece of a sed script that is working correctly and I'm not sure why. This bothers me."



To a layman like myself, that was funny.

[Kuciwalker]

I don't understand why the marked comma in the rule is matching the marked comma in the input, rather than the preceding comma.

.* means "any character, one or more number of times"

Comma is a character, so foo.* matches foo,

Has it occurred to you that I couldn't have possibly written the OP without understanding that?

My question was not 'why does this match' because I know that it matches; my question was 'why did it pick this particular match, given that the answer is ambiguous, and will it work this way consistently'.

[VetLegion]

What is ambiguous?

[VetLegion]

Sorry, I just skimmed your question. Onodera is probably right.

[Kuciwalker]

What is ambiguous?

Not fully specified.

e.g. given the string 'foo', '.*' matches '', 'f', 'o', 'fo', 'oo', and 'foo', thus it is ambiguous.

[Thue]

Of course, the regexp will misbehave if arg1 or arg2 contains a comma.

[Kuciwalker]

Yes. Thankfully, I'm pretty sure that's never the case.

[aneeshm]

By default, if there are multiple matches possible for a regex, it matches the longest one.

[Kuciwalker]

It would have been simpler and faster to quote my hypothesis and answer "yes"