Perl Regex匹配未用引号括起来的字符串

Question

I'm trying to write regex to match a string that is NOT inside quotes (double or single), but the best I can do so far is a loop to iterate through all characters of a string. There must be a simpler, more elegant solution.

Examples: If trying to replace foo with bar the string hello foo! would become hello bar! , but the string you said "my name is foo" would stay the same.

Could anyone help out with a regexp to achieve the above?

Answer 1

One way, using a negative lookahead:

perl -lane 's/foo(?![^"]*"(?:[^"]*"[^"]*")*[^"]*$)/bar/g; print' input

which means substitute if the number of quotes ahead is not odd. So this one assumes you have balanced quotes in the input.

Example input:

hello foo!
"foo" foo "foo"
foo "hello" foo
"foo" bar

Example output:

hello bar!
"foo" bar "foo"
bar "hello" bar
"foo" bar

Answer 2

Update: Quick-Summary: While you'd need "Balancing Groups" to really handle this, the short answer is, you can't do it if you require Single-Quotes too. Because those double as Apostrophes. So no matter what, this would really screw you up: That's when foo said, "That's my line!" balancing gets very thrown out of whack with apostrophes. You'll need to build a custom parsing engine.

Note: If this is for HTML Properties... I have written a Regex that properly parses them exactly as you say, and I believe would work in Perl. But that also relies on delimiters like the = sign and other HTML Structures. But in 90% of those cases, an XML/HTML Parser is the best option (That 10% is still a possibility).

As I mentioned in my comment to your question, more examples would give more concrete answers. This is the answer for your limited example:

^([^"']*?)foo([^"']*)$

Lookarounds are easy for an intermediate regex writer, but complicated for code maintainance and usually not what was needed. Also, anything that requires you to use a Dot . in regex is typically not as efficient as it could be.

Replace my example with $1bar$2 and you'll be golden. But again, as my comment says, this is based on your basic example which assumes your entire string is that which may start with and end with quotes. If you have different examples they would help.

Addition

Just for fun, I'm going to answer your question for two other options. Option 1 is my original answer above.

Option 2 (As mentioned by Floris):

Hi foo, I said "hello"

Or

"hello", said foo to his friend.

If this is the case, where quoted text will only appear BEFORE or AFTER your search-text (foo in this case), then the answer is like so:

^(?:([^"']*?)foo(.*)|(.*?)foo([^"']*))$

Option 3 (as seen in my comment below)

He said, "Hello", so then Foo told him, "Lawl, bye"

To do this, we'd have to count the number of Quotes before and after foo, to be sure they are either even, or that they "close out" known as "Balancing" in .NET Regex, neither of which options are available in your circumstance without some other custom functions.

Answer 3

Needed to do this as well, so solved it myself... This solution does not rely on balanced quotes, but obviously will not support apostrophes if they come in pairs.

#!/usr/bin/perl

my @test = ( 'hello foo!',
             '"my name is foo"',
             'foo test "test foo test" test foo test "test foo test" test foo',
             "foo test 'test foo test' test foo test 'test foo test' test foo",
             '"foo test foo"',
             'foo test " foo test' );

foreach ( @test )
{
  s!("[^"]*"|'[^']*')|foo!$1//'bar'!ge;
  print "$_\n";
}