Regular expression returns round brackets

[Eric]

I have a string that contains tokens surrounded by percentage signs,
e.g. %token1%

I want to return any instance of a token that is not surrounded by
single quotes. For example:

Using the input string: "This is %token1% and here is '%token2%'."
I want to match "%token1%" but not "%token2%".

My match pattern is: "[^\x27a-zA-Z0-9]%[^%]+%[^\x27a-zA-Z0-9]"

If my input string is: "This is %token1% and here is '%token2%' and
also (%token3%)."

I am getting back "%token1%" and "(%token3%)".

Why am I getting back round brackets around %token3%? How can I still
get back %token3%, but without the round brackets?

Thanks,
Eric

 

[Jon Skeet [C# MVP]]

I have a string that contains tokens surrounded by percentage signs,
e.g. %token1%

I want to return any instance of a token that is not surrounded by
single quotes. For example:

Using the input string: "This is %token1% and here is '%token2%'."
I want to match "%token1%" but not "%token2%".

My match pattern is: "[^\x27a-zA-Z0-9]%[^%]+%[^\x27a-zA-Z0-9]"

If my input string is: "This is %token1% and here is '%token2%' and
also (%token3%)."

I am getting back "%token1%" and "(%token3%)".

Why am I getting back round brackets around %token3%? How can I still
get back %token3%, but without the round brackets?

I suspect it's because of \x27a. Did you really mean Unicode U+027A,
or did you mean U+0027 followed by an a?

The \x escape sequence is a really, really bad one to use in almost
all situations. Especially in this case, as using a ' would be a lot
simpler to start with.

Jon

 

[Anthony Jones]

I have a string that contains tokens surrounded by percentage signs,
e.g. %token1%

I want to return any instance of a token that is not surrounded by
single quotes. For example:

Using the input string: "This is %token1% and here is '%token2%'."
I want to match "%token1%" but not "%token2%".

My match pattern is: "[^\x27a-zA-Z0-9]%[^%]+%[^\x27a-zA-Z0-9]"

If my input string is: "This is %token1% and here is '%token2%' and
also (%token3%)."

I am getting back "%token1%" and "(%token3%)".

Why am I getting back round brackets around %token3%? How can I still
get back %token3%, but without the round brackets?

I suspect it's because of \x27a. Did you really mean Unicode U+027A,
or did you mean U+0027 followed by an a?

The \x escape sequence is a really, really bad one to use in almost
all situations. Especially in this case, as using a ' would be a lot
simpler to start with.

Nope it has nothing to do with the \x27 but I agree there is no need to
escape it.

The pattern matches anything that isn't a ' or alphanumeric that preceeds a
% followed by anything isn't a % that is followed by a % and anything isn't
a ' or alphanumeric.

Therefore the spaces that preceed and follow the first token are included in
the match ( you have to look real careful to see that in the result) but in
the second ( ) also match because they're not ' or alphanumeric.

Try this instead:-

"(?<!')%(?!')[^%]+%(?!')"

 

[Jon Skeet [C# MVP]]

Nope it has nothing to do with the \x27 but I agree there is no need to
escape it.

The pattern matches anything that isn't a ' or alphanumeric that preceeds a
% followed by anything isn't a % that is followed by a % and anything isn't
a ' or alphanumeric.

No it doesn't. The escaping isn't just unnecessary, it's broken.

To see what I mean, compile and run this:

using System;
class Test
{
static void Main()
{
string x = "[^\x27a-zA-Z0-9]%[^%]+%[^\x27a-zA-Z0-9]";
Console.WriteLine(x);
}
}

You won't see a quote there at all, nor the 'a' of a-z.

The string is equivalent to

"[^\u027A-zA-Z0-9]%[^%]+%[^\u027A-zA-Z0-9]"

Definitely not what's intended. I dare say it's not responsible for the
brackets issue, but the regex isn't doing what it's meant to at the
moment.

 

[Eric]

I have a string that contains tokens surrounded by percentage signs,
e.g. %token1%
I want to return any instance of a token that is not surrounded by
single quotes.  For example:
Using the input string: "This is %token1% and here is '%token2%'."
I want to match "%token1%" but not "%token2%".
My match pattern is: "[^\x27a-zA-Z0-9]%[^%]+%[^\x27a-zA-Z0-9]"
If my input string is: "This is %token1% and here is '%token2%' and
also (%token3%)."
I am getting back "%token1%" and "(%token3%)".
Why am I getting back round brackets around %token3%?  How can I still
get back %token3%, but without the round brackets?

I suspect it's because of \x27a. Did you really mean Unicode U+027A,
or did you mean U+0027 followed by an a?

The \x escape sequence is a really, really bad one to use in almost
all situations. Especially in this case, as using a ' would be a lot
simpler to start with.

Nope it has nothing to do with the \x27 but I agree there is no need to
escape it.

The pattern matches anything that isn't a ' or alphanumeric that preceeds a
% followed by anything isn't a % that is followed by a % and anything isn't
a ' or alphanumeric.

Therefore the spaces that preceed and follow the first token are included in
the match ( you have to look real careful to see that in the result) but in
the second ( ) also match because they're not ' or alphanumeric.

Try this instead:-

"(?<!')%(?!')[^%]+%(?!')"

Thanks Anthony and Jon.

Yes, it makes sense that I shouldn't be unnecessarily escaping the
apostrophe.

Anthony, the pattern you provided above works great, thanks.

Eric

 

[Anthony Jones]

Nope it has nothing to do with the \x27 but I agree there is no need to
escape it.

The pattern matches anything that isn't a ' or alphanumeric that preceeds a
% followed by anything isn't a % that is followed by a % and anything isn't
a ' or alphanumeric.

No it doesn't. The escaping isn't just unnecessary, it's broken.

To see what I mean, compile and run this:

using System;
class Test
{
static void Main()
{
string x = "[^\x27a-zA-Z0-9]%[^%]+%[^\x27a-zA-Z0-9]";
Console.WriteLine(x);
}
}

You won't see a quote there at all, nor the 'a' of a-z.

The string is equivalent to

"[^\u027A-zA-Z0-9]%[^%]+%[^\u027A-zA-Z0-9]"

Based on the OPs described output it was clear that he wasn't getting an
error that RegExp would throw if that exact syntax was used. He was
probably using @"[^\x27a-zA-Z0-9]%[^%]+%[^\x27a-zA-Z0-9]" in which case the
RegExp parser would see the \x27 correctly.

 

[Jon Skeet [C# MVP]]

Based on the OPs described output it was clear that he wasn't getting an
error that RegExp would throw if that exact syntax was used. He was
probably using @"[^\x27a-zA-Z0-9]%[^%]+%[^\x27a-zA-Z0-9]" in which case the
RegExp parser would see the \x27 correctly.

Fair enough - although in that case it's just a bug waiting to occur
when some maintenance programmer decides to remove the @. It's also an
example of why it's a good idea to post complete, runnable code - it
removes sources of ambiguity like this.

I've been meaning to add the revolting \x escape (in C#, not regex) to
my brainteaser list. Something along the lines of:

// Character 9 = tab
Console.WriteLine("You say:\x9Good compiler!");
Console.WriteLine("You say:\x9Bad compiler!");

I wonder how many people would quickly spot that the second line will
actually try to output U+9BAD (which isn't in use as far as I can
tell).