regex challenge

[Jarlaxle]

I'd like to issue a challenge (I already have a class to do it so I'm not
asking just to get the code)...

Write one regex expression that removes all comments from a string.

Hints:

1. /* a block can contain newlines and must match till next */
2. // must match until end of line
3. I have seen some that claim to remove all comments but are inadequate for
one reason...(QUOTES!)

 

[Jon Skeet [C# MVP]]

I'd like to issue a challenge (I already have a class to do it so I'm not
asking just to get the code)...

Write one regex expression that removes all comments from a string.

Hints:

1. /* a block can contain newlines and must match till next */
2. // must match until end of line
3. I have seen some that claim to remove all comments but are inadequate for
one reason...(QUOTES!)

Well, you need to specify the quote behaviour as well. Hint: C# quote
handling would be different to Java handling. (In fact, that goes for
various other non-quote cases, given the way Java handles Unicode
escape sequences.)

 

[Jarlaxle]

we can keep it to c#:

1. /* a block can contain newlines and must match till next */
2. // must match until end of line
3. quotes must be supported
a. all text from a starting quote to closing quote must not be
matched
b. must support \" escape sequence inside quotes.

 

[Paul E Collins]

I'd like to issue a challenge (I already have a class to do it so I'm
not asking just to get the code)... Write one regex expression that
removes all comments from a string.

How does your class handle code like this?

/* string s = "/* hello */ // how are you?"; */ string s = "/* // */";

I'm pretty sure it's impossible with a regular expression; you'd need a
true C# parser.

Eq.

 

[Paul E Collins]

How does your class handle code like this?
/* string s = "/* hello */ // how are you?"; */ string s = "/* // */";

Hey, here's something weird.

If you type my (top-of-head) code above into Visual Studio, it produces
one long line that's somehow a single comment, and none of it gets
compiled as actual code.

/* string s = "/* hello */ // how are you?"; */ string s = "/* // */";

But if you put a line break where you'd expect the first multi-line
comment to end -- without changing anything else -- then the second
statement gets syntax-coloured and compiled as C# code.

/* string s = "/* hello */ // how are you?"; */
string s = "/* // */";

So I think I've accidentally found a bug. What do I win?

Eq.

P.S. My question to the original poster still stands, bug or none!

 

[Paul E Collins]

Blah, never mind I worked out what's going on with that line.

This can serve as a lesson to anyone who would combine // and /**/
comments.

Eq.

 

[Jon Skeet [C# MVP]]

we can keep it to c#:

1. /* a block can contain newlines and must match till next */
2. // must match until end of line
3. quotes must be supported
a. all text from a starting quote to closing quote must not be
matched
b. must support \" escape sequence inside quotes.

What about @"\"//This is a comment ?

 

[Jesse Houwing]

Hello Jarlaxle,

I'd like to issue a challenge (I already have a class to do it so I'm
not asking just to get the code)...

Write one regex expression that removes all comments from a string.

Hints:

1. /* a block can contain newlines and must match till next */
2. // must match until end of line
3. I have seen some that claim to remove all comments but are
inadequate for
one reason...(QUOTES!)

I would love to take this challenge on the one hand, but am lackign time
and incentive to do so at this time.

There's a lot to keep in mind even when only considering C# as target language
for this exercise.

I say:

A that this can be done with regular expressions.
B that it will be bloody hard and unreadable
C that regex isn't the best tool for this job
D that if you mix up comment like code with string constants as in the provided
samples, you're either crazy or need total job security

So for those that want to try:

A) Verbatim strings are a pain in the ass for this as it removes the security
of the normal line ends.
B) You need to understand balancing groups to get this to work
C) And a lot of look ahead's/behinds
D) And greedy matching to solve performance issues

The best way is probably to make a regex that matches from the start to the
end and use a MatchEvaluator to null all the comments found... but that wouldn't
be a pure regex solution would it?

 

[Ben Voigt [C++ MVP]]

Hey, here's something weird.

If you type my (top-of-head) code above into Visual Studio, it produces
one long line that's somehow a single comment, and none of it gets
compiled as actual code.

/* string s = "/* hello */ // how are you?"; */ string s = "/* // */";

But if you put a line break where you'd expect the first multi-line
comment to end -- without changing anything else -- then the second
statement gets syntax-coloured and compiled as C# code.

This just goes to show that where you'd expect the first comment to end is
not, in fact, where it does end.

Here is the first comment:
/* string s = "/* hello */

The */ inside quotes is NOT skipped because it is NOT inside a quoted string
literal because a quote inside a comment is a comment, not the beginning of
a string literal.

 

[Ben Voigt [C++ MVP]]

Blah, never mind I worked out what's going on with that line.

This can serve as a lesson to anyone who would combine // and /**/
comments.

You absolutely should combine // and /**/ comments. If you need to comment
out a block of code, you need to block prefix with // because /* */ style
comments do not nest.