[C] remove comments

[Tool]

K&R2 1-23: Write a program to remove all comments from a C program.
Don't forget to handle quoted strings and character constants
properly. C comments do not nest.

So i have to remove all comments from a C program.

I dont understand how i'm supposed to handle quoted strings and character constants. Delete them aswell or what do they mean?

[laserlight]

I think that it means is that you cannot blindly remove the contents of strings, even if they appear to contain comments.

[Tool]

So basicaly, i have to have 4 states - normal, comment, quoted string, and chracter constant and make sure i only delete comments which i find from normal state.

Or is there an easier fix to this problem?

[laserlight]

Well, that sounds about right to me, actually.

[Tool]

But im interested about the constant state.

I dont think a comment start can fit into a constant:

If you write '/*' debuger reports an error. So why am i supposed to bother with the constant part?

[laserlight]

I dont think a comment start can fit into a constant:

If you write '/*' debuger reports an error. So why am i supposed to bother with the constant part?

There are multi-character constants, though I am not well read about them.

[Tool]

I know about escape char constants, example

'\n'.

But that only works for escape character \ followed by any character.

I dont know if there are any other multi char consts, are there? One which could be filled with a comment?

[MK27]

I would say both constants and defines must use QUOTED string literals, so all you have to do is ignore "comments" that occur inside quotes.

AFAIK you cannot extent a quote over multiple lines in C code. Also, a single quote will not escape a comment*, so all you have to worry about is double quoted strings on a single line. I guess the thing about constants and defines is just to draw your attention to the issue.

* '\/*' is nonsense anyway, and it WILL open a comment.

[laserlight]

I know about escape char constants, example

'\n'.

But that only works for escape character \ followed by any character.

Yes, but that is nonetheless a single character constant.

I dont know if there are any other multi char consts, are there?

The standard allows for them.

One which could be filled with a comment?

I have my doubts. I think MK27's advice is sound: just concentrate on string literals. You can get back to handling multi-character constants '/*' and '*/' (and '//' if you are dealing with C99) as a remote possibility for extra credit or something later.

AFAIK you cannot extent a quote over multiple lines in C code.

Besides automatic concatenation, a string literal can span multiple lines in the source code by using a backslash at the end of the lines that it spans (other than the last). However, if I remember correctly this would become a single line after preprocessing.

[MK27]

Besides automatic concatenation, a string literal can span multiple lines in the source code by using a backslash at the end of the lines that it spans (other than the last). However, if I remember correctly this would become a single line after preprocessing.

Point.

char eg = "this and \
     that";

[Tool]

* '\/*' does not open a comment in my compiler.

So the only state i have to worry about is quote? And if im allready in the state quote, and i reach a \n newline signal, i have to return into normal state therefore, right? Only if there's \ and then \n, then i dont have to, right?

Since i wrote exercise 1-24 today, i didn't include that going into a newline after being in double quote state throws you out of double quote state.
Should i add that part aswell?


Sorry. But what are string literals?

[MK27]

* '\/*' does not open a comment in my compiler.

Well, it is some kind of error or mistake in any case.

So the only state i have to worry about is quote? And if im allready in the state quote, and i reach a \n newline signal, i have to return into normal state therefor, right? Only if there's \ and then \n, then i dont have to, right?

The first one would also be some kind of error/mistake. The ONLY exception to that would be '"' (single quoted double quote). I don't think you should have to account for how the code could appear if it is not going to compile properly anyhow.

String literals are strings which appear literally "as is" in the code. In C these are always double quoted (so a string literal could be a define or a variable assignment). AFAIK, the only place you can escape comments is in a string literal. Pretty nearly positive.

[Tool]

The first one would also be some kind of error/mistake. The ONLY exception to that would be '"' (single quoted double quote). I don't think you should have to account for how the code could appear if it is not going to compile properly anyhow.

Can you please simplify what you're saying.

I don't understand a thing what you said here. What would be an error?

What do i have to deal with here then.

I have to make a parser with states. States are normal and double_quote, you said i should ignore single quote constants.

Only from the normal state i delete comments.

The only exception for the double quote state is escape sequence \, which i know how it works and i will add it.

Other then that, am i supposed to take care of any other forms of exceptions?

AFAIK, the only place you can escape comments is in a string literal. Pretty nearly positive.

Yes.

[King Mir]

What you need to do is come up with a list of test cases that you want to support. The list should include all the complex possibility with single quotes, double quotes, multi line comments and any other possibilities you can think of. Do you need to worry about ??/ used in place of \? What about C++/C99 single line comments? If yes, write a few examples.

Only once you have that can you start thinking about how many state you need, and what constitutes a transition between them. Though I suspect that you're original four states are correct and all needed. Five if you support // comments.

[MK27]

Can you please simplify what you're saying.

I don't understand a thing what you said here. What would be an error?

My point was just '\/*' will not appear in a legal, sensible C program. The only purpose of single quotes in C that I can think of is to indicate character values ('\0', 'x', '\\', '"', etc) which is a finite set that does not include '\/*' '/*' or '//'.