Problem with K&R Exercise 1-23

So I am writing the solution [including nested comments and special cases, since I don't like half-assing]

I wrote a string that has special cases.

Code:

---

printf("\"test/*iello*/"); /*hello*/ //hello

---

It is coming out to

Code:

---

printf("\"test/*iello*/"); /

---

My code is

Code:

---

    void remcomments( char s[] )
    {
            printf("Removing comments\n");
            int slcflag, mlcflag, qflag;
            int readindex, writeindex;
            slcflag = mlcflag = readindex = qflag = writeindex = 0;
            int loop = 0;
            while( s[readindex] != EOF )
            {
                    printf("Loop %d\n", loop);
                    if( s[readindex] == '\n' && mlcflag == 0 )
                    {
                            s[writeindex] = s[readindex];
                            slcflag = 0;
                            writeindex++;
                    }
                    else
                    {
                            if( qflag == 0 )
                            {
                                    if( s[readindex] == '"' )
                                    {
                                            if( s[readindex - 1] != '\\' )
                                            {
                                                    qflag = 1;
                                            }
                                            else if( s[readindex - 1] == '\\' && s[readindex - 2] == '\\' )
                                            {
                                                    qflag = 1;
                                            }
                                    }
                                    if( qflag == 0 )
                                    {
                                        if( slcflag == 0 && mlcflag == 1 && s[readindex] == '*' && s[readindex + 1] == '/' )
                                            {
                                                    readindex++;
                                                    mlcflag = 0;
                                            }
                                            else if( s[readindex] == '/' )
                                            {
                                                    if( mlcflag == 0 && s[readindex + 1] == '/' )
                                                    {
                                                            readindex++;
                                                            slcflag = 1;
                                                    }
                                                    else if( slcflag == 0 && s[readindex + 1] == '*' )
                                                    {
                                                            readindex++;
                                                            mlcflag = 1;
                                                    }
                                            }
                                    }
                            }
                            else if( qflag == 1 )
                            {
                                    if( s[readindex] == '"')
                                    {
                                            if( s[readindex - 1] != '\\')
                                            {
                                                    qflag = 0;
                                            }
                                            else if( s[readindex - 1] == '\\' && s[readindex - 2] == '\\')
                                            {
                                                    qflag = 0;
                                            }
                                    }
                            }
                            if( mlcflag == 0 && slcflag == 0 )
                            {
                                    s[writeindex] = s[readindex];
                                    writeindex++;
                            }
                    }       
                    loop++;
                    readindex++;
                    printf("end of loop");
            }
            s[writeindex] = '\0';
    }

---

Where am I going wrong? I can't figure it out.

I think the error is very simple. Check around lines 34 through 38 of your program. When you are comming out of a multi-line comment, you are not moving the readindex far enough into the string, so consequently, you end up writing part of the comment.

I noticed this while watching your function execute a few times. Here's the proof:

Code:

---

(gdb) c
Continuing.
end of loopLoop 33
Hardware watchpoint 3: writeindex


Old value = 27
New value = 28
remcomments (s=0x61fb28 "printf(\"\\\"test/*iello*/\"); /*hello*/ //hello") at remcomments.c:78
78              loop++;
(gdb) print readindex
$13 = 35
(gdb) p s+readindex
$14 = 0x61fb4b "/ //hello"
(gdb) p s+writeindex
$15 = 0x61fb44 "*hello*/ //hello"
(gdb) p slcflag
$16 = 0
(gdb) p mlcflag
$17 = 0

---

Notice where $14 and $15 start. Since both slcflag and mlcflag are 0, your function wrote the slash at the end of the comment here.

Code:

---

printf("\"test/*iello*/"); /*hello*/ //hello

---

So once again if you bump readindex enough to skip past the whole comment token, I think you will have less of a problem.

Quote:

---

Originally Posted by whiteflags

I think the error is very simple. Check around lines 34 through 38 of your program. When you are comming out of a multi-line comment, you are not moving the readindex far enough into the string, so consequently, you end up writing part of the comment.

I noticed this while watching your function execute a few times. Here's the proof:

Code:

---

(gdb) c
Continuing.
end of loopLoop 33
Hardware watchpoint 3: writeindex


Old value = 27
New value = 28
remcomments (s=0x61fb28 "printf(\"\\\"test/*iello*/\"); /*hello*/ //hello") at remcomments.c:78
78              loop++;
(gdb) print readindex
$13 = 35
(gdb) p s+readindex
$14 = 0x61fb4b "/ //hello"
(gdb) p s+writeindex
$15 = 0x61fb44 "*hello*/ //hello"
(gdb) p slcflag
$16 = 0
(gdb) p mlcflag
$17 = 0

---

Notice where $14 and $15 start. Since both slcflag and mlcflag are 0, your function is going to write the slash at the end of the comment here.

Code:

---

printf("\"test/*iello*/"); /*hello*/ //hello

---

To fix it, try moving readindex ahead by two characters instead of one when you detect */ ; this will properly skip over those elements in the text.

---

I did that before, but I'm pretty sure there's an edge case where I'm going to miss something. So instead I added a skip flag at the line I'd add the readindex++ and at the write section at the end I just make sure skip flag == 0, and then after that I set skip flag to 0. I'm pretty sure reading ahead twice is going to skip a char since at the end of the function I increment the read index also.

You're right. I just did that and had a much more complex test and hell, it worked :) Thanks a lot for your help. Now I just want to figure out how I can remove the newline character if the only thing on that line was a comment. But I think that is going to make my program much larger.

Quote:

---

Originally Posted by whiteflags

I think the error is very simple. Check around lines 34 through 38 of your program. When you are comming out of a multi-line comment, you are not moving the readindex far enough into the string, so consequently, you end up writing part of the comment.

I noticed this while watching your function execute a few times. Here's the proof:

Code:

---

(gdb) c
Continuing.
end of loopLoop 33
Hardware watchpoint 3: writeindex


Old value = 27
New value = 28
remcomments (s=0x61fb28 "printf(\"\\\"test/*iello*/\"); /*hello*/ //hello") at remcomments.c:78
78              loop++;
(gdb) print readindex
$13 = 35
(gdb) p s+readindex
$14 = 0x61fb4b "/ //hello"
(gdb) p s+writeindex
$15 = 0x61fb44 "*hello*/ //hello"
(gdb) p slcflag
$16 = 0
(gdb) p mlcflag
$17 = 0

---

Notice where $14 and $15 start. Since both slcflag and mlcflag are 0, your function wrote the slash at the end of the comment here.

Code:

---

printf("\"test/*iello*/"); /*hello*/ //hello

---

So once again if you bump readindex enough to skip past the whole comment token, I think you will have less of a problem.

---

So, there was an edge case where it would error out.

Code:

---

"printf(/*test*/\"\\\"test/*iello*/  //more\"/*test*/);"

---

It will jump over the \, then next it would trigger an "out of quote" state, and output printf(""test, since it's thinking the /* in the string is a valid beginning of mlc marker.
I fixed it though.
I fixed it so hard, now it can even skip blank lines.
and I don't have to do readindex++.
The new code:

Code:

---

void remcomments(char s[])
{
        int slcflag, mlcflag, qflag;
        int readindex, writeindex;
        slcflag = mlcflag = qflag = readindex = writeindex = 0;
        while (s[readindex] != '\0')
        {
                if (slcflag == 1 && s[readindex] == '\n')
                        slcflag = 0;
                if (mlcflag == 0)
                {
                        if (s[readindex] == '"' && (s[readindex - 1] != '\\' || (s[readindex - 1] == '\\' && s[readindex - 2] == '\\')))
                        {
                                if (qflag == 0) qflag = 1;
                                else qflag = 0;
                        }
                        else if (qflag == 0)
                        {
                                if (s[readindex] == '/' && s[readindex + 1] == '/')
                                {
                                        readindex++;
                                        slcflag = 1;
                                }
                                if (s[readindex] == '/' && s[readindex + 1] == '*')
                                {
                                        readindex++;
                                        mlcflag = 1;
                                }
                        }
                        if (slcflag == 0 && mlcflag == 0 && !(s[readindex] == '\n' && s[writeindex - 1] == '\n'))
                        {
                                s[writeindex] = s[readindex];
                                writeindex++;
                        }
                }
                else if (s[readindex - 1] == '*' && s[readindex] == '/')
                                mlcflag = 0;
                readindex++;
        }
        s[writeindex] = '\0';
}

---

Quote:

---

Originally Posted by CjStaal

I did that before, but I'm pretty sure there's an edge case where I'm going to miss something. So instead I added a skip flag at the line I'd add the readindex++ and at the write section at the end I just make sure skip flag == 0, and then after that I set skip flag to 0. I'm pretty sure reading ahead twice is going to skip a char since at the end of the function I increment the read index also.

---

Code:

---

if (slcflag == 0 && mlcflag == 0 && !(s[readindex] == '\n' && s[writeindex - 1] == '\n'))

---

This looks like it works, until you consider a string that starts with \n. In that case, this test gets to s[writeindex - 1], and since writeindex == 0, you read s[-1] and potentially crash.

It is much safer to write the extra newlines and overwrite duplicates instead.

Code:

---

if (!slcflag && !mlcflag)
{
    s[writeindex] = s[readindex];
    writeindex++;
    if (writeindex >= 2 && s[writeindex - 2] == '\n' && s[writeindex - 1] == '\n')
    {
        writeindex--;
    }
}

---

In one of the simplest cases, "\n//comments\n", the result should be "\n" because of the fact that s was null terminated where the last character is, later.