String Trim in C

HashBox · 2009-06-09 21:47:50

Hello everyone, I have coded a small string trim function in C, and I was wondering if my method is flawed in any way.
Try to go easy on me though :\
I'm mostly just curious as to whether it can be optimized further.

This probably isn't even what you would call a real string trim since it doesn't completely alter the string, it basically returns a pointer to the start of string (without the spaces), and chops the end of the string with nulls.

Oh well I'm interested to see what others come up with.

/* Yay goto :D */
char *trim(char *buffer, char *stripchars)
{
        int i = 0;

        /* Left Side */
        char *start = buffer;

left:
        for (i = 0; i < strlen(stripchars); i++) {
                if (*start == stripchars[i]) {
                        start++;
                        goto left;
                }
        }

        /* Right Side */
        char *end = start + strlen(start) - 1;

right:
        for (i = 0; i < strlen(stripchars); i++) {
                if (*end == stripchars[i]) {
                        *end = '\0';
                        --end;
                        goto right;
                }
        }

        return start;
}

EDIT: You can use it like so,

char buffer[] = "    Hello World     ";
printf("%s", trim(buffer, " "));

Last edited by HashBox (2009-06-09 21:51:13)

Peasantoid · 2009-06-09 21:54:37

Only problem I see is that you're using goto.

HashBox · 2009-06-09 22:00:13

I knew someone had to say it

Consider this, without goto that function (as far as I can see) would need two new for loops (or two while loops), two new if statements, and a new variable, as that is how it was originally.

My only concern about using goto is the effect it could have when breaking out of a for loop, whether it might do something to the stack for example, I'm not terribly well versed about this.

iza · 2009-06-09 22:59:57

goto does have its uses, however in such a simple function it shouldn't really be necessary.

Here's another way you could do it without goto; had to add one variable (i wish C had booleans =/), and a do-while loop instead of each label/goto.

char *trim(char *buffer, char *stripchars)
{
    int i = 0;
    int flag;

    /* Left Side */
    char *start = buffer;

    do {
        flag = 0;
        for (i = 0; i < strlen(stripchars); i++) {
            if (*start == stripchars[i]) {
                start++;
                flag = 1;
                break;
            }
        }
    } while (flag);

    /* Right Side */
    char *end = start + strlen(start) - 1;

    do {
        flag = 0;
        for (i = 0; i < strlen(stripchars); i++) {
            if (*end == stripchars[i]) {
                *end = '\0';
                --end;
                flag = 1;
                break;
            }
        }
    } while (flag)

    return start;
}

Still pretty ugly -- I'm sure there is a better way of doing this.

Peasantoid · 2009-06-09 23:01:19

iza wrote:

(i wish C had booleans =/)

C does have booleans. It's called "int foo = 1;".

HashBox · 2009-06-09 23:14:52

Peasantoid wrote:

iza wrote:
(i wish C had booleans =/)
C does have booleans. It's called "int foo = 1;".

or

#define bool int
#define true 1
#define false 0

@iza, this is basically how I was doing it before I switched to using goto, although yours is slightly cleaner than mine was.

Procyon · 2009-06-09 23:16:18

What if you give it a string with all characters in stripchars, there's a small chance especially for the end that data is replaced before the string's original start, if I see it right.

voice · 2009-06-09 23:21:34

Here is some trim realizations:
http://gd.tuwien.ac.at/languages/c/cref … NIP/trim.c
http://agraja.wordpress.com/2007/12/13/ … nk-spaces/

HashBox · 2009-06-09 23:27:23

Procyon wrote:

What if you give it a string with all characters in stripchars, there's a small chance especially for the end that data is replaced before the string's original start, if I see it right.

Lucky me, it seems to hold up against that too
Assuming this is what you mean:

char buffer[] = "   hello world    ";
printf("'%s'\n", trim(buffer, " abcdefghijklmnopqrstuvwxyz"));

Returns ''
Edit: Now that I look at it, if you give it an empty string for either parameters it'll probably segfault. But checking for that is pretty simple.

@voice, thanks for that, interesting to see other ways that people do it

Last edited by HashBox (2009-06-09 23:41:16)

iza · 2009-06-09 23:46:57

Procyon wrote:

What if you give it a string with all characters in stripchars, there's a small chance especially for the end that data is replaced before the string's original start, if I see it right.

I think you might be right. The left side part would move the start pointer to the null at the end of the input string. Then, the end pointer would be placed one unit to the left of that (since strlen(start) is 0). The loop would then proceed back through the input string and null it all. If the memory directly to the left of the input string happens to match one of the stripchars, it would also null it. It's highly unlikely, but even the fact that it is reading memory in a place it shouldn't be is not good (possibility of unauthorized memory access error).

Last edited by iza (2009-06-09 23:47:58)

HashBox · 2009-06-09 23:53:47

iza wrote:

Procyon wrote:
What if you give it a string with all characters in stripchars, there's a small chance especially for the end that data is replaced before the string's original start, if I see it right.
I think you might be right. The left side part would move the start pointer to the null at the end of the input string. Then, the end pointer would be placed one unit to the left of that (since strlen(start) is 0). The loop would then proceed back through the input string and null it all. If the memory directly to the left of the input string happens to match one of the stripchars, it would also null it. It's highly unlikely, but even the fact that it is reading memory in a place it shouldn't be is not good (possibility of unauthorized memory access error).

If I'm not mistaken, this should only happen if the input is a zero length string? (strlen = 0 as you said)
Therefore, checking for a zero length string should solve this? or is there another simple check that would be better?

Edit: Even with a zero length input string and both a zero length stripchars and a stripchars with " abcdefghijklmnopqrstuvwxyz", it still does not crash, I guess that would be due to the memory before it not containing any of the stripchars though correct?

Still I think that's sorted most of the problems with it, thanks for all the input everyone

Last edited by HashBox (2009-06-10 00:02:27)

moljac024 · 2009-06-10 06:42:47

HashBox wrote:

Peasantoid wrote:
iza wrote:
(i wish C had booleans =/)
C does have booleans. It's called "int foo = 1;".
or
#define bool int
#define true 1
#define false 0
@iza, this is basically how I was doing it before I switched to using goto, although yours is slightly cleaner than mine was.

Integer still uses double the memory a boolean does on most machines

Last edited by moljac024 (2009-06-10 06:46:35)

HashBox · 2009-06-10 06:59:59

moljac024 wrote:

Integer still uses double the memory a boolean does on most machines

I decided to google this for the hell of it and yes it does seem to depend on the machine and also the version of GCC, interesting stuff. The more you know

Meillo · 2009-06-10 07:44:18

HashBox wrote:

Peasantoid wrote:
iza wrote:
(i wish C had booleans =/)
C does have booleans. It's called "int foo = 1;".
or
#define bool int

I think

typedef int bool

is better.

Procyon · 2009-06-10 08:19:49

Here is an example of how it will write over other data:

#include <stdio.h>
#include <string.h>

(trim function)

void main() {
    char buffer[] = "BBBBBBBBBBBBBBB";
    char array[4] = { 'Q','E','E','E'};
    printf("Array Before: %4.4s\n", array);
    printf("Buffer After: %s\n", trim(buffer, "EB"));
    printf("Array After: %4.4s\n", array);
}

--> gcc -o trimtest trimtest.c 
--> ./trimtest
Array Before: QEEE
Buffer After: 
Array After:    Q

HashBox · 2009-06-10 09:02:29

Ouch I see what you mean, thanks for providing that example.

kumyco · 2009-06-10 11:31:45

Procyon wrote:

Here is an example of how it will write over other data:
char array[4] = { 'Q','E','E','E'};

i don't think you want to pass that array to any string related function, it's asking for a segfault
----
i thought it'd be interesting to implement this (without the goto)

char* ky_trim (char *buffer, const char *stripchars)
{
    char *buf;
    register size_t i=0, j;
    if (!buffer || !*buffer || !stripchars || !*stripchars) return buffer;
    
    do {
        for (j=0; j < strlen(stripchars) && buffer[i] != stripchars[j]; ++j){};
    } while (j < strlen(stripchars) && (i++ < strlen(buffer)));
    
    buf = buffer + i;
    i = strlen(buffer) - 1;
    
    do {
        for (j=0; j < strlen(stripchars) && buffer[i] != stripchars[j]; ++j){};
    } while (j < strlen(stripchars) && (i-- > 0));
    
    buffer[i + 1] = '\0';
    if (buf == buffer) return buffer;
    return memmove(buffer, buf, &buffer[i] - buf + 2);
}

and using pointers

char* ky_trim_ptr (char *buffer, const char *stripchars)
{
    char *buf=buffer, *start;
    const char *stp;
    
    if (!buffer || !*buffer || !stripchars || !*stripchars) return buffer;
    
    do {
        stp = stripchars;
        while (*stp && *buf != *stp) ++stp;
        if (*stp) ++buf;
    } while (*buf && *stp);
    
    start = buf;
    while (*buf) ++buf;
    --buf;
    
    do {
        stp = stripchars;
        while (*stp && *buf != *stp) ++stp;
        if (*stp) --buf;
    } while (*buf && *stp);
    *(buf + 1) = '\0';
    
    if (start == buffer) return buffer;
    return memmove(buffer, start, buf - start + 2);
}

applied the standard -O2 and tested as-is (with the memmove), they were largely the same, thoguh the ptr version appeared to be slightly faster than the other two.
i used memmove because for larger buffers it's faster than a straight loop, you need to memove because if you use it on a malloc()'d buffer, you very likely to lose the pointer causing a memory leak

vacant · 2009-06-10 12:14:32

My C skills are very rusty, but assuming the string passed is null terminated and writeable, then a couple of "while" statements are about all that are needed for a "simple" trim (space, null, tabs are all less than ASCII 33)

#include "stdio.h"
#include "string.h"

char* trim (char* stripthis) {

    char *s = stripthis - 1, *e = stripthis + strlen (stripthis);

    while (++s < e && *s < 33);
    while (--e > s && *e < 33);
    *(++e) = (char) 0;

    return s;
    }

main () {

    char buffer[] = "    Hello World     ";
    printf(">%s<", trim(buffer));
    }

edited to tidy & deal with blank string

Last edited by vacant (2009-06-10 13:32:15)

Cerebral · 2009-06-10 13:09:28

kumyco wrote:

char* ky_trim (char *buffer, const char *stripchars)
...
...
applied the standard -O2 and tested as-is (with the memmove), they were largely the same, thoguh the ptr version appeared to be slightly faster than the other two.

The amount of times you used strlen there made me wince. If you want to improve the perf of this one, try calling strlen once, and using the result everywhere else. Like so:

char* ky_trim (char *buffer, const char *stripchars)
{
    char *buf;
    register size_t i=0, j;
    register size_t bufLen, stripLen;

    if (!buffer || !*buffer || !stripchars || !*stripchars) return buffer;

    bufLen = strlen(buffer);
    stripLen = strlen(stripchars);

    do {
        for (j=0; j < stripLen && buffer[i] != stripchars[j]; ++j){};
    } while (j < stripLen && (i++ < bufLen));
    
    buf = buffer + i;
    i = bufLen - 1;
    
    do {
        for (j=0; j < stripLen && buffer[i] != stripchars[j]; ++j){};
    } while (j < stripLen && (i-- > 0));
    
    buffer[i + 1] = '\0';
    if (buf == buffer) return buffer;
    return memmove(buffer, buf, &buffer[i] - buf + 2);
}

kumyco · 2009-06-10 13:12:49

gcc -O2 optimizes it away x)

Cerebral · 2009-06-10 13:19:26

kumyco wrote:

gcc -O2 optimizes it away x)

Maybe I'm missing something here, but how can that be possible when it doesn't even know the input string at compile time?

kumyco · 2009-06-10 13:25:45

i dunno the implementation details but I'd think it can see that there are really only 2 separate calls to strlen, so caches the first calls and reuse them -- or something like that, i didn't get it either (never even considered) until someone pointed it out, a quick

a quick compile with -S will show 6 calls to strlen, recompile with -S -O2 will show only 2

vacant · 2009-06-10 13:35:55

Cerebral wrote:

kumyco wrote:
gcc -O2 optimizes it away x)
Maybe I'm missing something here, but how can that be possible when it doesn't even know the input string at compile time?

The compiler sees that the address of the symbol & contents stripchars do not alter within the "do...while"? Not sure about async (e.g. interrupts/sensor reading).

Last edited by vacant (2009-06-10 13:40:05)

Aprz · 2009-06-10 13:43:29

As for the whole true and false thing, what about bit-fields?

Edit: I've never used a bit-field before, but just remember reading about it.

struct boolean {
    unsigned false:    0;
    unsigned true:     1;
} bool;

Wouldn't that be smaller than using an intenger?

Last edited by Aprz (2009-06-10 13:49:41)

kumyco · 2009-06-10 13:46:52

don't, 4 bytes is almost nothing and in some cases it's far more efficient to use the type that matches the native word-size than 1 byte etc.

Arch Linux

#1 2009-06-09 21:47:50

String Trim in C

#2 2009-06-09 21:54:37

Re: String Trim in C

#3 2009-06-09 22:00:13

Re: String Trim in C

#4 2009-06-09 22:59:57

Re: String Trim in C

#5 2009-06-09 23:01:19

Re: String Trim in C

#6 2009-06-09 23:14:52

Re: String Trim in C

#7 2009-06-09 23:16:18

Re: String Trim in C

#8 2009-06-09 23:21:34

Re: String Trim in C

#9 2009-06-09 23:27:23

Re: String Trim in C

#10 2009-06-09 23:46:57

Re: String Trim in C

#11 2009-06-09 23:53:47

Re: String Trim in C

#12 2009-06-10 06:42:47

Re: String Trim in C

#13 2009-06-10 06:59:59

Re: String Trim in C

#14 2009-06-10 07:44:18

Re: String Trim in C

#15 2009-06-10 08:19:49

Re: String Trim in C

#16 2009-06-10 09:02:29

Re: String Trim in C

#17 2009-06-10 11:31:45

Re: String Trim in C

#18 2009-06-10 12:14:32

Re: String Trim in C

#19 2009-06-10 13:09:28

Re: String Trim in C

#20 2009-06-10 13:12:49

Re: String Trim in C

#21 2009-06-10 13:19:26

Re: String Trim in C

#22 2009-06-10 13:25:45

Re: String Trim in C

#23 2009-06-10 13:35:55

Re: String Trim in C

#24 2009-06-10 13:43:29

Re: String Trim in C

#25 2009-06-10 13:46:52

Re: String Trim in C

Board footer