String parsing program

[pereges]

Hi I've a string input and I have to parse it in such a way that that
there can be only white space till a digit is reached and once a digit
is reached, there can be only digits or white space till the string
ends. Am I doing this correctly ? :

Code:

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

int main(void)
{
char s[50];
int i = 0;

gets(s);

while (isspace(s[i]))
i++;
while (isdigit(s[i]))
i++;
while (isspace(s[i]))
i++;
if (s[i] != '\0')
printf("\nIncor rect string\n");

return (0);
}

I want to actually convert a string to unsigned long. So this kind of
algorithm should be carried out prior to strtoul function to ensure
that some of the weakness from which the strtoul function suffers like
convertin 123aaaaa to 123 for eg or -123 to some unsigned value is
removed. This will also ensure that when you have a string like :

1234 78

1234 is not returned but an error message will be printed. Because a
string should only contain 1 number in my program.

[pereges]

On Jul 4, 2:14 am, Ben Bacarisse <ben.use...@bsb .me.ukwrote:

<snip>
The simplest way to scan for a number whilst reporting bad input is to
use the signed strtol function. You check that errno has not been set
to ERANGE and that the end-pointer is not the string you passed in.
If you like, you can now check that nothing but white space is left in
the string. Finally, you confirm the input is the range your program
expects. The signed version lets you detect input like -123. The down
side is that you loose half the range of possible inputs. If that
matters, you can (probably) go up to strtoll.
<snip>

My input is of following format :

45 5666 16000

^^ All of that is just a single string. I need to read the three
numbers into 3 different size_t variables. There can be white space
amongst them but no alphabets or any other characters. It is possible
check if *endp character is nothing but white space (This must be done
when errno != ERANGE and s != endp i.e. when one usually expects
correct output), but any character other than that means data is
erroneous. With unsigned long, you can check if the first non white
space character is a '-' or not. This should solve the problem of
negative numbers as well and prevent their conversion to some unsigned
value before hand.

[Default User]

santosh wrote:

Your code exhibits undefined behaviour because you have failed to
include ctype.h where the declarations for the is* functions are.

Not really. The default declarations will do for those. It's not good
practice, of course.


Brian

[Ben Bacarisse]

pereges <Br*****@gmail. comwrites:

On Jul 4, 2:14 am, Ben Bacarisse <ben.use...@bsb .me.ukwrote:

><snip>
The simplest way to scan for a number whilst reporting bad input is to
use the signed strtol function. You check that errno has not been set
to ERANGE and that the end-pointer is not the string you passed in.
If you like, you can now check that nothing but white space is left in
the string. Finally, you confirm the input is the range your program
expects. The signed version lets you detect input like -123. The down
side is that you loose half the range of possible inputs. If that
matters, you can (probably) go up to strtoll.
<snip>

My input is of following format :

45 5666 16000

^^ All of that is just a single string. I need to read the three
numbers into 3 different size_t variables. There can be white space
amongst them but no alphabets or any other characters. It is possible
check if *endp character is nothing but white space (This must be done
when errno != ERANGE and s != endp i.e. when one usually expects
correct output), but any character other than that means data is
erroneous. With unsigned long, you can check if the first non white
space character is a '-' or not. This should solve the problem of
negative numbers as well and prevent their conversion to some unsigned
value before hand.

Here is one way based on using the widest signed type for input. It
is not ideal, but then without very details specs, what could be? If
you need to accept input right up to SIZE_MAX and you have an
implementation where intmax_t can't hold that value, then you will
need to use strtoumax and check for the - manually, so to speak. That
would not be a big change to parse_size.

#include <stdio.h>
#include <stdbool.h>
#include <inttypes.h>
#include <stdint.h>
#include <errno.h>
#include <ctype.h>

size_t parse_size(cons t char *num, const char **endp, bool *error)
{
char *ep;
errno = 0;
intmax_t imax = strtoimax(num, &ep, 10);
if (errno == ERANGE || imax < 0 || imax SIZE_MAX) {
while (isspace(*num))
num++;
fprintf(stderr, "Input \"%.*s\" out of range.\n",
ep - num, num);
if (error)
*error = true;
}
else if (ep == num) {
/* Skip to the next space-delimited portion of the string. */
while (isspace(*ep))
ep++;
num = ep;
while (*ep != '\0' && !isspace(*ep))
ep++;
fprintf(stderr, "Input \"%.*s\" could not be converted.\n",
ep - num, num);
if (error)
*error = true;
}
if (endp)
*endp = ep;
return imax;
}

void parse_three_siz es(const char *input)
{
bool errors = false;
const char *ep;
size_t s1 = parse_size(inpu t, &ep, &errors);
size_t s2 = parse_size(ep, &ep, &errors);
size_t s3 = parse_size(ep, &ep, &errors);

if (!errors) {
/* Check that everything parsed. */
const char *save_ep = ep;
while (isspace(*ep))
ep++;
if (*ep != '\0')
fprintf(stderr, "Superfluou s input found: \"%s\"\n", save_ep);
printf("Got: %zu %zu %zu\n", s1, s2, s3);
}
}

int main(int argc, char **argv)
{
if (argc 1)
parse_three_siz es(argv[1]);
return 0;
}

--
Ben.

[Keith Thompson]

"Default User" <de***********@ yahoo.comwrites :

santosh wrote:

>Your code exhibits undefined behaviour because you have failed to
include ctype.h where the declarations for the is* functions are.

Not really. The default declarations will do for those. It's not good
practice, of course.

That works only in C90. Even though there are few full C99 compilers,
it doesn't hurt to write good C90 code that's also valid C99 code.

And the is* and to* functions are very likely to be implemented as
macros, with better performance than the function calls. By not
including the header, you miss out on the macro definitions.

--
Keith Thompson (The_Other_Keit h) ks***@mib.org <http://www.ghoti.net/~kst>
Nokia
"We must do something. This is something. Therefore, we must do this."
-- Antony Jay and Jonathan Lynn, "Yes Minister"

[Default User]

Keith Thompson wrote:

"Default User" <de***********@ yahoo.comwrites :

santosh wrote:

Your code exhibits undefined behaviour because you have failed to
include ctype.h where the declarations for the is* functions are.

Not really. The default declarations will do for those. It's not
good practice, of course.

That works only in C90. Even though there are few full C99 compilers,
it doesn't hurt to write good C90 code that's also valid C99 code.

So? In that case there's a required diagnostic for the missing
declaration.

And the is* and to* functions are very likely to be implemented as
macros, with better performance than the function calls. By not
including the header, you miss out on the macro definitions.

But not undefined behavior.


Brian

[Peter Nilsson]

pereges wrote:

Hi I've a string input and I have to parse it in such a way that that
there can be only white space till a digit is reached and once a digit
is reached, there can be only digits or white space till the string
ends. Am I doing this correctly ? :

Code:

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

int main(void)
{
char s[50];
int i = 0;

gets(s);

Real bad example.

>
while (isspace(s[i]))

while ((unsigned char) s[i])

i++;
while (isdigit(s[i]))
i++;

This does not _require_ a digit to be present.

while (isspace(s[i]))
i++;
if (s[i] != '\0')
printf("\nIncor rect string\n");
return (0);
}

I want to actually convert a string to unsigned long. So this
kind of algorithm should be carried out prior to strtoul
function to ensure that some of the weakness from which
the strtoul function suffers like convertin 123aaaaa to 123

That is not a weakness but a strength.

x = strtoul(str, &endp, 0);

if (endp != str)
while (isspace((unsig ned char) endp))
endp++;

if (endp != str && *endp == 0)
/* all good */;

for eg or -123 to some unsigned value is removed.

if (endp != str && *endp == 0 && strchr(str,'-') == 0)
/* all good */;

--
Peter

[pete]

Default User wrote:

Keith Thompson wrote:

>"Default User" <de***********@ yahoo.comwrites :

>>santosh wrote:
Your code exhibits undefined behaviour because you have failed to
include ctype.h where the declarations for the is* functions are.
Not really. The default declarations will do for those. It's not
good practice, of course.

That works only in C90. Even though there are few full C99 compilers,
it doesn't hurt to write good C90 code that's also valid C99 code.

So? In that case there's a required diagnostic for the missing
declaration.

>And the is* and to* functions are very likely to be implemented as
macros, with better performance than the function calls. By not
including the header, you miss out on the macro definitions.

But not undefined behavior.

The common subset of C90 and C99,
is like C90 with less allowable sloppy style.

It's a better way to write C90 code,
even if you don't intend it to be compiled as C99 code,

--
pete

[Keith Thompson]

"Default User" <de***********@ yahoo.comwrites :

Keith Thompson wrote:

>"Default User" <de***********@ yahoo.comwrites :

santosh wrote:
Your code exhibits undefined behaviour because you have failed to
include ctype.h where the declarations for the is* functions are.

Not really. The default declarations will do for those. It's not
good practice, of course.

That works only in C90. Even though there are few full C99 compilers,
it doesn't hurt to write good C90 code that's also valid C99 code.

So? In that case there's a required diagnostic for the missing
declaration.

>And the is* and to* functions are very likely to be implemented as
macros, with better performance than the function calls. By not
including the header, you miss out on the macro definitions.

But not undefined behavior.

You're right, there's no undefined behavior in either C90 or C99.

(Well, there's a constraint violation in C99; if the implementation
accepts the program in spite of that, after issuing the required
diagnostic, then the behavior is undefined. But that's stretching the
point.)

Failing to include <ctype.hwhen using the is* or to* functions is
still a bad idea, of course.

--
Keith Thompson (The_Other_Keit h) ks***@mib.org <http://www.ghoti.net/~kst>
Nokia
"We must do something. This is something. Therefore, we must do this."
-- Antony Jay and Jonathan Lynn, "Yes Minister"

[Richard Heathfield]

pereges said:

Hi I've a string input and I have to parse it in such a way that that
there can be only white space till a digit is reached and once a digit
is reached, there can be only digits or white space till the string
ends. Am I doing this correctly ? :

No.

gets(s);

Until you have learned why you must never call this function, there is
little point trying to learn anything else about C. This is your next
step. Trying to jump over it is most unwise.

<snip>

--
Richard Heathfield <http://www.cpax.org.uk >
Email: -http://www. +rjh@
Google users: <http://www.cpax.org.uk/prg/writings/googly.php>
"Usenet is a strange place" - dmr 29 July 1999

[Default User]

pete wrote:

Default User wrote:

But not undefined behavior.

It's a better way to write C90 code,
even if you don't intend it to be compiled as C99 code,

What does this have to do with what I said?

Brian