Strtok() - int or float it be?

So, I am utilizing the string library...

Code:

---

str = strtok(line, ",");
variable = atof(str);

if (sizeof(variable) == sizeof(float)) {
  var_i_or_f = 1;
  variable = atof(str);
  }
else if(sizeof(variable) == sizeof(int)) {
  var_i_or_f = 0;
  variable = atoi(str);
    }

---

My question is: is their a way to know or test to see if str will be a int or a float?

At first, I thought I could use sizeof() and conditional statements to test before realizing that I am using atof to get str. Thus can't check it if its one or the other since im using it... Trouble is that the data switches between int and float so their is no prescribed way to know which it will be beforehand.

Is this possible.

Sorry about my code: that is from memory - it compiles correctly but what I may have typed HERE may have an error in it.

I think you will have to do something like:

if (strchr(str,'.')) {

to decide.

Quote:

---

Originally Posted by MK27

I think you will have to do something like:

if (strchr(str,'.')) {

to decide.

---

Ok, strchr() returns a pointer to the first occurrence of character in the C string str. So, essentially, looks to see if str has a decimal point - thus if its a float value or an int. BUT, by using strchr do we lose what str originally was assigned/given? Does this string library function modify what str originally was, is what i mean...

I mean if we did it that way:

Code:

---

str = strtok(line, ",");
if (strchr(str,'.')) {
  variable = atof(str)
else (variable = atoi(str))

---

I feel that lost in this, is what str was originally assigned when it was given a value by strtok()....

Thoughts?

Nope. strchr takes a const char *, so it cannot change the passed string.

Neat - works groovy. So by its description of "pointer to the first occurrence of character in the C string str.", what it returns is the location of the "." and thus, preserving what str was assigned!

correct?

Yeah, if there's no "." it returns NULL. This is a "truth test":

Code:

---

if (strchr(str,'.'))

---

means:

Code:

---

if (strchr(str,'.') != NULL)

---

The opposite:

Code:

---

if (!strchr(str,'.'))
// means:
if (strchr(str,'.') == NULL)

---

Because you don't care where it is, just whether it is there. Or not.

Quote:

---

Originally Posted by towed

My question is: is their a way to know or test to see if str will be a int or a float?

---

Many ints will also be floats. I would choose to use strtol and strtof to attempt conversion of the text. If either is successful, you know what you got. If neither is successful, odds are it was neither. The order of calling these two allows you to have preference for one or the other. Something like this:

Code:

---

#include <stdio.h>
#include <stdlib.h>

union uType
{
  int  i;
  float f;
};

struct tType
{
  enum eType
  {
      TYPE_INT, TYPE_FLOAT, TYPE_OTHER, TYPE_TOTAL
  } type;
  union uType value;
};

int foo(struct tType *data, const char *text)
{
  char *end;
  data->value.i = strtol(text, &end, 10);
  if ( end > text && *end == '\0' )
  {
      data->type = TYPE_INT;
      return 1;
  }
  data->value.f = strtof(text, &end);
  if ( end > text && *end == '\0' )
  {
      data->type = TYPE_FLOAT;
      return 1;
  }
  data->type = TYPE_OTHER;
  return 0;
}

int main()
{
  const char *input[] =
  {
      "",
      "-123",
      "-123.45",
      "123E-2",
      "123abc",
      "xyz",
      "123E2",
      "1.23.E2",
  };
  size_t i;
  for ( i = 0; i < sizeof input / sizeof *input; ++i )
  {
      struct tType data;
      int result = foo(&data, input[i]);
      printf("\"%s\" : ", input[i]);
      switch ( data.type )
      {
      case TYPE_INT:  printf("TYPE_INT  %d", data.value.i); break;
      case TYPE_FLOAT: printf("TYPE_FLOAT %g", data.value.f); break;
      default:        fputs("TYPE_OTHER", stdout); break;
      }
      putchar('\n');
  }
  return 0;
}

/* my output
"" : TYPE_OTHER
"-123" : TYPE_INT  -123
"-123.45" : TYPE_FLOAT -123.45
"123E-2" : TYPE_FLOAT 1.23
"123abc" : TYPE_OTHER
"xyz" : TYPE_OTHER
"123E2" : TYPE_FLOAT 12300
"1.23.E2" : TYPE_OTHER
*/

---

Good one.

I would just check for the dot and then sscanf one way or the other tho -- if that fails: NAN. Who cares about those fancy smancy scientific numbers or whatever they are called.

Interesting - I am doing something akin to MK27. I do now ahead that it will be either float or int so I feel looking for the "." is efficient.

Anywho... New stumbling block in regards to what I was doing.

Code:

---

str = strtok(line, ",");

---

Is what I use, but I encountered a situation where there is an variable pass the last ","
i.e.
Line 1: 3.1, 2, 1
Line 2: 3, 14.25, 131

How can the last variable be read from Line 1 since it has no token to search?

I immediately looked at strtok(line, " ") and strtok(line, "\n") but got an compiler error (NULL is returned).

Secondly, tried strncpy, similar to strtok except precisely n characters are written into s1 and s1 is returned.
i.e.
strncpy(str, NULL, 1);

But this results in a seg fault - even though no error messages are given.


Any thoughts?

Quote:

---

Originally Posted by towed

Code:

---

str = strtok(line, ",");

---

Is what I use, but I encountered a situation where there is an variable pass the last ","
i.e.
Line 1: 3.1, 2, 1
Line 2: 3, 14.25, 131

How can the last variable be read from Line 1 since it has no token to search?

I immediately looked at strtok(line, " ") and strtok(line, "\n") but got an compiler error (NULL is returned).

---

Code:

---

strtok(line, " ,\n");

---

will break on a space, a comma, or a newline, this might work depending on if your line retains the newline or not. fgets does retain the newline.

Using strchr to differentiate a float from a int is hackish, IMO. Any parseable int is also a parsable float, so I'd use strtod (or equivalent) to parse floats.