Line matching

Hey,

I have a large textfile that contains data that changes over time. Each 'timestep', it starts a new section of the file, beginning with a ling that conatins a single integer value and nothing else. These lines are irregularly spaced since each timestep contains different amounts of information as the system evolves. I need a function that will count the number of timesteps (ei, count the number of lines that contain a single integer value and nothing else). How can I do this?


I though about using fgets to read in each line, but I have never worked with strings like that so I wouldn't know how to test the result properly. Is there a library function that can tell you if everything in a given string is a number with no decimal place?

Thanks

Look into the scanf (sscanf, fscanf, etc..) family of functions - they read formatted input. So what you would do is ask it to scan a decimal integer. If it's a match, it'll return 1 and place the number in the variable you specified. If not, it returns 0, and you'll know it's not that line.

An important element of this is: how much of a role can you play in the structure of the file? It sounds as if the answer is "none" because it is output by some other program not written by you, but I just want to check.

Originally Posted by KBriggs

I though about using fgets to read in each line, but I have never worked with strings like that so I wouldn't know how to test the result properly. Is there a library function that can tell you if everything in a given string is a number with no decimal place?

Very likely, using fgets or fread is the way to go. I would not bother looking for some special library to do this -- learning to implement it won't be any easier than just learning to parse a string, which if you have any desire to use the C language for programming anything, you should really get comfortable with strings.

One thing I have done for tasks like this is to write a custom "get line" type function using fread:

Code:

int fileline_base (int fp, char *line) {
      int c=0;
      char byte;
      while (fread(&byte, 1, 1, fp) {
           if (byte == ' ')   etc [ character by character processing ]
           line[c] = byte; c++;
      }
      line[c] = '\0';  /* must */
      return (some informative value, eg, the type of line)
}

I called this "fileline_base" because it is possible you will want to have different functions to process different kinds of lines *if* you can predict what kind of line is coming next based on the file structure. Otherwise, you could use fileline_base to get the line, but use the character by character processing to identify the "kind" of line, and pass it on to (eg) fileline_timestamp, which might break it up and put appropriate details into a struct.

There's a lot of useful things that working byte by byte makes possible. You could use fgets and then process the line the same way in a loop, but it is probably more efficient to do it on the initial read.

You could also use fgets() with sscanf(). That would allow you to distinguish one kind of line from another -- if the first sscanf fails, move on to the next one. That will be much less prone to error (and more flexible) than if you just fscanf the file.

Is there something I have to link at compile time to include sscanf? I #included <stdio.h> but I get an error saying:

/tmp/ccuKnQDb.o: In function `countTimeSteps':
chargestate.c.text+0x33): undefined reference to `sccanf'
collect2: ld returned 1 exit status

EDIT: LOL, a sad face in an error message


Also, will this work? I am unsure of my error-checking:

Code:

int countTimeSteps(FILE *input)
{
  char *line;
  int check,x,N_timeSteps;
  N_timeSteps = 0;
  
  while (fgets(line, MAXNUM, input) != NULL)
  {
    check = sccanf(input,"%d",&x);
    if (check != 0 && check != EOF)
    {
      N_timeSteps++;
    }
  }
  return N_timeSteps;
}

Spell it correctly: sscanf.

Remember to check spelling if you get an odd linker error like this.

Heh, silly me

No, it doesn't work - it gives me 0, and I get a warning:

chargestate.c: In function ‘countTimeSteps’:
chargestate.c:15: warning: passing argument 1 of ‘sscanf’ from incompatible pointer type


Which I don't know how to intepret

EDIT: it is reading in the lines, because adding a puts(line) prints the output file just fine...

You are probably after fscanf().

sscanf() reads from a string (char pointer), and you are passing it a FILE pointer. That's why it says incompatible pointer type.

Or... you should be passing it line instead of input. line is the string that was read, input was the file it came from. You've already read the line, so you wouldn't want fscanf reading more lines. fscanf is to validate the contents of a line.

Haha yes of course. Sorry about that.

Didn't read the whole code snippet.

oo, perfect. Wrong pointer haha. Works now

ACtually, I have another question about this. Here is my working code:

Code:

int countTimeSteps(FILE *input)
{
  char line [MAXNUM];
  int check,x,N_timeSteps;
  N_timeSteps = 1; //since there is one for time == 0, start at 1
  fgets(line, MAXNUM, input);
  
  while (fgets(line, MAXNUM, input) != NULL) //read in file 1 line at a time
  {
    //puts(line);
    check = sscanf(line,"%d",&x); //x is useless, only there to make sscanf stop whining about arguments
    if (check != 0 && check != EOF) //if the line contains an integer value at the start
    {
      ++N_timeSteps;
    }
  }
  return N_timeSteps;
}

My question is this:

right now, the code will count any line that starts with an integer. Is there a way to specify that it count only lines that contain ONLY a single integer, rather than lines the simply begin with one? For the current case it doesn't matter, but it would help make the code a little more robust.

Originally Posted by KBriggs

right now, the code will count any line that starts with an integer. Is there a way to specify that it count only lines that contain ONLY a single integer, rather than lines the simply begin with one? For the current case it doesn't matter, but it would help make the code a little more robust.

By "integer" you mean one number (eg, 234123) and not a single digit? Because a single digit will be simple.

Otherwise, depending on what else could be in the string, you could add a second parameter to sscanf and check the number of items returned.


BTW, you can use

Code:

sscanf("%*d");

The * means ignore, so you don't need x. Since you are not keeping the number, you could also test this mo' better:

Code:

if ((line[0] >= '0') && (line[0] <= '9'))

The ASCII values of digits are guaranteed to be sequential.

If you want to make sure there's nothing else, you could, instead of sscanf, use strtol -- strtol gives you a pointer to "what's left over".

Looking at strtol in the documentation, I really don't understand how it would be used

Got a one-liner that encapsulates the necessary test for me? :P

Code:

long strtol(const char* s, char** endp, int base);

You pass your line in as s, pass some pointer in as endp, and (if you want) the base of the number in as base (most likely 10).

The function will return the value of that number, and endp will be null unless there was data in your string IN ADDITION to the number. Make sense?