Thread: Reading a comma separated file

Unless all the lines are the same length, you'd have to record the position of each newline so you could seek to the right position in the file. Or you could read in the whole file into an array and address a specific line by the array element.

Well, if every line is, say, 10 characters long, then to access the fifth line you'd fseek to position 40, then read 10 characters in, probably with fgets():

Code:

void get_line(FILE *fp, int n, char *s) {
    fseek(fp, 10 * n, SEEK_SET);
    fgets(s, 9, fp);  /* not 10, to exclude the newline */
}

Maybe I shouldn't be using C?

Any language would have the same problem, unless it read in the file and recorded the positions of the newline for you.

Anyway, is it possible for me to read the entire file into an array that is still organized the way my file is set up?

Yes, read in each line with fgets() and store it in an array:

Code:

char data[LINES][BUFSIZ];
int line = 0;
FILE *fp = fopen("file.txt", "rt");

while(fgets(buffer[line++], BUFSIZ, fp));

/* now data[] contains every line of the file; lines is how many lines were read in */

(Note that I don't do any error checking in that code.)

Once you have picked the line you want, you could use sscanf() to parse numbers or whatever from the string.

[edit] Yes, I meant data[]. Oops. [/edit]

Originally Posted by nwr

I have a bunch of comma separated files, and each field is in quotation marks. Individual records are separated by carriage returns. I want to be able to to go a given record (a specific line in the file) and the read the data in a specific field in that record into an int or a char depending on whether the entry in that field is a number or text.

So, how can I do this?

Declare a char array large enough for the longest line, and a counter initialized to zero. Read each line in a loop, incrementing the counter each time -- just discard each line of text you don't want and break the loop when you've gotten the the line you wanted. Then parse the line.

But this is where it may get a little sticky. Since each field is in quotation marks, does this mean that commas may be embedded in a field?

Where you wrote buffer[line++], I think you meant data[line++], right?

He put that there on purpose, to make sure you did your share of homework and not just lounge around waiting to be spoonfed.

Adding in code to handle quotes shouldn't be an issue. It's even feasible to check for escaped quotes (eg. {\"} as a literal quote) without hurting performance. You're going to iterate over those characters anyway in a naive comma search.

Do you absolutely have to use C? Problems like these just scream "Perl". There are tons of CSV- and DSV-parsing modules out there.

In fact I managed to throw something together in 8 minutes, most of which spend find(1)'ing the CPAN module, and I haven't even read chapter 5 of the Llama Book yet.