setting a position in a binary file

I have a program that I am writing fro school. In this program, I create a Struct of type person[1000]. The struct holds various data char *, ints and floats. I have no problems writing to the binary file and displaying all the contents in the file.

In another function I need to open the binary file and retrieve particular data from it. The data I need lies in byte position 37-40. I would then need to increment the position by 56 bytes to get the data from 97-100 and basically keep doing this until the end of file is reached.

I have been reading up on fseek but it doesnt appear that I can actually pass it a manual position. There are only the three options for the whence. Do I need to use another option to set the position? Any help would be appreciated.

Thanks,

Michael

Apparently, you haven't read the documentation for fseek() all that closely. Either use fseek(file, 56, SEEK_CUR) after reading the data in position 37-40, or fseek(file, 97, SEEK_SET) to position relative to the start of the file.

If you're reading multiple data structures, each struct will be sizeof(your structure) bytes in length. So add an appropriate multiple and use positioning relative to the start of the file to position for the data in position 37-40.

Your description "manual position" is meaningless.

Note that if the structure contains pointers, the file will contain the value of the pointers, not whatever those pointers point at.

> The struct holds various data char *, ints and floats.
Therein lies the problem.

You can't just fwrite the contents of a struct, if it contains pointers.

Code:

---

struct foo {
  char *a;
} bar;
bar.a = malloc(100);
strcpy(bar.a,"this is a really long string");
fwrite(&bar,sizeof(bar),1,fp);

---

It will fwrite the pointer, but you won't see your string in the file.

I had actually tried that but I never did a fread to store the data and print it to the screen to verify that it was right. Way too many hours in front of the screen today. I've been running some tests and I am getting the values I want, now I just need to loop through the file to get all of my values. Do you know a good EOF alternative for binary files.

Quote:

---

Originally Posted by Salem

> The struct holds various data char *, ints and floats.
Therein lies the problem.

You can't just fwrite the contents of a struct, if it contains pointers.

Code:

---

struct foo {
  char *a;
} bar;
bar.a = malloc(100);
strcpy(bar.a,"this is a really long string");
fwrite(&bar,sizeof(bar),1,fp);

---

It will fwrite the pointer, but you won't see your string in the file.

---

That is actually not a problem, I have no problems storing the Struct in the binary file and retrieving it. The string name gets displayed fine for all my entries.

So exactly how are you serialising the struct then?

If you have (as a result) variable length records because you're outputting variable length strings, then using fseek() to go to a particular record suddenly becomes a far less easy thing to do.

Quote:

---

Originally Posted by generaltso78

That is actually not a problem, I have no problems storing the Struct in the binary file and retrieving it. The string name gets displayed fine for all my entries.

---

Really? That means you are using arrays rather than pointers (which is fine) or you have a critically flawed program with symptoms that have yet to emerge.

Are they actually char* s or fixed length char arrays?

Quote:

---

Originally Posted by generaltso78

Do you know a good EOF alternative for binary files.

---

Check that the return value from fread is what you expected it to be. If it's not, check ferror() and feof() like you would for a text file.

Yes, The structs are stored in a array and then written to binary file. The Struct contains a char[20], 4 ints and 6 floats. Every time a struct is stored the file size increments by 60 bytes. This is still uncharted territory for me, so bare with my terminology. I made a variable to contain my byte location for fseek. That way I can increment the position through a loop as I run through the file. I haven't figured out the exact increment size Im looking for yet, so I need to keep playing with it. My big concern is that when I increment my byte location variable than eventually fseek will move beyond the end of file and return a value that I am not looking for (would this happen before my EOF condition is triggered?)

This is where my program pushes my mental limits.

Code:

---

void totalRegularPay(EMPLOYEE person[]){
        int i = 0, bP = 36;
        float payrollTotals[1000];
        FILE *f;
        f = fopen("employeeRecords.bin", "rb");
       
        while (!feof(f)) {
                fseek(f, bP, SEEK_CUR);
                fread(&payrollTotals[i], sizeof(float), 1, f);
                printf("%.2f", payrollTotals[i]);
                i++;
                bP += 20;
        }
        fclose(f);
        pause;
}

---

The positions I want are at 36, 96, 156, 216, 276, etc..... Obviously, my sad attempt at += 20 only works for the next position. I'm thinking I may need to reset the binary position back to 0 after every fseek and then increment my bP(36) by 60.

sorry, obviously while (!feof(F)) is not going to work

Quote:

---

Originally Posted by generaltso78

The positions I want are at 36, 96, 156, 216, 276, etc..... Obviously, my sad attempt at += 20 only works for the next position. I'm thinking I may need to reset the binary position back to 0 after every fseek and then increment my bP(36) by 60.

---

As far as I understand your problem you want to get one specific float value from each record, right?
So why don't you read one full record after another (into a temporary struct object) and extract the value you are interested in? Should be very easy and straight forward.

Bye, Andreas

Quote:

---

Originally Posted by generaltso78

Yes, The structs are stored in a array and then written to binary file. The Struct contains a char[20], 4 ints and 6 floats. Every time a struct is stored the file size increments by 60 bytes. This is still uncharted territory for me, so bare with my terminology. I made a variable to contain my byte location for fseek. That way I can increment the position through a loop as I run through the file. I haven't figured out the exact increment size Im looking for yet, so I need to keep playing with it. My big concern is that when I increment my byte location variable than eventually fseek will move beyond the end of file and return a value that I am not looking for (would this happen before my EOF condition is triggered?)

---

You need to use the sizeof() operator to get the size of your struct. Manually adding the field sizes might not give the right size because of padding. You should then read a whole struct in and out of the binary file. Whilst you can read a field, it's easy to get it subtly wrong.
The equation is simply pos = sizeof(the_struct) * index;

Quote:

---

Originally Posted by generaltso78

This is where my program pushes my mental limits.

---

So your mental limits are pushed by the need to retrieve a particular field from a repeated series of structs? I've heard of pretty low limits in my time .....

Quote:

---

Originally Posted by generaltso78

The positions I want are at 36, 96, 156, 216, 276, etc..... Obviously, my sad attempt at += 20 only works for the next position. I'm thinking I may need to reset the binary position back to 0 after every fseek and then increment my bP(36) by 60.

---

If you really want to unnecessarily reduce performance of your program when reading, go for it.

Your basic problem is that you're not taking into account the fact that, after fread(), the file pointer will be incremented by the amount successfully read. Either that or you don't understand that relative and absolute positioning in a file are related by a subtraction.

If positioning with fseek() is beyond you, AndiPersti's approach will also work.

The problem turned out to be as easy as replacing my eof condition and implenting another fseek to run before the loop.

Code:

---

void totalRegularPay(EMPLOYEE person[]){
        int i = 0, bP =36;
        float payrollTotals[1000];
        FILE *f;
        f = fopen("employeeRecords.bin", "rb");
        fseek(f, 36, SEEK_SET);
        while (fread(&payrollTotals[i], sizeof(int), 1, f) == 1){
               
                printf("$%.2f", payrollTotals[i]);
                i++;
                bP += 60;
                fseek(f, bP, SEEK_SET);
        }
        fclose(f);
        pause;
}

---

I have only had it tested running through eight structs but i seems to be returning all the data I am looking for.

Why, if payrollTotals[1000] is of type float, are you looping on "fread(&payrollTotals[i], sizeof(int), 1, f)"?????

If you use SEEK_CURR rather than SEEK_SET, no need to keep track of the position using bP. Only need to account for the fact that fread() increases the file pointer (by the amount of data it reads).

Also, don't use magic values (36, 60, etc) if you can avoid it. 36 is presumably offsetof(your_struct, some_member_of_struct) and 60 is presumably sizeof(your_struct). Those simple changes will help your code work with different compilers.


Bear in mind that the size of structs (and struct members) is compiler dependent. So, if you build your code with a different compiler, odds are it will not be able to read your file correctly - the amount of data read and written by will vary with compiler.

Grumpy, well at least you acknowledge it

I guess I never changed the fread loop because float and int are the same byte size but if only for consistency I should, and if other compilers allocate more bytes to a float rather than an int, I would have probelms. What could I use instead of 36 to seek to that position. Is there a data type I can create to represent that many bytes or is there a way of telling it that I want the 6th value.

Quote:

---

Originally Posted by generaltso78

I guess I never changed the fread loop because float and int are the same byte size but if only for consistency I should, and if other compilers allocate more bytes to a float rather than an int, I would have probelms.

---

Your choice of words implies you think it is an irrelevant detail that can be ignored. That is an effective way of developing bad habits that WILL eventually bite you in real-world programming. Attention to detail is a primary weapon in the ongoing battle to generate fault-free code.

There are real-world compilers for which sizeof(int) != sizeof(float). Even better, there are real-world compilers for which sizeof(int) is configurable.

Quote:

---

Originally Posted by generaltso78

What could I use instead of 36 to seek to that position. Is there a data type I can create to represent that many bytes or is there a way of telling it that I want the 6th value.

---

I've already mentioned the offsetof() macro. It is specified in the standard header <stddef.h>.

The standard type used to represent sizes of types is called size_t. It is an unsigned integral type that, again, is specified in <stddef.h>.

@generaltso78
You need to post your actual struct, and the code you use to fwrite it to a file.
Without it, it's impossible to tell whether any of your fseek / fread code makes any sense at all.

Grumpy,

To play devils advocate, if everyone who programmed wrote fault-free code then would there not be a lot less programmers. That being said, I do want to be in the group of "good" programmers at some point. It is difficult because, my class doesn't even attempt to convey portability or stress good habits even when the bad ones produce the same result in that particular assignment. Granted this is the class which comes directly after the intro class. The assignment I have written, will "sadly" get me a 100 but only because it does everything the instructor wanted it to do. I could just stop there but I want it to be sound. I am still at the infant stage of writing/designing but I understand the importance of good fundamentals.

Here is the two structs that my program uses. I do have a EMPLOYEE person[10000];

Code:

---

typedef struct {
    int month;
    int day;
    int year;
}DATE;


typedef struct {
    char name[20];
    int age;
    float hourlyWage;
    float hoursWkd;
    float otHoursWkd;
    float regPay;//computed
    float otPay;//computed
    float totalPay;//computed
    DATE  payDate;                      //this is a struct inside a struct
}EMPLOYEE;

---

and here is the function where the person arrays are populated and written to file

Code:

---

    char sValidate[5] = "exit";
    FILE *f;
    int count = 0, i = 0;
    f = fopen("employeeRecords.bin", "ab");
   
    while (strcmp( sValidate, person[i].name)) {
        printf("Please enter name or type exit to return to main menu: ");
        scanf("%s", person[i].name);        //must use the '->' when passing by by refrence, must use '&' sign
        flush;


        if (!strcmp( sValidate, person[i].name))
            break;
       
        printf("\nPlease enter age of %s: ", person[i].name);
        scanf("%i", &person[i].age);
        flush;
        printf("\nPlease enter the hourlyWage for %s: ", person[i].name);
        scanf("%f", &person[i].hourlyWage);
        flush;
        printf("\nPlease enter the hours worked for %s: ", person[i].name);
        scanf("%f", &person[i].hoursWkd);
   
        if (person[i].hoursWkd > 40) {
            person[i].regPay = person[i].hourlyWage * 40.0;
            person[i].otHoursWkd = person[i].hoursWkd - 40.0;
            person[i].otPay = person[i].otHoursWkd * (person[i].hourlyWage * 1.5);
            person[i].totalPay = person[i].regPay + person[i].otPay;
        }
        else {
            person[i].totalPay = person[i].hoursWkd * person[i].hourlyWage;
            person[i].regPay = person[i].totalPay;
            person[i].otHoursWkd = 0.0; person[i].otPay = 0.0;
        }
        flush;
        printf("\nEnter 2 digit month: ");
        scanf("%i", &person[i].payDate.month);  //must use the '->' when passing by by refrence, must use '&' sign
        flush;
        printf("\nEnter 2 digit day: ");
        scanf("%i", &person[i].payDate.day);  //must use the '->' when passing by by refrence, must use '&' sign
        flush;
        printf("\nEnter 4 digit year: ");
        scanf("%i", &person[i].payDate.year);  //must use the '->' when passing by by refrence, must use '&' sign
        flush;
        fwrite(&person[i], sizeof(person[i]), 1, f);
        i++;
    }
    fclose(f);
           
}//end function loadPayRoll

---

I know this is probably not sound code, especially my exit implementation, but most of the sloppiness of my past assignments are not present in this one and the plan is to look back from future programs and know that a lot of issues above are not present in future programs.

Thanks,

Mike

Good - so you do have structs which don't contain pointers.

The first thing to look into is offsetof() which grumpy mentioned earlier. This will allow you to get to the field of interest in the first record.

After that, you use sizeof() (minus a bit - you can work out how much) to step from that field to the same field in the next record.

I implemented the offsetof and it seems to work beautifully, but am I not getting the same results. The first result (using offset) returns exactly what I want but when increment the position and use SEEK_CUR inside my loop the position is obviously not going where I want it to.

Code:

---

fseek(f, offsetof(EMPLOYEE, regPay), SEEK_SET);
        while (fread(&bufferTotals[i], sizeof(float), 1, f) == 1){
       
            payrollTotal += bufferTotals[i];
            fseek(f, sizeof(56), SEEK_CUR);
            i++;
        }

---

From my understanding, I should be getting the same results as before. As a follow up, for the best portability, is there an alternative to (56), like a way to represent sizeof(struct) - sizeof(struct value).

Thanks again,

Mike

Wow, I just got back. looked at my problem again and wow, cant believe I had sizeof in my fseek, sorry for wasting your time if you read my last post

> is there an alternative to (56), like a way to represent sizeof(struct) - sizeof(struct value).
Well yes, sizeof(struct) - sizeof(struct value) should work, given the right things to do sizeof on.