Compress a struct into a binary file? [C] - c

Question

This is part of my homework that I'm having difficults to solve.
I have a simple structure:
typedef struct Client {
char* lname;
unsigned int id;
unsigned int car_id;
} Client;
And the exercise is:
Create a text file named as the company name and then branch number with txt extention.
the file contain all clients' details.
The file you created in exercise 1 will be compressed. as a result, a binary file be created with .cmpr extention.
I don't really have an idea how to implement 2.
I remember at the lectures that the professor said we have to use "all" the variable, with binary operators (<< , >> , | , &, ~), but I don't know how to used it.
I'm using Ubuntu, under GCC and Eclipse. I'm using C.
I'd be glad to get helped. thanks!

Answer 1

Let's say the file from step 1 looks like:
user1798362
2324
462345
where the three fields were simply printed on three lines. Note that the above is the text/readable (i.e. ASCII) representation of that file.
Looking at the contents of this file in hex(adecimal) representation we get (with the ASCII character printed below each byte value):
75 73 65 72 31 37 39 38 33 36 32 0a 32 33 32 34 0a 34 36 32 33 34 35 0a
u s e r 1 7 9 8 3 6 2 nl 2 3 2 4 nl 4 6 2 3 4 5 nl
here nl is of course the newline character. You can count that there are 24 bytes.
In step 2 you have to invent another format that saves as many bits as possible. The simplest way to do this is to compress each of the three fields individually.
Similar to where the text format uses a nl to mark the end of a field, you also need a way to define where a binary field begins and ends. A common way is to put a length in front of the binary field data. As a first step we could replace the nl's with a length and get:
58 75 73 65 72 31 37 39 38 33 36 32 20 32 33 32 34 30 34 36 32 33 34 35
-- u s e r 1 7 9 8 3 6 2 -- 2 3 2 4 -- 4 6 2 3 4 5
For now we simply take a whole byte for the length in bits. Note that 58 is the hex representation of 77 (i.e. 11 characters * 8 bits), the bit length of lname',20hex equals 4 * 8 = 32, and30is 6 * 8 = 48. This does not compress anything, as it's still 24 bytes in total. But we already got a binary format because58,20and30` got a special meaning.
The next step would be to compress each field. This is where it gets tricky. The lname field consists of ASCII character. In ASCII only 7 of the 8 bits are needed/used; here's a nice table For example the letter u in binary is 01110101. We can safely chop off the leftmost bit, which is always 0. This yields 1110101. The same can be done for all the characters. So you'll end up with 11 7-bit values -> 77 bits.
These 77 bits now must be fit in 8-bit bytes. Here are the first 4 bytes user in binary representation, before chopping the leftmost bit off:
01110101 01110011 01100101 01110010
Chopping off a bit in C is done by shifting the byte (i.e. unsigned char) to the left with:
unsigned char byte = lname[0];
byte = byte << 1;
When you do this for all characters you get:
1110101- 1110011- 1100101- 1110010-
Here I use - to indicate the bits in these bytes that are now available to be filled; they became available by shifting all bits one place to the left. You now use one or more bit from the right side of the next byte to fill up these - gaps. When doing this for these four bytes you'll get:
11101011 11001111 00101111 0010----
So now there's a gap of 4 bits that should be filled with the bit from the character 1, etc.
Filling up these gaps is done by using the binary operators in C which you mention. We already use the shift left <<. To combine 1110101- and 1110011- for example we do:
unsigned char* name; // name MUST be unsigned to avoid problems with binary operators.
<allocated memory for name and read it from text file>
unsigned char bytes[10]; // 10 is just a random size that gives us enough space.
name[0] = name[0] << 1; // We shift to the left in-place here, so `name` is overwritten.
name[1] = name[1] << 1; // idem.
bytes[0] = name[0] | (name[1] >> 7);
bytes[1] = name[1] << 1;
With name[1] >> 7 we have 1110011- >> 7 which gives: 00000001; the right most bit. With the bitwise OR operator | we then 'add' this bit to 1110101-, resulting in 111010111.
You have to do things like this in a loop to get all the bits in the correct bytes.
The new length of this name field is 11 * 7 = 77, so we've lost a massive 11 bits :-) Note that with a byte length, we assume that the lname field will never be more than 255 / 7 = 36 characters long.
As with the bytes above, you can then coalesce the second length against the final bits of the lname field.
To compress the numbers you first read 'em in with (fscanf(file, %d, ...)) in an unsigned int. There will be many 0s at the left side in this 4-byte unsigned int. The first field for example is (shown in chunks of 4 bit only for readability):
0000 0000 0000 0000 0000 1001 0001 0100
which has 20 unused bits at the left.
You need to get rid of these. Do 32 minus the number of zero's at the left, and you get the bit-length of this number. Add this length to the bytes array by coalescing its bits against those of previous field. Then only add the significant bits of the number to the bytes. This would be:
1001 0001 0100
In C, when working with the bits of an 'int' (but also 'short', 'long', ... any variable/number larger than 1 byte), you must take byte-order or endianness into account.
When you do the above step twice for both numbers, you're done. You then have a bytes array you can write to a file. Of course you must have kept where you were writing in bytes in the steps above; so you know the number of bytes. Note that in most cases there will be a few bits in the last byte that are not filled with data. But that doesn't hurt and it simply unavoidable waste of the fact that files are stored in chunks of 8 bits = 1 byte minimally.
When reading the binary file, you'll get a reverse process. You'll read in a unsigned char bytes array. You then know that the first byte (i.e. bytes[0]) contains the bit-length of the name field. You then fill in the bytes of the 'lname' byte-by-byte by shifting and masking. etc....
Good luck!