Problem with Huffman Algo implementation...

[gandhi_rohit]

I am trying to implement the Huffman algorithm,
The problem that I am facing is that while decompressing I get almost
95% of the file correct,but it fills up some of the junk characters in between
the output file at different places.Also I do not see any regular pattern for
the places and the characters that are filled up in the file,when i tried this on
various files.
but it does work perfectly for smaller files,say upto 1KB files,the problem increases with
larger files.

I understand that there is no point sending the whole code(which is quite lengthy)

A general idea abt the data structures used...
--A 2-d array and a linked list initially contains the character with their respective frequencies.
--This linked list is then converted to a binary tree,based on the frequency of the characters.
--A struct which contains the 1/0 traversed code of the character,along with the character.

Using this struct the compressed file is produced,with bits packed into a byte.

Can u tell me why that could be happening ?
Also can u suggest me a way of debugging the code,so that I can know where the exact problem lies,
as it is not possible to use Trace i.e F7 for larger files.

Tell me if u need more information on this.

I am using TurboC 3.0 compiler.

Thanking You,
Rohit.

[Salem]

> Can u tell me why that could be happening ?
Do you read the file using the same method?
I'm thinking that your first scan of the file doesn't count all characters, and when you subsequently perform a lookup, you generate a bogus code.

Check that your tree is a valid tree - like all leaf nodes have NULL pointers. Write a debug routine to print a text representation of this tree to make sure its valid.
You can also do this for your frequency distribution array.

[gandhi_rohit]

Salem,

I had checked the whole tree for small files,that is working perfectly,Null pointers r properly assigned,the compressed file is also perfect.

Pls suggest smthing else.
Thanx for ur help.