Difference between revisions of "CISC220 F2023 Lab8"

Revision as of 22:13, 2 November 2023

Lab #8

1. File compression with Huffman coding

As discussed in class on Nov. 2, tries and Huffman coding can be used to compress files/messages by analyzing character frequencies and choosing bitcodes accordingly.

A nearly-complete Huffman class is defined for you in huffman.hh (code here). It uses a TrieNode class that stores parent and child links as well as the character being encoded and its frequency in the file being compressed.

The main work of Huffman happens in the following functions:

• compress()
  • Reads input file character by character and counts occurrences (aka computes frequency) for each, storing the results by ASCII index in the char_frequency vector
  • Applies the Huffman trie-building algorithm presented in class 11/2 in build_optimal_trie(). This function is TO BE WRITTEN BY YOU
  • Creates code look-up tables (string to char decompression_map and char to string compression_map) from the optimal trie in compute_all_codes_from_trie(). This function is TO BE WRITTEN BY YOU
  • Writes decompression_map as a header to the output file
  • Re-reads input file character by character and, using compression_map, writes encoded versions to the output file body (in ASCII or binary depending on do_binary flag)
• decompress
  • Reads the code look-up table from header of input file to decompression_map
  • Reads rest of input file bitstream, gets character for each variable-length code read using decompression_map, and writes decoded versions to output file. The ASCII part of this function, ascii_decompress_body() is TO BE WRITTEN BY YOU

The "forest of tries" discussed in lecture is implemented in the Huffman class with an STL priority queue of TrieNode pointers (each representing the root of a subtrie), ordered on the combined frequency of each subtrie. This priority queue is called trie_pq

Several major functions are unfinished. In order to separate them from the finished ones, they are located in main.cpp -- this is where you will be writing code.

The Huffman member functions to finish in main.cpp:

• [2 points] merge_two_least_frequent_subtries() : This function should do the following (there are further hints in the comments):
  • (1) remove trie in PQ with SMALLEST frequency, assign to first
  • (2) remove trie in PQ with NEXT SMALLEST frequency, assign to second
  • (3) make new trie node (new_root) and plug first and second tries in as children, set new_root's frequency to first->frequency + second->frequency, and insert in new_root in PQ
• [1 point] compute_all_codes_from_trie(TrieNode *T) : The function should recursively traverse the binary tree (trie) rooted at T to determine the huffcode string at every leaf. If T is a leaf, make sure to execute the compression_map and decompression_map assignments
• [1 point] calculate_huffman_file_size() : Return how many bits Huffman-coded version of the file takes. This is the sum of the frequency * huffcode length over every leaf

2. Testing [1 point]

• DOI (about 1K total words), BTS (~14K words), and GE (~185K words) require 8588 bits, 83964 bits, and 1013761 bytes, respectively according to the Unix ls command
• How many *bytes* are required for each of these files after compression when you set the -debug flag in (a) "custom" form (shortest fixed-length code) and (b) using the variable-length Huffman code that your completed class generates?

3. Submission

• Make a PDF file <Your Name>_Lab9_README.pdf with your answers to the testing questions above
• Rename your code directory <Your Last Name>_Lab9 and create a single tar/zip/rar file out of it named <Your Last Name>_Lab9.tar (or .zip or .rar, etc.).
• Submit it in Canvas by midnight at the end of Tuesday, November 16