Sliding Window Search

Problem #141

Tags: data-compression strings external-file

Who solved this?

The part of LZ77 compression algorithm is about searching of a substring in the text directly preceeding it.

Let us see an example with the following text:

in the town where i was born lived a man who sailed to sea
and he told us of his life in a land of submarines

starting from the words and he told we can discover the following repeating fragments:

This exercise asks you to find the longest of preceeding matches for several positions in the text. Note that while for real compression algorithm it is crucial to implement this efficiently, for the given exercise it is not needed - it is only important to get the idea of encoding with back-references.


Preceeding match should be sought for only within the block of preceeding 4096 bytes and the match of no more than 15 bytes is required. This will allow us to encode the position of the match with 12-bit value and the length of the match with 4-bit value - i.e. only two bytes in total.

Position (or rather offset) of the match should be calculated relatively backward to current position, so that the first preceeding symbol has the index 0, the next 1 and so on up to 4095 (i.e. it is the difference of positions of the current substring and its previous match decreased by 1).

Returning to our example above:

string   off-len

an        20-2
d_        10-2
he to     58-5
l         19-1
d_         7-2
u          0-0

In the last case the length 0 shows that even substring of the length 0 could not be found.

If more than one prefix of the same length could be found, please choose the smallest offset.

As a source text to fetch input data from please use the sample data file:

Click with right button and choose 'save as' to download file doyle.txt

It is not large so it is recommended to store it in a single long byte array in your program to handle easily.

Input/Output specification

Input data have the number of testcases N in the first line.
Second line contains exactly such amount of positions Pi for i=1..N for which the search should be done.
Answer to each case should be 2-byte hexadecimal with its lowest 12 bits (or 3 digits) holding the offset and the highest 4 bits (one digit) storing the length (so that offset 20 and length 2 are encoded as 2014, while 58 and 5 as 503A).

Example (with real doyle.txt file):

input data
22255 22947 36587 9700 31780 21915 4676 25611 34708 13155 7630 46121 20855

10D7 59A4 3167 3007 301E 6AE7 3334 4721 3C77 7AD1 7F21 214A 3B04

Here the substring from position 22255 reads kable and no better match could be found before rather than single letter k at position 22039.

The substring from position 22947 is led to listen and match of 5 letters is found at position 20478 inside the phrase disentang[led t]he most.

You may trace other matches manually to get the idea better.

You need to login to get test data and submit solution.