How to find common strings among two very large files?

Question

I have two very large files (and neither of them would fit in memory). Each file has one string (which doesn\'t have spaces in

Answer 1

Is there any order to the data in the files? The reason I ask is that though a line by line comparison would take an eternity, going through one file line by line whilst doing a binary search in the other would be much quicker. This can only work if the data is sorted in a particular way though.

Answer 2

To do it in windows, its pretty simple .. lets say , you have two files A and B. 'A' files contains the strings you want to search in file B. just open command prompt and use the following command

FINDSTR /G:A B > OUTPUT

this command is pretty fast and can compare two files very efficiently. The file OUTPUT will contain the strings common in A and B.

if you want to perform the OR operations (strings in B other than A) then use

FINDSTR /V /G:A B > OUTPUT

Answer 3

I would sort each file, then use a Balanced Line algorithm, reading one line at a time from one file or the other.

Answer 4

I would load both files into two database tables so that each string in the file became a row in the table and use SQL queries to find duplicate rows using a join.

Answer 5

Depending on how similar the entries within one file is, it might be possible to create a Trie (not tree) from it. Using this trie you can iterate the other file and check each entry if it is inside the trie.

When you have more than 2 files, iterate over one file and build a new trie from the matches. This way the last trie you have will contain all the matches that are contained in all files.

Answer 6

A hash based solution might look like this (in python pseudocode):

hashes = dict()
for file in files:
    for line in lines:
        h = md5(line)
        hashes[h] += 1

Then loop over again, printing matching lines:

for file in files:
    for line in lines:
        h = md5(line)
        if hashes[h] == nfiles:
            print line
            del hashes[h]  # since we only want each once.

There are two potential problems.

1. potential hash collisions (which can be mitigated some, but is a risk. )
2. needs to be able to handle a dict (associative array) of size: |uniq lines in all files|

This is O(lines * cost(md5) ).

(if people a fuller python implementation, it's pretty easy to write, I don't know java though!).