Lab - Finding Anagrams

In this writeup it is shown how the UNIX philosophy of combining simple programs using pipes can solve an interesting problem. The problem is to find all anagrams in a file containing words. The UNIX tools that will be used are: (1) a small C program that uses library calls to do string operations and sorting, (2) the UNIX sort command, (3) a small AWK program (maybe 2), and finally the UNIX head or tail command.

The directory for this lab is tux:/home/kschmidt/public_html/CS265/Labs/Anagram

Your job is to write a single script to pull all of this together. Most of the machinery mentioned above is given to you, in this directory, or right here in the sheet.

Problem Statement and Solution

Two words are anagrams if one can be obtained from the other by permuting the letters. For example, the words stop and pots are anagrams since the letters "s", "t", "o", and "p" in stop can be rearranged to obtain the word pots.

A naive way to test if two words are anagrams is to produce all permutations of the letters of one of the words and compare each possible permuted sequence of letters to the other word. If any sequence is equal to the other word then they are anagrams. For example, to determine if "ant" and "tan" are anagrams, we first generate the six possible permutations of the letters in ant:


and compare each to tan. Since the 5th permuted sequence is equal to tan, ant and tan are anagrams.

The problem with this approach is that as the number of letters increases, the number of permuted sequences gets very big. For example, there are 3,628,800 permutations of a 10 letter word.

A better approach to test if two words are anagrams is to sort the letters in each word and then compare the sorted words. If the words are anagrams then they must have the same letters and if they have the same letters then they are anagrams. Therefore this test determines if two words are anagrams with one comparison. For example, if we sort the letters in stop and pots, we obtain the letter sequence opts for both words, which shows that they are anagrams.

Given a file containing a list words we want to find all of the anagrams. We will do this by dividing the words into anagram classes. An anagram class containing a particular word is the collection of words in the input list that are anagrams of the given word. Note that if two words are anagrams to a given word then they are anagrams to each other. Thus any two words in an anagram class are anagrams. Moreover, all of the words in the input file that are anagrams to any of the words in the anagram class are contained in the anagram class.

A naive way to find all of the anagram classes is for each word to search through all of the other words and check each to see if it is an anagram. This approach is very time consuming. A better approach would be to:

  1. For each word in the file sort the letters to produce a key.
  2. Sort all of the words in the input file by their key.
  3. In the sorted list find all adjacent words with the same key (these are the anagram classes).

This will be implemented in UNIX in the following steps:

  • Use a C program called sign to read the input file (the dictionary on our machines, /usr/share/dict/words) and to produce an output file where each line contains the key (the letters of the word sorted) followed by the word. This program will read from standard input and write to standard output so that we do not have to specify the input file in the program and we can use the output in a pipe. The name of the the executable program will be sign . To create this:
  • gcc -osign sign.c
  • The temporary file containing words and keys produced in (1) will be sorted using the UNIX sort command.
  • Use an AWK program to collect all adjacent words with the same key. The result will be a file containing an anagram class per line. The file containing the awk program is squash.awk .
  • Here is the UNIX command to do this. The file /usr/share/dict/words contains a list of just under 100,000 words. To verify this use the command wc.

    % ./sign < /usr/share/dict/words | sort | awk -f squash.awk > out

    The output will be in the file out. The awk command can be put into a file called squash. Such a file containing UNIX commands is called a shell script and is itself interpred as a command as soon as the execute permission has been set. The resulting UNIX pipe, which uses the shell script squash instead of the call to awk, becomes

    % ./sign < /usr/share/dict/words | sort | ./squash > out

    To find the largest anagram classes, we can first count the number of words in each class (I'm thinkin' another AWK script here) and then sort the file by these counts. Here is a UNIX command to do this:

    % awk '{ print NF " " $0}' < out | sort -n | tail

    In this command the AWK program is given between the single quotes. The utility tail prints the last 10 lines of a file. Here is some sample output (from an old dictionary):

    4 glare lager large regal
    4 lament mantel mantle mental
    4 lascar rascal sacral scalar
    4 leap pale peal plea
    4 leapt petal plate pleat
    4 least slate stale steal
    4 mate meat tame team
    4 pare pear rape reap
    4 resin rinse risen siren
    5 caret carte cater crate trace

    Students should try these commands to see if they can produce the same output. Note that you can view the output of any intermediate stage by piping the results through more. For example:

    % sign < /usr/share/dict/words | more

    You can quit more by typing "q" ("man more" for more info on more).


    Submit all the files needed for this lab (and only those needed).

    Your top-level script will put this all together; it'll check for the needed files, compile the C program (and anything else you need to have done), and put it all together, to print (to the screen) the 10 largest anagram classes.