Jumbled-word puzzler solver

A program to help solve Jumbled word puzzles

The user interface is pathetic. I am not a UI/UX specialist, and I don't want to put the effort into learning how to do decent user interfaces.

Building

Has no dependencies outside of the Go standard library.

$ go build runserver.go

Executing

It's a webapp that runs correctly probably only under Linux, maybe under a BSD. You DO NOT want to expose this to the internet. I made no effort to keep it safe, or to have the web app part validate user input. It is a security risk of unknown magnitude.

$ ./runserver -d /usr/share/dict/words -s stopwords.dat

The file specified with "-d" is a text file, one word per line. This is the dictionary, the words the server will choose from to solve the puzzle.

The file specified with "-s" is a text file, one word per line. These are "stop words", words that won't be considered in solving the puzzle. This repo includes a stopwords.dat file, but its contents aren't the usual, very common short words that a stopwords list usually has. It contains a mix of words that the Linux words file has, but aren't English, like "viii", or "ct" viii and a few other strings are Roman Numerals often used to number introductions to books. ct and others are US postal state abbreviations. The words file includes them because it's mainly used to get spellchecks correct. The other category of stop words include spavined, damn, English words that the Jumble authors won't use for fear of either offending or mystifying their audience.

You'll have to access the server with a URL like http://localhost:8012/jumble

Design and Algorithm

I used HTML and JavaScript to build the user interface because I don't want to invest the time to figure out how to do a native UI. This damages the UI, but I sincerely doubt anyone will care, as nobody will try it.

It works by sorting the letters of every word in a file (/usr/share/dict/words, for example) that has one word per line. The words "meat", "team" and "meta" all end up alphabetized as "aemt". It makes a hashtable/dictionary/associative array using an alphabetized word as a key, and an array of words that alphabetize to that key as the value.

The program can alphabetize the characters in a jumbled input word, and use the resulting string to look up all words that potentially match. A jumbled word can match more than one unjumbled alternate word. That lets it "unjumble" words.

The same thing happens in the final solution. Marked characters of the un-jumbled words get used to create a key to look up possible matches in the dictionary.

One complication happens when a letter-alphabetized key matches more than one word. More than one marked character is possible. The code creates keys by iterating through all the possible combinations. That is, more than one possible unjumbled word can cause the program to deal with more than one key for the final solution.

Multi-word solutions also cause complications. The code deals with it by creating all possible combinations of the characters for the first word, then recursively treating the leftover characters as the other word(s) of the solution, making combinations for them. A single set of marked characters from the unjumbled words can end up as many partitioned-into-words keys for the solution. Each partitioned-into-words combination has to match a dictionary entry (remember them from above?) so may of the paritioned-into-words keys can be discarded. Each of those dictionary answers can be a group of words that alphabetize to the key.

Example

1. Fire up the web app: ./runserver -d /usr/share/dict/words
2. Use a web browser to access the HTML generated by runserver: http://localhost:8012/jumble
3. In the web browser, select the number of words from the pull-down list
4. Set the number of letters in each individual word
5. Check the letters of each solved word that will be carried forward
6. Enter the jumbled letters into the individual fields of each word
7. Click "Unjumble Words". The web app will find all the real words whose letters can be moved around to match the jumbled words you enter.
8. Sometimes, the letters in the jumbled words don't match any words in the dictionary you used. You can unjumble the word youself, and click the "use as-is" checkbox. I found the Jumble authors using scrambled letters for "monkees", the mid-60s synthetic rock band, in two separate puzzles.
9. Select the number of words in the solution, and the size(s) of the words in the solution. This is a crap interface: you enter a number for the number of words, and space-separated numbers for the number of characters in each of the words of the solution.
10. Click the Solve button. The webapp will use the letters in the unjumbled words to create combinations of sub-words as specified in the space-separated numbers. Some times more than 1 word in the dictionary will match the jumbled characters, so some puzzles will have several lists of letters carried forward. The webapp will create combinations of each of the several lists of letters. The webapp will try to find words in the dictionary file that have the same letters as the combinations of letters from the lists of letters carried forward.

Oddly there can be only one solution for a given puzzle, or hundreds can exist.