For several months I’ve been playing a “hangman” style game on Mastodon, HangTrek, where words and phrases are from the various Star Trek television shows and films. The game starts with a phrase like

_ _ _ / _ _ _ _ _ _ _ / _ _ / _ _ _ / _ _ _ _ _ _ / _ _ _ _ / _ _ _ / _ _ _ _ _ _ _ _ _ _

Wrong guesses: none

The underscore (_) characters are unknown (to the players) letters, and the forward slashes (/) are word separators. For each round a poll with 4 choices is run for some amount of time, typically 24 hours. The poll winner is the guess for that round. If the winner matches one of the unknown letters, for the next round, that letter is filled in. The the winner isn’t one of the unknown letters, that’s noted below the puzzle in the following rounds, so it doesn’t get guessed again.

For example, if ‘E’ is a valid guess, in the next round it will show up like this:

_ _ _ / _ _ E _ _ _ _ / _ _ / _ _ _ / _ _ _ _ _ _ / _ _ _ _ / _ _ _ / _ _ _ _ _ _ _ _ _ _

Wrong guesses: none

In another example, if ‘Q’ was the poll winner but doesn’t match any letters, it’s noted in the wrong guesses.

_ _ _ / _ _ _ _ _ _ _ / _ _ / _ _ _ / _ _ _ _ _ _ / _ _ _ _ / _ _ _ / _ _ _ _ _ _ _ _ _ _

Wrong guesses: Q

Some number of Star Trek television and film scripts are available for download, so I took to searching them with regular expressions. For example, for a 10 letter word where none of the letters are known, I’d search for

[[:alpha:]]\{10\}

in the scripts, using a tool like grep.

Over time, I realized I could write a program to create these regular expressions, so I did. I have a long history of turning my non-programming homework into programming homework, back to my college days 😃

For a whole phrase, there are some things we want to do:

  • anchor words with ‘\<’ and ‘\>’, so we don’t match words within larger words
  • allow for multiple spaces between words in a phrase, using ‘\+’
  • shrink the original ‘[:alpha:]’ character class as letters are guessed. For correct answers, we see them in the phrase. For incorrect letters, add a way to specify them, so they’re also eliminated from the character class

So, I wrote a Python program hangman-regexp.py to accomplish this. It takes the phrase as a command line argument, and you can optionally specify wrong guesses with a -x Q argument.

hangman-regexp.py

For the first phrase above, the program runs like this:

hangman-regexp.py '_ _ _ / _ _ _ _ _ _ _ / _ _ / _ _ _ / _ _ _ _ _ _ / _ _ _ _ / _ _ _ / _ _ _ _ _ _ _ _ _ _'

which provides this output:

\<[abcdefghijklmnopqrstuvwxyz]\{3\}\>[[:space:]]\+\<[abcdefghijklmnopqrstuvwxyz]\{7\}\>[[:space:]]\+\<[abcdefghijklmnopqrstuvwxyz]\{2\}\>[[:space:]]\+\<[abcdefghijklmnopqrstuvwxyz]\{3\}\>[[:space:]]\+\<[abcdefghijklmnopqrstuvwxyz]\{6\}\>[[:space:]]\+\<[abcdefghijklmnopqrstuvwxyz]\{4\}\>[[:space:]]\+\<[abcdefghijklmnopqrstuvwxyz]\{3\}\>[[:space:]]\+\<[abcdefghijklmnopqrstuvwxyz]\{10\}\>

For the second example, where we’ve successfully guessed ‘E’, here’s the command and the output:

hangman-regexp.py '_ _ _ / _ _ E _ _ _ _ / _ _ / _ _ _ / _ _ _ _ _ _ / _ _ _ _ / _ _ _ / _ _ _ _ _ _ _ _ _ _'
\<[abcdfghijklmnopqrstuvwxyz]\{3\}\>[[:space:]]\+\<[abcdfghijklmnopqrstuvwxyz]\{2\}e[abcdfghijklmnopqrstuvwxyz]\{4\}\>[[:space:]]\+\<[abcdfghijklmnopqrstuvwxyz]\{2\}\>[[:space:]]\+\<[abcdfghijklmnopqrstuvwxyz]\{3\}\>[[:space:]]\+\<[abcdfghijklmnopqrstuvwxyz]\{6\}\>[[:space:]]\+\<[abcdfghijklmnopqrstuvwxyz]\{4\}\>[[:space:]]\+\<[abcdfghijklmnopqrstuvwxyz]\{3\}\>[[:space:]]\+\<[abcdfghijklmnopqrstuvwxyz]\{10\}\>

You’ll note that the character class no longer contains ‘e’, and in the second word we have the actual ‘e’ embedded in the regular expression.

For the third example, of ‘Q’ which is not in the phrase, the command and output are like this:

hangman-regexp.py -x q '_ _ _ / _ _ _ _ _ _ _ / _ _ / _ _ _ / _ _ _ _ _ _ / _ _ _ _ / _ _ _ / _ _ _ _ _ _ _ _ _ _'
\<[abcdefghijklmnoprstuvwxyz]\{3\}\>[[:space:]]\+\<[abcdefghijklmnoprstuvwxyz]\{7\}\>[[:space:]]\+\<[abcdefghijklmnoprstuvwxyz]\{2\}\>[[:space:]]\+\<[abcdefghijklmnoprstuvwxyz]\{3\}\>[[:space:]]\+\<[abcdefghijklmnoprstuvwxyz]\{6\}\>[[:space:]]\+\<[abcdefghijklmnoprstuvwxyz]\{4\}\>[[:space:]]\+\<[abcdefghijklmnoprstuvwxyz]\{3\}\>[[:space:]]\+\<[abcdefghijklmnoprstuvwxyz]\{10\}\>

Running a game: make-hangman-phrase.py

On the other end of this game is the game runner. They think up the phrase and run the rounds. I thought it would be nice to have a tool that, given a phrase and zero or more guesses, output the phrase “hangman style”, along with the “wrong guesses” line. This tool is make-hangman-phrase.py.

Let’s suppose you want to run a game with the phrase “There are four lights”. At the beginning of the game, you would run

make-hangman-phrase.py 'There are four lights'

Here’s the output:

Phrase:		_ _ _ _ _ / _ _ _ / _ _ _ _ / _ _ _ _ _ _
Occurrences:	None
Wrong guess:	None

This provides the phrase and wrong guesses that can be published on Mastodon. The “Occurrences” line here shows how many times a guess shows up in the phrase. We’ll see this below.

Let’s suppose for the first round the letter ‘T’ is guessed. In this case, we run

make-hangman-phrase.py -g T 'There are four lights'

We add -g T to specify that ‘T’ was guessed. The output is this:

Phrase:		T _ _ _ _ / _ _ _ / _ _ _ _ / _ _ _ _ T _
Occurrences:	{'t': 2}
Wrong guess:	None

So now the game runner can say “T was the guess, and there are 2 T’s in the phrase” in their next post.

Let’s say the next poll winner is ‘V’. There’s no ‘V’ in the phrase. Let’s see what happens.

make-hangman-phrase.py -g T -g V 'There are four lights'

Here’s the output:

Phrase:		T _ _ _ _ / _ _ _ / _ _ _ _ / _ _ _ _ T _
Occurrences:	{'t': 2, 'v': 0}
Wrong guesses:	V

Code availability

The code for these tools is at https://github.com/davemq/hangman-tools.