Unwordle

I wrote this to get around the nerd snipe that is wordle (It's awesome and fun, I just have neither the time to do it every day nor the willpower to resist). So here's a corpus based solution that i hope will let me sleep at night.

Also, I've wanted to take a look at the babashka task runner for a while now and this problem seems like a good opportunity.

Requirements

• https://github.com/babashka/babashka
• curl and tar installed, to fetch and unpack the corpus.

Usage

Commands are available as babashka tasks:

• Use corpus to fetch a corpus of words, or go directly to
• solve (will ensure the corpus is available if it's missing). Feedback obtained from wordle can be passed on the command line to receive the next guess (see example below). Provide a strategy (g -> global-freq-strategy, p -> positional-freq-strategy)
• When you're done, you can clean everything up using clean (removes the corpus).

Example

Let's say the wordle is toast.

We have no information so the script picks a word (using the most frequently used letters in the corpus).

> bb run solve g
rates

we type it into wordle and get:

Nice! That's already quite a good start. OK, how do we pass that information back? We provide the feedback in the following form: the letter in our guess word (r, a, t, e, s) plus the feedback we got from wordle: we receive colored tiles, we'll be using the colors' first letter.

tile	meaning	short
	not present	b
	present	y
	correct	g

So our r in first position that is absent from our target word is encoded as rb. For the a in second position that is found in the target word (but at a different position) we write ay. We follow through for each letter and pass this to the task:

> bb run solve g rb ay ty eb sy
stain

Finally! A correct placement!

We can just append this to our previous feedback to receive the next guess:

> bb run solve g rb ay ty eb sy sy ty ag ib nb
coast

so close...

> bb run solve g rb ay ty eb sy sy ty ag ib nb cb og ag sg tg
toast

... and we're done. Enjoy!