jonadsimon
commented
2 years ago Tried this again, didn't help. Functionally-identical programs with/without constraint:
- without constraint took 2.13s
- with constraint took 2.66s
Adding the constrain made the runtime take LONGER
Used an upper-bound of number of each letter occurring across all words, and used a lower-bound of this quantity / 2
Solutions found by the two programs were IDENTICAL, the bounded one just took longer to run
Tried alternate search formulations, but it didn't help
Implies that the program does not have an issue with finding appropriate letters. Make sense since it searchers via the words, and the words directly imply the letters
Since #letter > #squares, can strictly upper-bound the letter counts with a frequency constraint.
Unclear how helpful it will be, but it's at least a well-defined global constraint. If anything breaks it, it will be the overly-common number of e.g. a's