Post #1268786

2026-03-15 02:36 UTC

@reconbot @castaway The clue DSL is a tiny subset of Python, and I use Python's `eval()`. Constants: cards: a1, b1, ... d5 rows: row1..row5 columns: colA..colD neighbors: na1, nb1, ... nd5 above, below, left, right: ab4, bc1, ld3, rd1, etc. all edge cards: edges unsigned integers Functions: cc: criminal count ic: innocent count even odd Operators: + - & | = > and or #CluesBySam

Replies (2)

@kbob@chaos.social 2026-03-15 14:21

@reconbot @castaway Solving a puzzle takes about 7 seconds on my computer. This morning I worked out most of a scheme to do the brute force part of the Clues by Sam solver in numpy, which should speed it up 100X. But I'm not doing it (for now). Translating and typing in the clues takes several minutes. So CPU time is unimportant. I'm kind of thinking about putting a natural language front end onto it so I can type the clues as written. 1/2 #CluesBySam

Open ##1268787
@kbob@chaos.social 2026-03-17 23:02

@reconbot @castaway My solver language keeps growing. Yesterday I added 'cco' and 'ico', criminals/innocents are connected. "All criminals in column B are connected" => cco(colB). Today I added some horrible macros for the "Only one row/column has ..." type clues. They're ugly.

Open ##1268789