https://github.com/shivansh/sudokusat
https://github.com/shivansh/sudokusat
Last synced: 4 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/shivansh/sudokusat
- Owner: shivansh
- License: mit
- Created: 2017-02-03T20:26:02.000Z (over 8 years ago)
- Default Branch: master
- Last Pushed: 2018-02-01T09:48:50.000Z (over 7 years ago)
- Last Synced: 2025-01-07T16:27:20.051Z (5 months ago)
- Language: Python
- Homepage:
- Size: 18.6 KB
- Stars: 0
- Watchers: 3
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# SAT-sudoku
## Initial setup
Following environment variables are to be exported before running z3py python scripts -
```
export LD_LIBRARY_PATH=${z3_location}/build
export PYTHONPATH=${z3_location}/build/python
```
Ideally, the python scripts should be run inside `build/python` under the local z3 installation directory. However, the directory `build/python/z3` can be symlinked to the project root and all the python scripts can then be invoked where they are located.
All the above instructions are encapsulated in [`setup.sh`](setup.sh), and the initial setup can be done via `./setup.sh`.
**NOTE:** Update the location variables used in the script before proceeding to run it.
## Running
[`pattern.py`](pattern.py) contains encoding for sudoku solving patterns (WIP), and running it should generate an incompletely solved sudoku puzzle with a few fixed and rest pencilled entries.
----
### _Everything below this point was written a year back and may no longer be valid._
## Grammar definitions
The grammar used is defined in [grammar.py](grammar.py).
## Results and Benchmarks
* **Sudoku-Generator:**
For generating a valid 9x9 grid (satisfying all the 4 constraints) -
```
python sudoku-generator.py 25.28s user 0.02s system 99% cpu 25.333 total
```
For multiple iterations, same variation of sudoku was generated.
For generating a valid sudoku using the generated 9x9 grid -
Program was taking too long to halt, possibly because the **random selection** of indices `(i,j)` to be removed from the generated grid contributes greatly to the non-determinism, since as the grid gets close to being sparsely populated (many empty cells), the probability of a randomly generated non-empty cell keeps decreasing
## Todo
* Reduce the grid into a 3-D binary form which can be fed directly to the SAT solver.
* Find a better way of selecting cells to be removed from a valid 9x9 grid for generating a sudoku. A random selection (not optimal) is being used currently.