netwalk

Solver for netwalk, reading in image files of game state,

Status - TBC

• Game currently unsolvable at the expert level
• Having exhausted all guidance in the tutorial, I'm open to adding new solver logic if I get suggestions or bright ideas
• Having tried to solve the partly solved game state (see data/lgo_netwalk_example_game_expert_part_solved.png) the only route I can see is to see whether taking arbitrary decisions sets off a 'chain reaction' of tile solving (as is ensured to happen at easy and intermediate difficulty levels).
• Having attempted this manually, I've noted that when setting fix_connection(a), it's crucial to supply as many possible values to the a direction vector as there are connections possible on a tile (e.g. freezing the 0,0 [corner wire] tile in the above example using a=[0,3] will initiate a chain reaction whereas simply a=[0] or a=[3] will lead to errors, as it effectively puts the game 'out of sync').
• It's not immediately clear however whether this approach will lead to full solutions - however I suspect trying it with the tiles with the maximum number of connections (i.e. the T-wires) bordering an already solved region will be most likely to produce a full solution state...
• This however is dependent on already having a solved region from the initial solve step, which appears to be the minority of generated game states at expert level!

Development progress

• Solver now works for the easier game format, where there are blank tiles on the grid periphery
• Solver now works for the medium game format (for the one game I tried, uncertain if it will solve all)
• Solver for the expert level is under development...
• Solver can now display the results by rotating the tiles of the input image

Example game states

Beginner

Initial state 1:

Solved:

Initial state 2:

Solved:

Medium

Initial state:

Solved:

Expert

Initial state 1:

Part solved:

Initial state 2:

Not solved:

About

• terminology for components etc. as in De Biasi 2012*

* A couple of exceptions:

• N.B. the power unit shown in the figure here [from a paper] has two outputs, whereas the one coded for in this library has from one to three, in line with the game at this link

Usage

Interactive

• python -im netwalk
• from .imaging import give_me_the_tiles (or show_me_the_solved_tiles to open a viewer)

Command line

• python -c "from netwalk import imaging; imaging.show_me_the_solved_tiles()"
• (There's no full [expert level] solving functionality yet but will add a proper CLI interface when there is)

TODO

• write tests for all components
  • give all wire components on states
    • check this hasn't broken any tests
• use relative data path rather than hardcoded
• read in image file and parse tiling/component state
• detect output directions of component wiring and add to component state
• write tests for imaging and tiling
• add to the tile initialisation:
  • a 'solved' property per tile
  • a 'fixed' vector per component output direction
  • an 'avoid' vector per tile [maximum length = 4 - number of output directions]
• add methods that modify these attributes per tile over the tileset
• implement an interface list with methods to access both tiles across an interface
• implement tile level check_solvable, solve, check_solved, set_fixed, and set_avoid methods
• add a series of solvers for a tile which can be run in succession as applicable
  • just implemented a single class in solve_tile in the end with logical checks
• add further solvers until the expert level can be solved
• calculate and display tile rotations
  • apply the rotations to the input image and display a solved form on screen
• add a command to automatically save a game solution to the data directory based on its input filename