Skip to content

Latest commit

 

History

History
94 lines (49 loc) · 7.3 KB

syntax.md

File metadata and controls

94 lines (49 loc) · 7.3 KB

Grid

New grids are created either at root, or with wfc or map nodes. Example: MazeMap has 2 grids. Grids have one required attribute values and one optional attribute folder.

values="BRGUY" says that black, red, green, blue and yellow values are possible on the grid, and the starting value is black.

folder="DungeonGrowth" says that the interpreter should look for rule files in the DungeonGrowth folder.

For root grids, there is one more optional boolean attribute called origin, equal False by default. If origin is set True, the interpreter creates a pixel at the center of the grid with the value equal to the second value in values. For example, values="YRBAN" origin="True" creates a yellow screen with a red dot in the center.

Symmetry

With most nodes, you can specify the symmetry attribute. Possible symmetry values are listed in SymmetryHelper.cs. By default, the largest symmetry group is used. Example: in Flowers we need flowers growing vertically and not side-to-side, so we specify that rules should only be mirrored along the x axis.

Rules

Rule attributes:

  • in="BBB/BWB" - specifies input part of a rule.
  • out="RR DA FR" - specifies output part of a rule.
  • fin="filename" - loads input from filename.png or filename.vox.
  • fout="filename" - loads output from filename.png or filename.vox.
  • file="filename" - loads a glued input + output box from file. Example: Circuit.
  • p - the probability that the rule will be applied. Equals 1.0 by default. Example: in Apartemazements only 25% of ceiling locations are converted into light sources.

Slashes / are y-separators, spaces are z-separators.

If a file is referenced, the legend should be specified. Legend lists used values in a scanline order. See example in DungeonGrowth.

Rulenodes

There are 3 kinds of rulenodes:

  1. one nodes, also called (exists) nodes.
  2. all nodes, also called {forall} nodes.
  3. prl (parallel) nodes are similar to all nodes, but they are applied independently of each other, they don't care about rule overlaps. Often executing a prl node leads to exactly the same result as with an all node, but prl is more performant.

Rulenode attributes:

  • steps="60" - limits node execution to 60 steps. See an example in River.

Unions

See examples of union use in DungeonGrowth.

Inference

Inference in MarkovJunior allows to impose constraints on the future state, and generate only those runs that lead to the constrained future. Inference is triggered by putting observe elements inside rulenodes (inside one or all nodes, to be precise). Observe elements have 3 attributes: value, from, to.

For example, <observe value="W" to="BR"/> means that squares that are currently white should become black or red after a chain of rule applications. <observe value="I" from="B" to="W"/> means that squares that are currently indigo are turned black immediately, and then should become white after a chain of rule applications.

In SokobanLevel1 we say that the goal I-squares should become white - this would mean that the puzzle is solved. We also help the inference engine by explicitly saying that the current black, white and red squares should not be white in the end. Since we don't have I in the ruleset, we say that current indigo squares should be treated as black by setting from="B".

Rulenodes with inference have a boolean search attribute, false by default. If search is set false, the interpreter follows the rule propagation field greedily. If search is set true, the interpreter searches the state graph using the rule propagation field as a heuristic.

If search is set false, the interpreter can be made to follow the goal more strictly or less strictly by varying the floating point temperature parameter. If search is set true, the attributes are:

  1. Integer limit attribute sets the maximum number of states searched. By default, the number of states is not limited.
  2. Floating point depthCoefficient attribute interpolates between breadth-first search and depth-first search.

See examples of inference use in MultiSokoban9, SokobanLevel1, StairsPath, KnightPatrol, CrossCountry, RegularPath, DiagonalPath, EuclideanPath, BishopParity, SnellLaw, SequentialSokoban, CompleteSAW, CompleteSAWSmart, Island.

Map

See examples of map node use in MazeMap, MarchingSquares, OddScale, OddScale3D, SeaVilla, ModernHouse, CarmaTower.

Path

See examples of path node use in BasicDijkstraFill, BasicDijkstraDungeon, BernoulliPercolation, Percolation, Circuit, DungeonGrowth.

Convolution

See examples of convolution node use in GameOfLife, Cave, ConnectedCaves, ConstrainedCaves, OpenCave, OpenCave3D, Counting, CarmaTower.

WaveFunctionCollapse

See examples of tile WFC in TileDungeon, Knots2D, Knots3D, Surface, EscherSurface, PillarsOfEternity, Apartemazements, SeaVilla, ModernHouse.

See examples of overlap WFC in WaveFlowers, WaveBrickWall, WaveDungeon.

ConvChain

See examples of convchain node use in ChainMaze, ChainDungeon, ChainDungeonMaze.

Questions and Answers

Q: How to make a loop? How to make a sequence repeat?
A: To make a sequence repeat, put a sequence node inside a markov node or inside another sequence node. Examples of this: MultiHeadedWalk, HamiltonianPath, SelectLargeCaves, SelectLongKnots, FireNoise, SmartSAW, FindLongCycle. Counters in markov/sequence nodes are not supported right now. Instead, you may want to repeat the sequence until some node is matched.