Merge pull request #4 from pik-gane/viz_compat

Compatibility with web book and viz

.github/workflows/test.yml

@@ -0,0 +1,14 @@
+---
+name: "Test"
+on:
+  pull_request:
+  push:
+jobs:
+  tests:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v3
+      - uses: cachix/install-nix-action@v25
+        with:
+          nix_path: nixpkgs=channel:nixos-unstable
+      - run: nix-shell --run "npm i && npm run test"

README.md

@@ -4,51 +4,21 @@ Extension of webppl-agents with [additional non-maximizing agent types](/src/age
 
 To test:
 ```
-webppl --require webppl-dp --require . examples/runVerySimpleGW.wppl -- --gw GW3 --verbose
+./run.sh examples/runVerySimpleGW.wppl -- --gw GW3 --verbose
 ```
 
-(original README.md:)
-
-# webppl-agents
-
-This package provides constructors for MDP and POMDP agents, for gridworld and bandit environments, and a function for visualizing gridworlds:
-
-- Environments:
-  - [`makeGridWorldMDP`](https://github.com/agentmodels/webppl-agents/blob/master/src/environments/makeGridWorldMDP.wppl)
-  - [`makeGridWorldPOMDP`](https://github.com/agentmodels/webppl-agents/blob/master/src/environments/makeGridWorldPOMDP.wppl)
-  - [`makeBanditPOMDP`](https://github.com/agentmodels/webppl-agents/blob/master/src/environments/makeBanditPOMDP.wppl)
-- Agents:
-  - [`makeMDPAgent`](https://github.com/agentmodels/webppl-agents/blob/master/src/agents/makeMDPAgent.wppl) + [`simulateMDP`](https://github.com/agentmodels/webppl-agents/blob/master/src/simulation/simulateMDP.wppl)
-  - [`makePOMDPAgent`](https://github.com/agentmodels/webppl-agents/blob/master/src/agents/makePOMDPAgent.wppl) + [`simulatePOMDP`](https://github.com/agentmodels/webppl-agents/blob/master/src/simulation/simulatePOMDP.wppl)
-- Visualization:
-  - [`GridWorld.draw`](https://github.com/agentmodels/webppl-agents/blob/master/src/visualization/gridworld.js) (also available as `viz.gridworld` if [webppl-viz](https://github.com/probmods/webppl-viz) is used)
-
-## Installation
-
-To globally install `webppl-agents`, run:
-
-    mkdir -p ~/.webppl
-    npm install --prefix ~/.webppl webppl-agents
-
-This may print warnings (`npm WARN ENOENT`...) which can be ignored.
-
-To upgrade to the latest version, run:
-
-    npm install --prefix ~/.webppl webppl-agents --force
-
-For the agent functions, you will also need to install [webppl-dp](https://github.com/stuhlmueller/webppl-dp).
 
 ## Usage
 
 Once installed, you can make the environment and agent functions available to `program.wppl` by running:
 
-    webppl --require webppl-dp --require webppl-agents program.wppl
+    ./run.sh program.wppl
 
 ## Testing
 
 Run the included test using:
 
-    webppl --require webppl-dp --require . tests/tests.wppl
+    ./test.sh
 
 ## License
 

default.nix

@@ -0,0 +1,18 @@
+{ pkgs ? import <nixpkgs> { } }:
+with pkgs;
+let
+  node = nodejs-18_x;
+in
+mkShell {
+  name = "env";
+  buildInputs = [
+    nodePackages.typescript-language-server
+    python3
+    # nodePackages.prettier
+    # nodePackages.eslint
+    node
+  ];
+  LD_LIBRARY_PATH = lib.makeLibraryPath (with pkgs; [
+    libuuid
+  ]);
+}

examples/runVerySimpleGW.wppl

@@ -1,5 +1,9 @@
 // run via  webppl --require webppl-dp --require . examples/runVerySimpleGW.wppl
-
+
+var _W = webpplAgents;
+var _SU = webpplAgents.satisfia;
+var _PP = webpplAgents.pretty;
+
 var env = getEnv(),
     argv = env.argv,
     params = extend({
@@ -45,6 +49,7 @@ var simulate = function(state, aleph, _t) {
     if (options.verbose || options.debug) console.log(pad(state),"SIMULATE, t",t,"state",prettyState(state),"aleph4state",aleph4state,": localPolicy",JSON.stringify(localPolicy.params),"\n"+pad(state),"| action",action,"aleph4action",aleph4action,"Edel",Edel,"(terminal)");
     return { 
       trajectory: [stepData], // sequence of [state, action] pairs
+      tr4viz: [state],
       conditionalExpectedIndicator: Edel // expected indicator conditional on this trajectory
     };
   } else {
@@ -54,17 +59,18 @@ var simulate = function(state, aleph, _t) {
     var nextOut = simulate(nextState, nextAleph4state, t+1);
     return { 
       trajectory: [stepData].concat(nextOut.trajectory), 
+      tr4viz: [state].concat(nextOut.tr4viz),
       conditionalExpectedIndicator: Edel + nextOut.conditionalExpectedIndicator
     };
   }
 };
 
 console.log("aleph0", asInterval(aleph0));
 
-var t0 =  webpplAgents.time();
+var t0 =  _SU.time();
 // verify meeting of expectations:
 console.log("V", V(startState, aleph0));
-console.log("TIME:", webpplAgents.time() - t0, "ms");
+console.log("TIME:", _SU.time() - t0, "ms");
 console.log("cupLoss", cupLoss(mdp.startState, aleph0));
 console.log("entropy", entropy(mdp.startState, aleph0));
 console.log("KLdiv", KLdiv(mdp.startState, aleph0));
@@ -74,20 +80,22 @@ var gd = agent.getData, agentData = gd();
 
 // estimate distribution of trajectories:
 
+var sym = simulate(mdp.startState, aleph0);
+
 var trajDist = Infer({ model() {
-  return simulate(mdp.startState, aleph0).trajectory;
+  return sym.trajectory;
 }}).getDist();
 
 console.log("\nDATA FOR REGRESSION TESTS: \ntrajDist");
-var regressionTestData = webpplAgents.trajDist2simpleJSON(trajDist);
+var regressionTestData = _PP.trajDist2simpleJSON(trajDist);
 console.log(JSON.stringify(regressionTestData));
 console.log("END OF DATA FOR REGRESSION TESTS\n");
 
 var trajData = trajDist2TrajData(trajDist, agent);
 
 //console.log("trajData", trajData);
 
-var locActionData = webpplAgents.trajDist2LocActionData(trajDist, trajData);
+var locActionData = _SU.trajDist2LocActionData(trajDist, trajData);
 console.log("locActionData", locActionData);
 
 console.log("\nminAdmissibleQ:");
@@ -96,16 +104,22 @@ console.log("\nmaxAdmissibleQ:");
 console.log(stateActionFct2ASCII(agent.maxAdmissibleQ, agentData.stateActionPairs));
 
 console.log("\nQ:");
-console.log(webpplAgents.locActionData2ASCII(locActionData.Q));
+console.log(_PP.locActionData2ASCII(locActionData.Q));
 console.log("\ncupLoss:");
-console.log(webpplAgents.locActionData2ASCII(locActionData.cupLoss));
+console.log(_PP.locActionData2ASCII(locActionData.cupLoss));
 console.log("\nmessingPotential:");
-console.log(webpplAgents.locActionData2ASCII(locActionData.messingPotential));
+console.log(_PP.locActionData2ASCII(locActionData.messingPotential));
 console.log("\ncombinedLoss:");
-console.log(webpplAgents.locActionData2ASCII(locActionData.combinedLoss));
+console.log(_PP.locActionData2ASCII(locActionData.combinedLoss));
 
 console.log("\naction frequencies:");
-console.log(webpplAgents.locActionData2ASCII(locActionData.actionFrequency));
+console.log(_PP.locActionData2ASCII(locActionData.actionFrequency));
+
 
+var tr=sym.tr4viz;
+console.log(tr);
+console.log(tr[0].loc);
+console.log(tr[1].loc);
 
+_W.draw(world, {trajectory:tr});