# qlearningAgents.py
# ------------------
# Licensing Information:  You are free to use or extend these projects for
# educational purposes provided that (1) you do not distribute or publish
# solutions, (2) you retain this notice, and (3) you provide clear
# attribution to UC Berkeley, including a link to http://ai.berkeley.edu.
# 
# Attribution Information: The Pacman AI projects were developed at UC Berkeley.
# The core projects and autograders were primarily created by John DeNero
# (denero@cs.berkeley.edu) and Dan Klein (klein@cs.berkeley.edu).
# Student side autograding was added by Brad Miller, Nick Hay, and
# Pieter Abbeel (pabbeel@cs.berkeley.edu).


from game import *
from learningAgents import ReinforcementAgent
from featureExtractors import *

import random,util,math

class QLearningAgent(ReinforcementAgent):
    """
      Q-Learning Agent

      Functions you should fill in:
        - computeValueFromQValues
        - computeActionFromQValues
        - getQValue
        - getAction
        - update

      Instance variables you have access to
        - self.epsilon (exploration prob)
        - self.alpha (learning rate)
        - self.discount (discount rate)

      Functions you should use
        - self.getLegalActions(state)
          which returns legal actions for a state
    """
    def __init__(self, **args):
        "You can initialize Q-values here..."
        ReinforcementAgent.__init__(self, **args)

        "*** YOUR CODE HERE ***"
        # Initialize all Q_Values to 0
        self.qValues = util.Counter()

    def getQValue(self, state, action):
        """
          Returns Q(state,action)
          Should return 0.0 if we have never seen a state
          or the Q node value otherwise
        """
        "*** YOUR CODE HERE ***"
        # Get the current Q-Value
        return self.qValues[(state, action)]


    def computeValueFromQValues(self, state):
        """
          Returns max_action Q(state,action)
          where the max is over legal actions.  Note that if
          there are no legal actions, which is the case at the
          terminal state, you should return a value of 0.0.
        """
        "*** YOUR CODE HERE ***"
        qValues = []

        # get a list of the Q-Values from the legal actions
        for action in self.getLegalActions(state):
          qValues.append(self.getQValue(state, action))

        # return the max Q-Value or 0 if there are no qValues
        return 0.0 if len(qValues) == 0 else max(qValues)

    def computeActionFromQValues(self, state):
        """
          Compute the best action to take in a state.  Note that if there
          are no legal actions, which is the case at the terminal state,
          you should return None.
        """
        "*** YOUR CODE HERE ***"
        # Computes the best action of a state using Q-Values

        # initialize values
        maxAction = None
        maxQValue = None

        # for every action
        for action in self.getLegalActions(state):
          # get the Q-Value
          qValue = self.getQValue(state, action)
          # If it is greater than the current max, set it as the new max
          if qValue > maxQValue:
            maxAction = action
            maxQValue = qValue

        # return the max
        return maxAction


    def getAction(self, state):
        """
          Compute the action to take in the current state.  With
          probability self.epsilon, we should take a random action and
          take the best policy action otherwise.  Note that if there are
          no legal actions, which is the case at the terminal state, you
          should choose None as the action.

          HINT: You might want to use util.flipCoin(prob)
          HINT: To pick randomly from a list, use random.choice(list)
        """
        # Pick Action
        legalActions = self.getLegalActions(state)
        action = None

        "*** YOUR CODE HERE ***"
        # Use Epsilon to determine whether to take a random action or the policy action
        if util.flipCoin(self.epsilon):
          # A random action was selected
          action = random.choice(legalActions)
        else:
          # The policy action was taken
          action = self.computeActionFromQValues(state)

        return action

    def update(self, state, action, nextState, reward):
        """
          The parent class calls this to observe a
          state = action => nextState and reward transition.
          You should do your Q-Value update here

          NOTE: You should never call this function,
          it will be called on your behalf
        """
        "*** YOUR CODE HERE ***"
        # This function updates the Q-Values using Q-Value Iteration

        # initialize the sample with the base reward
        sample = reward
        
        # if the state is not terminal
        if len(self.getLegalActions(nextState)) > 0:
          # initialize max action to None (converted to 0)
          maxAction = None
          
          # iterate through each legal action and find the max value action
          for nextAction in self.getLegalActions(nextState):
            maxAction = max(maxAction, self.getQValue(nextState, nextAction))
            
          # This is the sample portion of the Q-Value Iteration function
          sample += self.discount * maxAction

        # This is where the Q-Values are updated using the Q-Value Iteration function
        self.qValues[(state, action)] = ((1 - self.alpha) * self.getQValue(state,action)) + (self.alpha * sample)

    def getPolicy(self, state):
        return self.computeActionFromQValues(state)

    def getValue(self, state):
        return self.computeValueFromQValues(state)


class PacmanQAgent(QLearningAgent):
    "Exactly the same as QLearningAgent, but with different default parameters"

    def __init__(self, epsilon=0.05,gamma=0.8,alpha=0.2, numTraining=0, **args):
        """
        These default parameters can be changed from the pacman.py command line.
        For example, to change the exploration rate, try:
            python pacman.py -p PacmanQLearningAgent -a epsilon=0.1

        alpha    - learning rate
        epsilon  - exploration rate
        gamma    - discount factor
        numTraining - number of training episodes, i.e. no learning after these many episodes
        """
        args['epsilon'] = epsilon
        args['gamma'] = gamma
        args['alpha'] = alpha
        args['numTraining'] = numTraining
        self.index = 0  # This is always Pacman
        QLearningAgent.__init__(self, **args)

    def getAction(self, state):
        """
        Simply calls the getAction method of QLearningAgent and then
        informs parent of action for Pacman.  Do not change or remove this
        method.
        """
        action = QLearningAgent.getAction(self,state)
        self.doAction(state,action)
        return action


class ApproximateQAgent(PacmanQAgent):
    """
       ApproximateQLearningAgent

       You should only have to overwrite getQValue
       and update.  All other QLearningAgent functions
       should work as is.
    """
    def __init__(self, extractor='IdentityExtractor', **args):
        self.featExtractor = util.lookup(extractor, globals())()
        PacmanQAgent.__init__(self, **args)
        self.weights = util.Counter()

    def getWeights(self):
        return self.weights

    def getQValue(self, state, action):
        """
          Should return Q(state,action) = w * featureVector
          where * is the dotProduct operator
        """
        "*** YOUR CODE HERE ***"
        # In Approximate Q-Learning, Q Values are computed by linearly summing weight and feature pairs
        
        # get the features
        feats = self.featExtractor.getFeatures(state, action)
        # get the weights
        weights = self.getWeights()
        
        # return the dot product of the weights and the featureVector
        return weights * feats

    def update(self, state, action, nextState, reward):
        """
           Should update your weights based on transition
        """
        "*** YOUR CODE HERE ***"
        # This function is where the weights are updated using the Approximate Q-Learning linear-function
        
        # get the features
        feats = self.featExtractor.getFeatures(state, action)
        # this is the difference from the Approximate Q-Learning linear-function
        difference = reward + self.discount * self.getValue(nextState) - self.getQValue(state, action)

        # updating of weights occurs here
        for feat in feats:
          # this is the Approximate Q-Learning linear-function
          self.weights[feat] += self.alpha * difference * feats[feat]


    def final(self, state):
        "Called at the end of each game."
        # call the super-class final method
        PacmanQAgent.final(self, state)

        # did we finish training?
        if self.episodesSoFar == self.numTraining:
            # you might want to print your weights here for debugging
            "*** YOUR CODE HERE ***"
            # print(self.weights)
            # print("Num weights: " + str(len(self.weights)))
            pass