From 7946483f096206f85bf648e10443b77bc7999b43 Mon Sep 17 00:00:00 2001
From: dehnert <dehnert@cs.rwth-aachen.de>
Date: Tue, 1 Sep 2015 21:53:27 +0200
Subject: [PATCH] added function to modify state-action rewards in
 nondeterministic model

Former-commit-id: d3a4c3870722278f84160137976f3459381078af
---
 src/models/sparse/NondeterministicModel.cpp | 18 ++++++++++++++++++
 src/models/sparse/NondeterministicModel.h   |  9 +++++++++
 src/models/sparse/StandardRewardModel.cpp   |  5 +++++
 src/models/sparse/StandardRewardModel.h     | 13 ++++++++++++-
 4 files changed, 44 insertions(+), 1 deletion(-)

diff --git a/src/models/sparse/NondeterministicModel.cpp b/src/models/sparse/NondeterministicModel.cpp
index 340400dde..c5e3e9ecf 100644
--- a/src/models/sparse/NondeterministicModel.cpp
+++ b/src/models/sparse/NondeterministicModel.cpp
@@ -4,6 +4,8 @@
 
 #include "src/adapters/CarlAdapter.h"
 
+#include "src/exceptions/InvalidOperationException.h"
+
 namespace storm {
     namespace models {
         namespace sparse {
@@ -45,6 +47,22 @@ namespace storm {
                 return indices[state+1] - indices[state];
             }
             
+            template<typename ValueType, typename RewardModelType>
+            void NondeterministicModel<ValueType, RewardModelType>::modifyStateActionRewards(RewardModelType& rewardModel, std::map<std::pair<uint_fast64_t, LabelSet>, typename RewardModelType::ValueType> const& modifications) const {
+                STORM_LOG_THROW(rewardModel.hasStateActionRewards(), storm::exceptions::InvalidOperationException, "Cannot modify state-action rewards, because the reward model does not have state-action rewards.");
+                STORM_LOG_THROW(this->hasChoiceLabeling(), storm::exceptions::InvalidOperationException, "Cannot modify state-action rewards, because the model does not have an action labeling.");
+                std::vector<LabelSet> const& choiceLabels = this->getChoiceLabeling();
+                for (auto const& modification : modifications) {
+                    uint_fast64_t stateIndex = modification.first.first;
+                    for (uint_fast64_t row = this->getNondeterministicChoiceIndices()[stateIndex]; row < this->getNondeterministicChoiceIndices()[stateIndex + 1]; ++row) {
+                        // If the action label of the row matches the requested one, we set the reward value accordingly.
+                        if (choiceLabels[row] == modification.first.second) {
+                            rewardModel.setStateActionRewardValue(row, modification.second);
+                        }
+                    }
+                }
+            }
+            
             template<typename ValueType, typename RewardModelType>
             void NondeterministicModel<ValueType, RewardModelType>::reduceToStateBasedRewards() {
                 for (auto& rewardModel : this->getRewardModels()) {
diff --git a/src/models/sparse/NondeterministicModel.h b/src/models/sparse/NondeterministicModel.h
index 9c5a01b05..5bacf6dd7 100644
--- a/src/models/sparse/NondeterministicModel.h
+++ b/src/models/sparse/NondeterministicModel.h
@@ -73,6 +73,15 @@ namespace storm {
                  */
                 uint_fast64_t getNumberOfChoices(uint_fast64_t state) const;
                 
+                /*!
+                 * Modifies the state-action reward vector of the given reward model by setting the value specified in
+                 * the map for the corresponding state-action pairs.
+                 *
+                 * @param rewardModel The reward model whose state-action rewards to modify.
+                 * @param modifications A mapping from state-action pairs to the their new reward values.
+                 */
+                void modifyStateActionRewards(RewardModelType& rewardModel, std::map<std::pair<uint_fast64_t, LabelSet>, typename RewardModelType::ValueType> const& modifications) const;
+                
                 virtual void reduceToStateBasedRewards() override;
                 
                 virtual void printModelInformationToStream(std::ostream& out) const override;
diff --git a/src/models/sparse/StandardRewardModel.cpp b/src/models/sparse/StandardRewardModel.cpp
index b8fa2e3be..5788dae10 100644
--- a/src/models/sparse/StandardRewardModel.cpp
+++ b/src/models/sparse/StandardRewardModel.cpp
@@ -197,6 +197,11 @@ namespace storm {
                 return result;
             }
             
+            template<typename ValueType>
+            void StandardRewardModel<ValueType>::setStateActionRewardValue(uint_fast64_t row, ValueType const& value) {
+                this->optionalStateActionRewardVector.get()[row] = value;
+            }
+            
             template<typename ValueType>
             bool StandardRewardModel<ValueType>::empty() const {
                 return !(static_cast<bool>(this->optionalStateRewardVector) || static_cast<bool>(this->optionalStateActionRewardVector) || static_cast<bool>(this->optionalTransitionRewardMatrix));
diff --git a/src/models/sparse/StandardRewardModel.h b/src/models/sparse/StandardRewardModel.h
index 457d4df5b..eb396d90a 100644
--- a/src/models/sparse/StandardRewardModel.h
+++ b/src/models/sparse/StandardRewardModel.h
@@ -10,9 +10,11 @@
 namespace storm {
     namespace models {
         namespace sparse {
-            template <typename ValueType>
+            template <typename CValueType>
             class StandardRewardModel {
             public:
+                typedef CValueType ValueType;
+                
                 /*!
                  * Constructs a reward model by copying the given data.
                  *
@@ -213,6 +215,15 @@ namespace storm {
                  */
                 std::vector<ValueType> getTotalStateActionRewardVector(uint_fast64_t numberOfRows, std::vector<uint_fast64_t> const& rowGroupIndices, storm::storage::BitVector const& filter) const;
                 
+                /*!
+                 * Sets the given value in the state-action reward vector at the given row. This assumes that the reward
+                 * model has state-action rewards.
+                 *
+                 * @param row The row at which to set the given value.
+                 * @param value The value to set.
+                 */
+                void setStateActionRewardValue(uint_fast64_t row, ValueType const& value);
+                
                 /*!
                  * Retrieves whether the reward model is empty, i.e. contains no state-, state-action- or transition-based
                  * rewards.