WIP Step 3

Ajout d'un jeu de test minimaliste LearningTest\TestKhiops\Standard\BUGEncryptRules

Modification temporaire de run-standard-tests.yaml pour n'executer les test que sous linux en release

KWDataGridClusteringCosts::ComputeAttributeCost: ajout de trace pour le calcul des couts des attributs

KWDREncrypt::EncryptString: trace pour verifier le comportement de isprint

.github/workflows/run-standard-tests.yml

@@ -29,10 +29,8 @@ jobs:
     strategy:
       matrix:
         build-setup:
-          - {os: windows-2022, cmake-preset: windows-msvc}
           - {os: ubuntu-latest, cmake-preset: linux-gcc}
-          - {os: macos-latest, cmake-preset: macos-clang}
-        config: [debug, release]
+        config: [release]
     runs-on: ${{ matrix.build-setup.os }}
     env:
       PRESET_NAME: ${{ matrix.build-setup.cmake-preset }}-${{ matrix.config }}

src/Learning/KWDRRuleLibrary/KWDRStringEncrypt.cpp

@@ -50,6 +50,20 @@ const ALString KWDREncrypt::EncryptString(ALString& sStringToEncrypt, const Symb
 	int nBlockCode;
 	char cLastChar;
 
+	// DDD
+	static boolean bTraceOnce = false;
+	if (not bTraceOnce)
+	{
+		cout << "Index\tChar\tisdigit\tisalnum\tisprint\n";
+		for (i = 0; i < 256; i++)
+		{
+			cout << i << "\t";
+			if (isprint(i))
+				cout << (char)i;
+			cout << "\t" << isdigit(i) << "\t" << isalnum(i) << "\t" << isprint(i) << endl;
+		}
+	}
+
 	// Memorisation de la graine aleatoire initiale
 	nRandomSeed = GetRandomSeed();
 

src/Learning/KWDataPreparation/KWAttributeSubsetStats.cpp

@@ -226,6 +226,12 @@ boolean KWAttributeSubsetStats::ComputeStats(const KWTupleTable* tupleTable)
 	    optimizedDataGrid->GetAttributeAt(0)->GetAttributeName() == "SPetalLength" and
 	    optimizedDataGrid->GetAttributeAt(1)->GetAttributeName() == "SepalWidth")
 	{
+		SetGlobalTraceOn(true);
+		dataGridCosts->ComputeAttributeCost(optimizedDataGrid->GetAttributeAt(0),
+						    optimizedDataGrid->GetAttributeAt(0)->GetPartNumber());
+		dataGridCosts->ComputeAttributeCost(optimizedDataGrid->GetAttributeAt(1),
+						    optimizedDataGrid->GetAttributeAt(1)->GetPartNumber());
+		SetGlobalTraceOn(false);
 		cout << "All costs" << endl;
 		dataGridCosts->WriteDataGridAllCosts(optimizedDataGrid, cout);
 	}

src/Learning/KWDataPreparation/KWDataGridCosts.cpp

@@ -4,6 +4,19 @@
 
 #include "KWDataGridCosts.h"
 
+// DDD
+static boolean bGlobalTraceOn = false;
+
+boolean GetGlobalTraceOn()
+{
+	return bGlobalTraceOn;
+}
+
+void SetGlobalTraceOn(boolean bValue)
+{
+	bGlobalTraceOn = bValue;
+}
+
 ////////////////////////////////////////////////////////////////////////////////////////
 // Classe KWDataGridCosts
 
@@ -1398,6 +1411,15 @@ double KWDataGridClusteringCosts::ComputeAttributeCost(const KWDGAttribute* attr
 	// Une partition avec poubelle contient au moins 2 parties informatives + 1 groupe poubelle
 	require(attribute->GetGarbageModalityNumber() == 0 or nPartitionSize >= 3);
 
+	// DDD
+	if (GetGlobalTraceOn())
+	{
+		cout << "KWDataGridClusteringCosts::ComputeAttributeCost\t" << attribute->GetAttributeName() << endl;
+		cout << "\tnPartileNumber\t" << nPartileNumber << endl;
+		cout << "\tKWFrequencyTable::GetMinimumNumberOfModalitiesForGarbage()\t"
+		     << KWFrequencyTable::GetMinimumNumberOfModalitiesForGarbage() << endl;
+	}
+
 	// Cout nul si partition nulle
 	if (nPartitionSize == 1)
 		dAttributeCost = 0;
@@ -1425,38 +1447,90 @@ double KWDataGridClusteringCosts::ComputeAttributeCost(const KWDGAttribute* attr
 			if (nPartileNumber > KWFrequencyTable::GetMinimumNumberOfModalitiesForGarbage())
 				dAttributeCost += log(2.0);
 
+			if (GetGlobalTraceOn()) // DDD
+			{
+				cout << "\tattribute->GetCost()\t" << attribute->GetCost() << endl;
+				cout << "\tnGarbageModalityNumber\t" << nGarbageModalityNumber << endl;
+				cout << "\tdAttributeCost1\t" << dAttributeCost << endl;
+			}
+
 			// Cout de codage du choix des modalites informatives (hors poubelle)
 			if (nGarbageModalityNumber > 0)
 			{
 				// Cout de codage de la taille de la non-poubelle
 				dAttributeCost += KWStat::BoundedNaturalNumbersUniversalCodeLength(
 				    nPartileNumber - nGarbageModalityNumber - 1, nPartileNumber - 2);
+				if (GetGlobalTraceOn()) // DDD
+				{
+					cout << "\tdAttributeCost2\t" << dAttributeCost << "\t"
+					     << KWStat::BoundedNaturalNumbersUniversalCodeLength(
+						    nPartileNumber - nGarbageModalityNumber - 1, nPartileNumber - 2)
+					     << endl;
+				}
 
 				// Choix des modalites hors poubelle selon un tirage multinomial avec un tirage par
 				// variable
 				dAttributeCost +=
 				    (nPartileNumber - nGarbageModalityNumber) * log(nPartileNumber * 1.0) -
 				    KWStat::LnFactorial(nPartileNumber - nGarbageModalityNumber);
+				if (GetGlobalTraceOn()) // DDD
+				{
+					cout << "\tdAttributeCost3\t" << dAttributeCost << "\t"
+					     << (nPartileNumber - nGarbageModalityNumber) * log(nPartileNumber * 1.0) -
+						    KWStat::LnFactorial(nPartileNumber - nGarbageModalityNumber)
+					     << endl;
+				}
 
 				// Cout de codage du nombre de parties informatives (nPartitionSize-1)
 				dAttributeCost += KWStat::BoundedNaturalNumbersUniversalCodeLength(
 				    nPartitionSize - 2, nPartileNumber - nGarbageModalityNumber - 1);
+				if (GetGlobalTraceOn()) // DDD
+				{
+					cout << "\tdAttributeCost4\t" << dAttributeCost << "\t"
+					     << KWStat::BoundedNaturalNumbersUniversalCodeLength(
+						    nPartitionSize - 2, nPartileNumber - nGarbageModalityNumber - 1)
+					     << endl;
+				}
 
 				// Cout de codage du choix des parties informatives (nPartitionSize - 1)
 				dAttributeCost +=
 				    KWStat::LnBell(nPartileNumber - nGarbageModalityNumber, nPartitionSize - 1);
+				if (GetGlobalTraceOn()) // DDD
+				{
+					cout << "\tdAttributeCost5\t" << dAttributeCost << "\t"
+					     << KWStat::LnBell(nPartileNumber - nGarbageModalityNumber,
+							       nPartitionSize - 1)
+					     << endl;
+				}
 			}
 			else
 			{
 				// Cout de codage du nombre de parties
 				dAttributeCost += KWStat::BoundedNaturalNumbersUniversalCodeLength(nPartitionSize - 1,
 												   nPartileNumber - 1);
+				if (GetGlobalTraceOn()) // DDD
+				{
+					cout << "\tdAttributeCostBis2\t" << dAttributeCost << "\t"
+					     << KWStat::BoundedNaturalNumbersUniversalCodeLength(nPartitionSize - 1,
+												 nPartileNumber - 1)
+					     << endl;
+				}
 
 				// Cout de codage du choix des parties
 				dAttributeCost += KWStat::LnBell(nPartileNumber, nPartitionSize);
+				if (GetGlobalTraceOn()) // DDD
+				{
+					cout << "\tdAttributeCostBis3\t" << dAttributeCost << "\t"
+					     << KWStat::LnBell(nPartileNumber, nPartitionSize) << endl;
+				}
 			}
 		}
 	}
+	if (GetGlobalTraceOn()) // DDD
+	{
+		cout << "\tdAttributeCostFinal\t" << dAttributeCost << endl;
+	}
+
 	return dAttributeCost;
 }
 

src/Learning/KWDataPreparation/KWDataGridCosts.h

@@ -14,6 +14,12 @@ class KWDataGridGeneralizedClassificationCosts;
 #include "KWDataGridMerger.h"
 #include "KWStat.h"
 
+// DDD
+//  Indicateur de trace globale
+//  A positionner et utiliser temporairement le temps d'un debugage
+boolean GetGlobalTraceOn();
+void SetGlobalTraceOn(boolean bValue);
+
 ////////////////////////////////////////////////////////////////////////////
 // Definition de la structure des couts d'une grille de donnees
 // Les couts par entite, nuls par defaut, sont redefinissable dans des sous-classes

src/Learning/MODL/MODL.cpp

@@ -36,7 +36,8 @@ int main(int argc, char** argv)
 	// Choix du repertoire de lancement pour le debugage sous Windows (a commenter apres fin du debug)
 	// SetWindowsDebugDir("Standard", "IrisLight");
 	// SetWindowsDebugDir("Standard", "Iris2D");
-	SetWindowsDebugDir("z_Work", "BUGIrisU");
+	// SetWindowsDebugDir("z_Work", "BUGIrisU");
+	SetWindowsDebugDir("z_Work", "BUGEncryptRules");
 
 	// Parametrage des logs memoires depuis les variables d'environnement, pris en compte dans KWLearningProject
 	//   KhiopsMemStatsLogFileName, KhiopsMemStatsLogFrequency, KhiopsMemStatsLogToCollect

test/LearningTest/TestKhiops/Standard/BUGEncryptRules/EncryptRules.kdic

@@ -0,0 +1,8 @@
+#Khiops 7.7.2i
+
+Dictionary	EncryptRules
+{
+	Categorical	Message		;	
+	Categorical	Key		;	
+	Categorical	AEncrypt	 = Encrypt(Message, Key)	;	
+};

test/LearningTest/TestKhiops/Standard/BUGEncryptRules/EncryptRules.txt

@@ -0,0 +1,3 @@
+Message	Key
+_	
++2.99e-1	

test/LearningTest/TestKhiops/Standard/BUGEncryptRules/results.ref/T_EncryptRules.txt

@@ -0,0 +1,3 @@
+Message	Key	AEncrypt
+_		oHKW
++2.99e-1		mHnS4d6s1RmbB

test/LearningTest/TestKhiops/Standard/BUGEncryptRules/results.ref/err.txt

@@ -0,0 +1,5 @@
+Deploy model EncryptRules
+Input database ./EncryptRules.txt: Read records: 2
+Output database ./results/T_EncryptRules.txt: Written records: 2
+Model deployment time: 0:00:00.01
+