21949: Fixes issue when emitting and parsing strings that start looki…

…ng numeric but are not (e.g., 1.0.0), and fixes issue with conditioned code features (#294)

src/Amalgam/Parser.cpp

@@ -417,10 +417,22 @@ std::string Parser::GetNextIdentifier(bool allow_leading_label_marks)
 	}
 }
 
-EvaluableNode *Parser::GetNextToken(EvaluableNode *parent_node, EvaluableNode *new_token)
+EvaluableNode *Parser::GetNextToken(EvaluableNode *parent_node, EvaluableNode *reuse_assoc_token_as_value)
 {
-	if(new_token == nullptr)
+	EvaluableNode *new_token = nullptr;
+	bool parsing_assoc_key = false;
+
+	if(reuse_assoc_token_as_value == nullptr)
+	{
 		new_token = evaluableNodeManager->AllocNode(ENT_NULL);
+		//if parsing an assoc but haven't been passed a value to reuse, it's a key
+		if(parent_node != nullptr && parent_node->IsAssociativeArray())
+			parsing_assoc_key = true;
+	}
+	else
+	{
+		new_token = reuse_assoc_token_as_value;
+	}
 
 	SkipWhitespaceAndAccumulateAttributes(new_token);
 	if(pos >= code->size())
@@ -497,7 +509,8 @@ EvaluableNode *Parser::GetNextToken(EvaluableNode *parent_node, EvaluableNode *n
 		FreeNode(new_token);
 		return nullptr;
 	}
-	else if(StringManipulation::IsUtf8ArabicNumerals(cur_char) || cur_char == '-' || cur_char == '.')
+	else if(!parsing_assoc_key
+		&& (StringManipulation::IsUtf8ArabicNumerals(cur_char) || cur_char == '-' || cur_char == '.'))
 	{
 		size_t start_pos = pos;
 		SkipToEndOfIdentifier();

src/Amalgam/Parser.h

@@ -166,8 +166,8 @@ class Parser
 
 	//Returns a EvaluableNode containing the next token, null if none left in current context
 	// parent_node is primarily to check for errors or warnings
-	//if new_token is not nullptr, it will put the token in the EvaluableNode provided, otherwise will return a new one
-	EvaluableNode *GetNextToken(EvaluableNode *parent_node, EvaluableNode *new_token = nullptr);
+	//if reuse_assoc_token_as_value is not nullptr, it will put the token in the EvaluableNode provided, otherwise will return a new one
+	EvaluableNode *GetNextToken(EvaluableNode *parent_node, EvaluableNode *reuse_assoc_token_as_value = nullptr);
 
 	//deallocates the current node in case there is an early exit or error
 	void FreeNode(EvaluableNode *node);

src/Amalgam/SeparableBoxFilterDataStore.h

@@ -543,15 +543,32 @@ class SeparableBoxFilterDataStore
 		SortedIntegerSet &entity_indices, size_t query_feature_index, size_t absolute_feature_index, bool high_accuracy)
 	{
 		size_t num_entity_indices = entity_indices.size();
+		size_t max_index = num_entity_indices;
 
 		auto &partial_sums = parametersAndBuffers.partialSums;
 		const auto accum_location = partial_sums.GetAccumLocation(query_feature_index);
+		size_t max_element = partial_sums.numInstances;
+
+		auto &entity_indices_vector = entity_indices.GetIntegerVector();
+
+		//it's almost always faster to just accumulate an index than to check if it is a valid index
+		// and then only accumulate if it is valid
+		//however, indices beyond the range of partial_sums will cause an issue
+		//therefore, only trim back the end if needed, and trim back to the largest possible element id (max_element - 1)
+		if(entity_indices.GetEndInteger() >= max_element)
+		{
+			max_index = entity_indices.GetFirstIntegerVectorLocationGreaterThan(max_element - 1);
+			num_entity_indices = max_index - 1;
+		}
 
 		auto &column_data = columnData[absolute_feature_index];
 
 		//for each found element, accumulate associated partial sums
-		for(size_t entity_index : entity_indices)
+		#pragma omp parallel for schedule(static) if(max_index > 300)
+		for(int64_t i = 0; i < static_cast<int64_t>(max_index); i++)
 		{
+			const auto entity_index = entity_indices_vector[i];
+
 			//get value
 			auto other_value_type = column_data->GetIndexValueType(entity_index);
 			auto other_value = column_data->GetResolvedValue(other_value_type, GetValue(entity_index, absolute_feature_index));