convertQuery.go

/*
This program has been developed by students from the bachelor Computer Science at Utrecht University within the Software Project course.
© Copyright Utrecht University (Department of Information and Computing Sciences)
*/

package aql

import (
	"errors"
	"fmt"
	"strconv"

	"git.science.uu.nl/graphpolaris/query-conversion/entity"
)

// Version 1.13

/*
ConvertQuery converts an IncomingQueryJSON object into AQL
	JSONQuery: *entity.IncomingQueryJSON, the query to be converted to AQL
	Returns: *string, the AQL query and a possible error
*/
func (s *Service) ConvertQuery(JSONQuery *entity.IncomingQueryJSON) (*string, error) {

	// Check to make sure all indexes exist
	// The largest possible id for an entity
	largestEntityID := len(JSONQuery.Entities) - 1
	// The largest possible id for a relation
	largestRelationID := len(JSONQuery.Relations) - 1

	// Make sure no entity should be returned that is outside the range of that list
	for _, e := range JSONQuery.Return.Entities {
		// If this entity references an entity that is outside the range
		if e > largestEntityID || e < 0 {
			return nil, errors.New("non-existing entity referenced in return")
		}
	}

	// Make sure that no relation mentions a non-existing entity
	for _, r := range JSONQuery.Relations {
		if r.EntityFrom > largestEntityID || r.EntityTo > largestEntityID {
			return nil, errors.New("non-exisiting entity referenced in relation")
		}
	}

	// Make sure no non-existing relation is tried to be returned
	for _, r := range JSONQuery.Return.Relations {
		if r > largestRelationID || r < 0 {
			return nil, errors.New("non-existing relation referenced in return")
		}
	}
	result := createQuery(JSONQuery)
	return result, nil
}

/*
createQuery generates a query based on the json file provided
	JSONQuery: *entity.IncomingQueryJSON, this is a parsedJSON struct holding all the data needed to form a query,
	Return: *string, a string containing the corresponding AQL query and an error
*/
func createQuery(JSONQuery *entity.IncomingQueryJSON) *string {
	// Note: Case #4, where there is an edge only query (without any entity), is not supported by frontend

	// If a modifier is used, disable the limit
	if len(JSONQuery.Modifiers) > 0 {
		JSONQuery.Limit = -1
	}

	var (
		relationsToReturn []string
		nodesToReturn     []string
		nodeUnion         string
		relationUnion     string
	)
	// If we've already used an entity we can set the value to true so we skip it in the result later
	entityDone := make(map[int]bool)
	for o := range JSONQuery.Entities {
		entityDone[o] = false
	}
	// Loop over all relations
	ret := ""

	// Add a WITH statement for entityTo
	includedTypes := make(map[string]bool)
	allTypes := make(map[string]bool)
	for _, relation := range JSONQuery.Relations {
		if relation.EntityFrom >= 0 {
			includedTypes[JSONQuery.Entities[relation.EntityFrom].Type] = true
			allTypes[JSONQuery.Entities[relation.EntityFrom].Type] = true

			// If the type is in the entityTo it is a valid type but not yet included
			if relation.EntityTo >= 0 {
				allTypes[JSONQuery.Entities[relation.EntityTo].Type] = true
			}
		}
		if relation.EntityFrom == -1 && relation.EntityTo >= 0 {
			includedTypes[JSONQuery.Entities[relation.EntityTo].Type] = true
			allTypes[JSONQuery.Entities[relation.EntityTo].Type] = true
		}
	}

	// Include all types that are not yet included
	first := true
	for k := range allTypes {
		if !includedTypes[k] {
			if first {
				ret += fmt.Sprintf("WITH %v", k)
				first = false
			} else {
				ret += fmt.Sprintf(", %v", k)
			}
		}
	}
	if !first {
		ret += "\n"
	}

	for i, relation := range JSONQuery.Relations {

		relationName := fmt.Sprintf("r%v", i)

		if relation.EntityFrom >= 0 {
			// if there is a from-node
			// create the let for this node
			// IF WE'VE ALREADY SEEN THIS ENTITY WE DON'T HAVE TO REQUERY IT, WE CAN JUST REUSE THE LET BINDING
			if !entityDone[relation.EntityFrom] {
				fromName := fmt.Sprintf("n%v", relation.EntityFrom)

				ret += *createNodeLet(&JSONQuery.Entities[relation.EntityFrom], &fromName)
				entityDone[relation.EntityFrom] = true
			}
			var function *entity.QueryFunctionStruct
			for _, f := range JSONQuery.Functions {
				if (f.GroupID == relation.EntityFrom && f.ByID == relation.EntityTo) || (f.GroupID == relation.EntityTo && f.ByID == relation.EntityFrom) {
					function = &f
				}
			}
			ret += *createRelationLetWithFromEntity(&relation, relationName, &JSONQuery.Entities, JSONQuery.Limit, function)

		} else if relation.EntityTo >= 0 {
			fmt.Println("Joris are you a madman! How did this happen?")
			// if there is only a to-node
			if !entityDone[relation.EntityTo] {
				toName := fmt.Sprintf("n%v", relation.EntityTo)

				ret += *createNodeLet(&JSONQuery.Entities[relation.EntityTo], &toName)
				entityDone[relation.EntityTo] = true
			}

			ret += *createRelationLetWithOnlyToEntity(&relation, relationName, &JSONQuery.Entities, JSONQuery.Limit)
			// Add this relation to the list
		} else {
			fmt.Println("Relation-only queries are currently not supported")
			continue
		}

		// Add this relation to the list
		relationsToReturn = append(relationsToReturn, relationName)
	}

	// Add node let statements for nodes that are not yet returned
	// Create a set from all the entity-from's and entity-to's, to check if they are returned
	nodeSet := make(map[int]bool)
	for _, relation := range JSONQuery.Relations {
		nodeSet[relation.EntityFrom] = true
		nodeSet[relation.EntityTo] = true
	}

	// Check if the entities to return are already returned
	for _, entityIndex := range JSONQuery.Return.Entities {
		if !nodeSet[entityIndex] {
			// If not, return this node
			name := fmt.Sprintf("n%v", entityIndex)
			ret += *createNodeLet(&JSONQuery.Entities[entityIndex], &name)

			// Add this node to the list
			nodesToReturn = append(nodesToReturn, name)
		}
	}

	if len(JSONQuery.Functions) == 0 {
		//If there are modifiers within the query, we run a different set of checks which focus on quantifiable aspects
		if len(JSONQuery.Modifiers) > 0 {
			modifier := JSONQuery.Modifiers[0]
			// There is a distinction between (relations and entities) and (relations or entities)
			if len(JSONQuery.Return.Relations) > 0 && len(JSONQuery.Return.Entities) > 0 {

				var pathDistinction string // .vertices or .edges

				// Select the correct addition to the return of r0[**]
				if modifier.SelectedType == "entity" {
					// ASSUMING THERE IS ONLY 1 RELATION
					if JSONQuery.Relations[0].EntityFrom == modifier.SelectedTypeID {
						// This should always be 0, because that is the start of the path
						pathDistinction = ".vertices[0]"

					} else {
						// Otherwise take the depth.max -1 to get the last
						pathDistinction = fmt.Sprintf(".vertices[%v]", JSONQuery.Relations[0].Depth.Max)

					}
				} else {
					pathDistinction = ".edges[**]"
				}

				// Getting the attribute if there is one
				if modifier.AttributeIndex != -1 {
					if modifier.SelectedType == "entity" {
						pathDistinction += fmt.Sprintf(".%v", JSONQuery.Entities[modifier.SelectedTypeID].Constraints[modifier.AttributeIndex].Attribute)

					} else {
						pathDistinction += fmt.Sprintf(".%v", JSONQuery.Relations[modifier.SelectedTypeID].Constraints[modifier.AttributeIndex].Attribute)

					}
				}

				// If count is used it has to be replaced with Length + unique else use the modifier type
				if modifier.Type == "COUNT" {
					ret += fmt.Sprintf("RETURN LENGTH (unique(r0[*]%v))", pathDistinction)

				} else {
					ret += fmt.Sprintf("RETURN %v (r0[*]%v)", modifier.Type, pathDistinction)

				}

			} else {
				// Check if the modifier is on an attribute
				if modifier.AttributeIndex == -1 {
					ret += fmt.Sprintf("RETURN LENGTH (n%v)", modifier.SelectedTypeID)
				} else {
					var attribute string

					// Selecting the right attribute from either the entity constraint or relation constraint
					if modifier.SelectedType == "entity" {
						attribute = JSONQuery.Entities[modifier.SelectedTypeID].Constraints[modifier.AttributeIndex].Attribute

					} else {
						attribute = JSONQuery.Relations[modifier.SelectedTypeID].Constraints[modifier.AttributeIndex].Attribute

					}

					// If count is used it has to be replaced with Length + unique else use the modifier type
					if modifier.Type == "COUNT" {
						ret += fmt.Sprintf("RETURN LENGTH (unique(n%v[*].%v))", modifier.SelectedTypeID, attribute)

					} else {
						ret += fmt.Sprintf("RETURN %v (n%v[*].%v)", modifier.Type, modifier.SelectedTypeID, attribute)

					}
				}
			}

		} else {

			// Create UNION statements that create unique lists of all the nodes and relations
			// Thus removing all duplicates
			nodeUnion = "\nLET nodes = first(RETURN UNION_DISTINCT("
			for _, relation := range relationsToReturn {
				nodeUnion += fmt.Sprintf("flatten(%v[**].vertices), ", relation)
			}

			for _, node := range nodesToReturn {
				nodeUnion += fmt.Sprintf("%v,", node)
			}
			nodeUnion += "[],[]))\n"

			relationUnion = "LET edges = first(RETURN UNION_DISTINCT("
			for _, relation := range relationsToReturn {
				relationUnion += fmt.Sprintf("flatten(%v[**].edges), ", relation)
			}
			relationUnion += "[],[]))\n"

			ret += nodeUnion + relationUnion
			ret += "RETURN {\"vertices\":nodes, \"edges\":edges }"

		}
	} else {
		ret += createTableWithFunctions(JSONQuery.Functions, JSONQuery.Relations, JSONQuery.Entities)
	}

	return &ret
}

/*
createNodeLet generates a 'LET' statement for a node related query
	node: *entity.QueryEntityStruct, node is an entityStruct containing the information of a single nod,
	name: *string, is the autogenerated name of the node consisting of "n" + the index of the node,
	Return: *string, a string containing a single LET-statement in AQL
*/
func createNodeLet(node *entity.QueryEntityStruct, name *string) *string {
	header := fmt.Sprintf("LET %v = (\n\tFOR x IN %v \n", *name, node.Type)
	footer := "\tRETURN x\n)\n"
	constraints := *createConstraintStatements(&node.Constraints, "x", false)

	ret := header + constraints + footer
	return &ret
}

/*
createRelationLetWithFromEntity generates a 'LET' statement for relations with an 'EntityFrom' property and optionally an 'EntitiyTo' property
	relation: *entity.QueryRekationStruct, relation is a relation struct containing the information of a single relation,
	name: string, is the autogenerated name of the node consisting of "r" + the index of the relation,
	entities: *[]entity.QueryEntityStrucy, is a list of entityStructs that are needed to form the relation LET-statement,
	Return: *string, a string containing a single LET-statement in AQL
*/
func createRelationLetWithFromEntity(relation *entity.QueryRelationStruct, name string, entities *[]entity.QueryEntityStruct, limit int, function *entity.QueryFunctionStruct) *string {
	header := fmt.Sprintf("LET %v = (\n\tFOR x IN n%v \n", name, relation.EntityFrom)
	forStatement := fmt.Sprintf("\tFOR v, e, p IN %v..%v OUTBOUND x %s \n", relation.Depth.Min, relation.Depth.Max, relation.Type)

	// Guarantees that there is no path returned with a duplicate edge
	// This way there are no cycle paths possible, TODO: more research about this needed
	optionStmtn := "\tOPTIONS { uniqueEdges: \"path\" }\n"

	vFilterStmnt := ""
	if relation.EntityTo != -1 {
		// If there is a to-node, generate the filter statement
		toConstraints := (*entities)[relation.EntityTo].Constraints
		vFilterStmnt += *createConstraintStatements(&toConstraints, "v", false)
	}

	relationFilterStmnt := *createConstraintStatements(&relation.Constraints, "p", true)

	// Dont use a limit on quantifing queries
	footer := ""
	if limit != -1 {
		footer += fmt.Sprintf("\tLIMIT %v \n", limit)
	}
	if function != nil {
		footer += "RETURN DISTINCT v )\n"
	} else {
		footer += "RETURN DISTINCT p )\n"
	}

	ret := header + forStatement + optionStmtn + vFilterStmnt + relationFilterStmnt + footer
	return &ret
}

/*
createRelationLetWithOnlyToEntity generates a 'LET' statement for relations with only an 'EntityTo' property
	relation: *entity.QueryRelationStruct, relation is a relation struct containing the information of a single relation,
	name: string, is the autogenerated name of the node consisting of "r" + the index of the relation,
	entities: *[]entity.QueryEntityStruct, is a list of entityStructs that are needed to form the relation LET-statement,
	Return: *string, a string containing a single LET-statement in AQL
*/
func createRelationLetWithOnlyToEntity(relation *entity.QueryRelationStruct, name string, entities *[]entity.QueryEntityStruct, limit int) *string {
	header := fmt.Sprintf("LET %v = (\n\tFOR x IN n%v \n", name, relation.EntityTo)
	forStatement := fmt.Sprintf("\tFOR v, e, p IN %v..%v INBOUND x %s \n", relation.Depth.Min, relation.Depth.Max, relation.Type)

	// Guarantees that there is no path returned with a duplicate edge
	// This way there are no cycle paths possible, TODO: more research about this needed
	optionStmtn := "\tOPTIONS { uniqueEdges: \"path\" }\n"

	relationFilterStmnt := *createConstraintStatements(&relation.Constraints, "p", true)

	// Dont use a limit on quantifing queries
	footer := ""
	if limit != -1 {
		footer += fmt.Sprintf("\tLIMIT %v \n", limit)
	}
	footer += "RETURN DISTINCT p )\n"

	ret := header + forStatement + optionStmtn + relationFilterStmnt + footer
	return &ret
}

type variableNameGeneratorToken struct {
	token int
}

func newVariableNameGeneratorToken() *variableNameGeneratorToken {
	v := variableNameGeneratorToken{token: 0}
	return &v
}

func variableNameGenerator(vngt *variableNameGeneratorToken) string {
	result := "variable_" + strconv.Itoa(vngt.token)
	vngt.token++
	return result
}

func createTableWithFunctions(functions []entity.QueryFunctionStruct, relations []entity.QueryRelationStruct, entities []entity.QueryEntityStruct) string {
	result := ""
	v := newVariableNameGeneratorToken()
	for _, function := range functions {
		for j, relation := range relations {
			if (function.GroupID == relation.EntityFrom && function.ByID == relation.EntityTo) || (function.GroupID == relation.EntityTo && function.ByID == relation.EntityFrom) {
				if function.Type == "groupBy" {
					a := variableNameGenerator(v)
					b := variableNameGenerator(v)
					c := variableNameGenerator(v)
					d := variableNameGenerator(v)
					e := variableNameGenerator(v)
					f := variableNameGenerator(v)
					g := variableNameGenerator(v)
					h := variableNameGenerator(v)
					rName := fmt.Sprintf("r%v", j)
					nName := fmt.Sprintf("n%v", relation.EntityFrom)

					result += "LET " + a + " = (\n\tFOR r IN " +
						relation.Type + "\n\tLET " + b +
						" = (\n\t\tFOR c IN " + rName +
						" \n\t\tFILTER c._id == r._to \n\t\tRETURN c." + function.GroupAttribute + "\n\t) " +
						"\n\tLET " + c + " = " + b + "[0] \n\tLET " + d + " = (\n\t\t" +
						"FOR p in " + nName +
						" \n\t\tFILTER p._id == r._from \n\t\tRETURN p." + function.ByAttribute + "\n\t) " +
						"\n\tLET " + e + " = " + d + "[0] \n\tRETURN {\n\t\t\"" + f + "\" : " + c + ", \n\t\t" +
						"\"" + g + "\" : " + e + "\n\t}\n) \n" +
						"LET function_" + strconv.Itoa(function.TypeID) + " = (\n\tFOR r in " + a + " \n\tCOLLECT c = r." + f + " INTO groups = r." + g + " \n\t\t" +
						"LET " + h + " = " + function.AppliedModifier + "(groups) \n\t"
					if len(function.Constraints) > 0 {
						result += "FILTER " + h + " " + wordsToLogicalSign(function.Constraints[0].MatchType) + " " + function.Constraints[0].Value + " \n\t"
					}
					result += "RETURN {\n\t\t" + function.GroupAttribute + " : c, \n\t\t" +
						function.AppliedModifier + "_" + function.ByAttribute + " : " + h + "\n\t}\n) \n"
				}
			}

		}
	}
	result += "RETURN {"
	if len(functions) > 1 {
		for l := 0; l < len(functions)-1; l++ {
			result += "function_" + strconv.Itoa(functions[l].TypeID) + ", "
		}
	}
	result += "function_" + strconv.Itoa(functions[len(functions)-1].TypeID) + "}"
	return result
}

func wordsToLogicalSign(word string) string {
	if word == "LT" {
		return "<"
	} else if word == "LTE" {
		return "<="
	} else if word == "EQ" {
		return "=="
	} else if word == "GTE" {
		return ">="
	} else {
		return ">"
	}
}