Skip to content

Commit

Permalink
support ngql reader (#78)
Browse files Browse the repository at this point in the history
  • Loading branch information
Nicole00 authored Jan 9, 2023
1 parent 0e037b7 commit abf030e
Show file tree
Hide file tree
Showing 12 changed files with 554 additions and 16 deletions.
Original file line number Diff line number Diff line change
Expand Up @@ -164,4 +164,25 @@ object NebulaSparkReaderExample {
vertex.show(20)
println("vertex count: " + vertex.count())
}

def readEdgeWithNgql(spark: SparkSession): Unit = {
LOG.info("start to read nebula edge with ngql")
val config =
NebulaConnectionConfig
.builder()
.withMetaAddress("127.0.0.1:9559")
.withGraphAddress("127.0.0.1:9669")
.withConenctionRetry(2)
.build()
val nebulaReadConfig: ReadNebulaConfig = ReadNebulaConfig
.builder()
.withSpace("test")
.withLabel("friend")
.withNgql("match (v)-[e:friend]-(v2) return e")
.build()
val edge = spark.read.nebula(config, nebulaReadConfig).loadEdgesToDfByNgql()
edge.printSchema()
edge.show(20)
println("veedgertex count: " + edge.count())
}
}
Original file line number Diff line number Diff line change
Expand Up @@ -670,20 +670,45 @@ object WriteNebulaEdgeConfig {
* you can set noColumn to true to read no vertex col, and you can set returnCols to read the specific cols, if the returnCols is empty, then read all the columns.
* you can set partitionNum to define spark partition nums to read nebula graph.
*/
class ReadNebulaConfig(space: String,
label: String,
returnCols: List[String],
noColumn: Boolean,
partitionNum: Int,
limit: Int)
extends Serializable {
def getSpace = space
def getLabel = label
def getReturnCols = returnCols
def getNoColumn = noColumn
def getPartitionNum = partitionNum
def getLimit = limit
class ReadNebulaConfig extends Serializable {
var getSpace: String = _
var getLabel: String = _
var getReturnCols: List[String] = _
var getNoColumn: Boolean = _
var getPartitionNum: Int = _
var getLimit: Int = _
var getNgql: String = _
// todo add filter
def this(space: String,
label: String,
returnCols: List[String],
noColumn: Boolean,
partitionNum: Int,
limit: Int) = {
this()
this.getSpace = space
this.getLabel = label
this.getReturnCols = returnCols
this.getNoColumn = noColumn
this.getPartitionNum = partitionNum
this.getLimit = limit
}

def this(space: String,
label: String,
returnCols: List[String],
noColumn: Boolean,
ngql: String,
limit: Int) = {
this()
this.getNgql = ngql
this.getSpace = space
this.getLabel = label
this.getReturnCols = returnCols
this.getNoColumn = noColumn
this.getLimit = limit
this.getPartitionNum = 1
}
}

/**
Expand All @@ -699,6 +724,7 @@ object ReadNebulaConfig {
var noColumn: Boolean = false
var partitionNum: Int = 100
var limit: Int = 1000
var ngql: String = _

def withSpace(space: String): ReadConfigBuilder = {
this.space = space
Expand Down Expand Up @@ -740,9 +766,18 @@ object ReadNebulaConfig {
this
}

def withNgql(ngql: String): ReadConfigBuilder = {
this.ngql = ngql
this
}

def build(): ReadNebulaConfig = {
check()
new ReadNebulaConfig(space, label, returnCols.toList, noColumn, partitionNum, limit)
if (ngql != null && !ngql.isEmpty) {
new ReadNebulaConfig(space, label, returnCols.toList, noColumn, ngql, limit)
} else {
new ReadNebulaConfig(space, label, returnCols.toList, noColumn, partitionNum, limit)
}
}

private def check(): Unit = {
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -108,11 +108,19 @@ class NebulaOptions(@transient val parameters: CaseInsensitiveMap[String])(
var partitionNums: String = _
var noColumn: Boolean = _
var limit: Int = _
var ngql: String = _
if (operaType == OperaType.READ) {
returnCols = parameters(RETURN_COLS)
noColumn = parameters.getOrElse(NO_COLUMN, false).toString.toBoolean
partitionNums = parameters(PARTITION_NUMBER)
limit = parameters.getOrElse(LIMIT, DEFAULT_LIMIT).toString.toInt
ngql = parameters.getOrElse(NGQL,EMPTY_STRING)
ngql = parameters.getOrElse(NGQL,EMPTY_STRING)
if(ngql!=EMPTY_STRING){
require(parameters.isDefinedAt(GRAPH_ADDRESS),
s"option $GRAPH_ADDRESS is required for ngql and can not be blank")
graphAddress = parameters(GRAPH_ADDRESS)
}
}

/** write parameters */
Expand Down Expand Up @@ -235,6 +243,9 @@ object NebulaOptions {
val PARTITION_NUMBER: String = "partitionNumber"
val LIMIT: String = "limit"

/** read by ngql **/
val NGQL: String = "ngql"

/** write config */
val RATE_LIMIT: String = "rateLimit"
val VID_POLICY: String = "vidPolicy"
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -9,7 +9,11 @@ import java.util.Map.Entry
import java.util.Optional

import com.vesoft.nebula.connector.exception.IllegalOptionException
import com.vesoft.nebula.connector.reader.{NebulaDataSourceEdgeReader, NebulaDataSourceVertexReader}
import com.vesoft.nebula.connector.reader.{
NebulaDataSourceEdgeReader,
NebulaDataSourceNgqlEdgeReader,
NebulaDataSourceVertexReader
}
import com.vesoft.nebula.connector.writer.{NebulaDataSourceEdgeWriter, NebulaDataSourceVertexWriter}
import org.apache.spark.sql.SaveMode
import org.apache.spark.sql.catalyst.util.CaseInsensitiveMap
Expand Down Expand Up @@ -46,6 +50,8 @@ class NebulaDataSource

if (DataTypeEnum.VERTEX == DataTypeEnum.withName(dataType)) {
new NebulaDataSourceVertexReader(nebulaOptions)
} else if (nebulaOptions.ngql != null && nebulaOptions.ngql.nonEmpty) {
new NebulaDataSourceNgqlEdgeReader(nebulaOptions)
} else {
new NebulaDataSourceEdgeReader(nebulaOptions)
}
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -110,6 +110,45 @@ package object connector {
dfReader.load()
}

/**
* Reading edges from Nebula Graph by ngql
* @return DataFrame
*/
def loadEdgesToDfByNgql(): DataFrame = {
assert(connectionConfig != null && readConfig != null,
"nebula config is not set, please call nebula() before loadEdgesToDfByNgql")

val dfReader = reader
.format(classOf[NebulaDataSource].getName)
.option(NebulaOptions.TYPE, DataTypeEnum.EDGE.toString)
.option(NebulaOptions.SPACE_NAME, readConfig.getSpace)
.option(NebulaOptions.LABEL, readConfig.getLabel)
.option(NebulaOptions.RETURN_COLS, readConfig.getReturnCols.mkString(","))
.option(NebulaOptions.NO_COLUMN, readConfig.getNoColumn)
.option(NebulaOptions.LIMIT, readConfig.getLimit)
.option(NebulaOptions.PARTITION_NUMBER, readConfig.getPartitionNum)
.option(NebulaOptions.NGQL, readConfig.getNgql)
.option(NebulaOptions.META_ADDRESS, connectionConfig.getMetaAddress)
.option(NebulaOptions.GRAPH_ADDRESS, connectionConfig.getGraphAddress)
.option(NebulaOptions.TIMEOUT, connectionConfig.getTimeout)
.option(NebulaOptions.CONNECTION_RETRY, connectionConfig.getConnectionRetry)
.option(NebulaOptions.EXECUTION_RETRY, connectionConfig.getExecRetry)
.option(NebulaOptions.ENABLE_META_SSL, connectionConfig.getEnableMetaSSL)
.option(NebulaOptions.ENABLE_STORAGE_SSL, connectionConfig.getEnableStorageSSL)

if (connectionConfig.getEnableStorageSSL || connectionConfig.getEnableMetaSSL) {
dfReader.option(NebulaOptions.SSL_SIGN_TYPE, connectionConfig.getSignType)
SSLSignType.withName(connectionConfig.getSignType) match {
case SSLSignType.CA =>
dfReader.option(NebulaOptions.CA_SIGN_PARAM, connectionConfig.getCaSignParam)
case SSLSignType.SELF =>
dfReader.option(NebulaOptions.SELF_SIGN_PARAM, connectionConfig.getSelfSignParam)
}
}

dfReader.load()
}

/**
* read nebula vertex edge to graphx's vertex
* use hash() for String type vertex id.
Expand Down
Original file line number Diff line number Diff line change
@@ -0,0 +1,158 @@
/* Copyright (c) 2022 vesoft inc. All rights reserved.
*
* This source code is licensed under Apache 2.0 License.
*/

package com.vesoft.nebula.connector.reader

import com.vesoft.nebula.Value
import com.vesoft.nebula.client.graph.data.{Relationship, ResultSet, ValueWrapper}
import com.vesoft.nebula.connector.NebulaUtils.NebulaValueGetter
import com.vesoft.nebula.connector.nebula.GraphProvider
import com.vesoft.nebula.connector.{NebulaOptions, NebulaUtils}
import org.apache.spark.sql.catalyst.InternalRow
import org.apache.spark.sql.catalyst.expressions.SpecificInternalRow
import org.apache.spark.sql.sources.v2.reader.InputPartitionReader
import org.apache.spark.sql.types.StructType
import org.slf4j.{Logger, LoggerFactory}

import scala.collection.JavaConversions.asScalaBuffer
import scala.collection.mutable
import scala.collection.mutable.ListBuffer

/**
* create reader by ngql
*/
class NebulaNgqlEdgePartitionReader extends InputPartitionReader[InternalRow] {

private val LOG: Logger = LoggerFactory.getLogger(this.getClass)

private var nebulaOptions: NebulaOptions = _
private var graphProvider: GraphProvider = _
private var schema: StructType = _
private var resultSet: ResultSet = _
private var edgeIterator: Iterator[ListBuffer[ValueWrapper]] = _

def this(nebulaOptions: NebulaOptions, schema: StructType) {
this()
this.schema = schema
this.nebulaOptions = nebulaOptions
this.graphProvider = new GraphProvider(
nebulaOptions.getGraphAddress,
nebulaOptions.timeout,
nebulaOptions.enableGraphSSL,
nebulaOptions.sslSignType,
nebulaOptions.caSignParam,
nebulaOptions.selfSignParam
)
// add exception when session build failed
graphProvider.switchSpace(nebulaOptions.user, nebulaOptions.passwd, nebulaOptions.spaceName)
resultSet = graphProvider.submit(nebulaOptions.ngql)
edgeIterator = query()
}

def query(): Iterator[ListBuffer[ValueWrapper]] = {
val edges: ListBuffer[ListBuffer[ValueWrapper]] = new ListBuffer[ListBuffer[ValueWrapper]]
val properties = nebulaOptions.getReturnCols
for (i <- 0 until resultSet.rowsSize()) {
val rowValues = resultSet.rowValues(i).values()
for (j <- 0 until rowValues.size()) {
val value = rowValues.get(j)
val valueType = value.getValue.getSetField
if (valueType == Value.EVAL) {
val relationship = value.asRelationship()
if (checkLabel(relationship)) {
edges.append(convertToEdge(relationship, properties))
}
} else if (valueType == Value.LVAL) {
val list: mutable.Buffer[ValueWrapper] = value.asList()
edges.appendAll(
list.toStream
.filter(e => checkLabel(e.asRelationship()))
.map(e => convertToEdge(e.asRelationship(), properties))
)
} else {
LOG.error(s"Exception convert edge type ${valueType} ")
throw new RuntimeException(" convert value type failed");
}
}
}
edges.iterator
}

def checkLabel(relationship: Relationship): Boolean = {
this.nebulaOptions.label.equals(relationship.edgeName())
}

def convertToEdge(relationship: Relationship,
properties: List[String]): ListBuffer[ValueWrapper] = {
val edge: ListBuffer[ValueWrapper] = new ListBuffer[ValueWrapper]
edge.append(relationship.srcId())
edge.append(relationship.dstId())
edge.append(new ValueWrapper(new Value(3, relationship.ranking()), "utf-8"))
if (properties == null || properties.isEmpty)
return edge
else {
for (i <- properties.indices) {
edge.append(relationship.properties().get(properties(i)))
}
}
edge
}

override def next(): Boolean = {
edgeIterator.hasNext
}

override def get(): InternalRow = {
val getters: Array[NebulaValueGetter] = NebulaUtils.makeGetters(schema)
val mutableRow = new SpecificInternalRow(schema.fields.map(x => x.dataType))

val edge = edgeIterator.next();
for (i <- getters.indices) {
val value: ValueWrapper = edge(i)
var resolved = false
if (value.isNull) {
mutableRow.setNullAt(i)
resolved = true
}
if (value.isString) {
getters(i).apply(value.asString(), mutableRow, i)
resolved = true
}
if (value.isDate) {
getters(i).apply(value.asDate(), mutableRow, i)
resolved = true
}
if (value.isTime) {
getters(i).apply(value.asTime(), mutableRow, i)
resolved = true
}
if (value.isDateTime) {
getters(i).apply(value.asDateTime(), mutableRow, i)
resolved = true
}
if (value.isLong) {
getters(i).apply(value.asLong(), mutableRow, i)
}
if (value.isBoolean) {
getters(i).apply(value.asBoolean(), mutableRow, i)
}
if (value.isDouble) {
getters(i).apply(value.asDouble(), mutableRow, i)
}
if (value.isGeography) {
getters(i).apply(value.asGeography(), mutableRow, i)
}
if (value.isDuration) {
getters(i).apply(value.asDuration(), mutableRow, i)
}
}
mutableRow

}

override def close(): Unit = {
graphProvider.close();
}
}
Original file line number Diff line number Diff line change
Expand Up @@ -21,3 +21,9 @@ class NebulaEdgePartition(index: Int, nebulaOptions: NebulaOptions, schema: Stru
override def createPartitionReader(): InputPartitionReader[InternalRow] =
new NebulaEdgePartitionReader(index, nebulaOptions, schema)
}

class NebulaNgqlEdgePartition(nebulaOptions: NebulaOptions, schema: StructType)
extends InputPartition[InternalRow] {
override def createPartitionReader(): InputPartitionReader[InternalRow] =
new NebulaNgqlEdgePartitionReader(nebulaOptions, schema)
}
Original file line number Diff line number Diff line change
Expand Up @@ -131,3 +131,16 @@ class NebulaDataSourceEdgeReader(nebulaOptions: NebulaOptions)
partitions.map(_.asInstanceOf[InputPartition[InternalRow]]).asJava
}
}

/**
* DataSourceReader for Nebula Edge by ngql
*/
class NebulaDataSourceNgqlEdgeReader(nebulaOptions: NebulaOptions)
extends NebulaSourceReader(nebulaOptions) {

override def planInputPartitions(): util.List[InputPartition[InternalRow]] = {
val partitions = new util.ArrayList[InputPartition[InternalRow]]()
partitions.add(new NebulaNgqlEdgePartition(nebulaOptions, getSchema))
partitions
}
}
Loading

0 comments on commit abf030e

Please sign in to comment.