mcvsubbu commented on a change in pull request #4747: Data Anonymizer Tool URL: https://github.com/apache/incubator-pinot/pull/4747#discussion_r340776589
########## File path: pinot-tools/src/main/java/org/apache/pinot/tools/PinotDataAndQueryAnonymizer.java ########## @@ -0,0 +1,1332 @@ +/** + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ +package org.apache.pinot.tools; + +import com.google.common.annotations.VisibleForTesting; +import com.google.common.base.Preconditions; +import com.google.common.base.Stopwatch; +import java.io.BufferedReader; +import java.io.BufferedWriter; +import java.io.File; +import java.io.FileInputStream; +import java.io.FileWriter; +import java.io.InputStream; +import java.io.InputStreamReader; +import java.io.PrintWriter; +import java.util.Arrays; +import java.util.Comparator; +import java.util.HashMap; +import java.util.HashSet; +import java.util.List; +import java.util.Map; +import java.util.Random; +import java.util.Set; +import java.util.concurrent.TimeUnit; +import org.apache.avro.SchemaBuilder; +import org.apache.avro.file.DataFileWriter; +import org.apache.avro.generic.GenericData; +import org.apache.avro.generic.GenericDatumWriter; +import org.apache.commons.lang.RandomStringUtils; +import org.apache.pinot.common.data.DateTimeFieldSpec; +import org.apache.pinot.common.data.DimensionFieldSpec; +import org.apache.pinot.common.data.FieldSpec; +import org.apache.pinot.common.data.MetricFieldSpec; +import org.apache.pinot.common.data.Schema; +import org.apache.pinot.common.data.TimeFieldSpec; +import org.apache.pinot.common.segment.ReadMode; +import org.apache.pinot.core.data.GenericRow; +import org.apache.pinot.core.data.readers.PinotSegmentRecordReader; +import org.apache.pinot.core.indexsegment.immutable.ImmutableSegment; +import org.apache.pinot.core.indexsegment.immutable.ImmutableSegmentLoader; +import org.apache.pinot.core.segment.index.ColumnMetadata; +import org.apache.pinot.core.segment.index.SegmentMetadataImpl; +import org.apache.pinot.core.segment.index.readers.Dictionary; +import org.apache.pinot.pql.parsers.Pql2Compiler; +import org.apache.pinot.pql.parsers.pql2.ast.AstNode; +import org.apache.pinot.pql.parsers.pql2.ast.BetweenPredicateAstNode; +import org.apache.pinot.pql.parsers.pql2.ast.BooleanOperatorAstNode; +import org.apache.pinot.pql.parsers.pql2.ast.ComparisonPredicateAstNode; +import org.apache.pinot.pql.parsers.pql2.ast.FunctionCallAstNode; +import org.apache.pinot.pql.parsers.pql2.ast.GroupByAstNode; +import org.apache.pinot.pql.parsers.pql2.ast.IdentifierAstNode; +import org.apache.pinot.pql.parsers.pql2.ast.InPredicateAstNode; +import org.apache.pinot.pql.parsers.pql2.ast.LiteralAstNode; +import org.apache.pinot.pql.parsers.pql2.ast.OutputColumnAstNode; +import org.apache.pinot.pql.parsers.pql2.ast.OutputColumnListAstNode; +import org.apache.pinot.pql.parsers.pql2.ast.PredicateAstNode; +import org.apache.pinot.pql.parsers.pql2.ast.PredicateListAstNode; +import org.apache.pinot.pql.parsers.pql2.ast.SelectAstNode; +import org.apache.pinot.pql.parsers.pql2.ast.StarColumnListAstNode; +import org.apache.pinot.pql.parsers.pql2.ast.StarExpressionAstNode; +import org.apache.pinot.pql.parsers.pql2.ast.StringLiteralAstNode; +import org.apache.pinot.pql.parsers.pql2.ast.WhereAstNode; +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; + + +/** + * The goal of this tool is to generate test dataset (as Avro files) with + * characteristics similar to a given source dataset. The source dataset is + * a set of Pinot segments. The tool can be used in situations where actual + * source data isn't allowed to be used for the purpose of testing (regression, + * performance, functional, evaluation of other OLAP systems etc). + * + * The tool understands the characteristics of the given dataset (Pinot segments) + * and generate corresponding random data while preserving those characteristics. + * The tool can then also be used to generate queries for the random data. + * + * So if we have a set of production data which you want to use for testing + * but are unable to do so (because of security restrictions etc), then this tool + * can be used to generate corresponding anonymous data and queries. Users can then + * use the anonymized dataset (avro files) and generated queries for their testing. + * + * One avro file is generated per input Pinot segment. The tool also randomizes the + * column names (and table name) so that source schema is not revealed. The user is also + * allowed to provide a set of columns for which they want the data to be retained + * as is (not anonymized). User should be careful when choosing these columns. Ideally + * these should be time (or time related) columns since they don't reveal anything and so + * it is fine to copy them as is from souce segments into Avro files. + * + * Steps to use this tool are as follows: + * + * STEP 1 - Download a day’s queries (same day when we downloaded the segments) from + * Pinot broker query log. You may have to post-process the log to remove some noise + * and just keep queries + * + * STEP 2 - Point the tool to the query file. It will parse each query + * and output the set of columns participating in filters (WHERE clause). + * + * STEP 3 - Point the tool to directory containing Pinot segments along with + * the set of filter columns (identified in previous step) and their cardinalities + * to anonymize data. This step will first read dictionaries from each input segment + * and build a sorted global dictionary containing 1-1 mapping between an original + * column value and corresponding column value. Global dictionary will be built + * only for the set of filter columns. For remaining columns, we can generate any + * arbitrary random value directly into the Avro file directly. + * + * Please see the implementation notes further in the code explaining the global + * dictionary building phase and data generation phase in detail. + * + * STEP 4 - Point the tool to directory containing global dictionaries and column + * name mapping -- this directory will be the same as the outputDir used in previous + * step where the tool wrote avro files. Also point the tool to queryDir containing + * source query files. The tool will load the global dictionaries and use them to + * rewrite the WHERE part of the query. Similarly, the column name mapping file + * will be used to rewrite the select list of the original query. A file named + * "queries.generated" will be written into queryDir. + * + * Please see implementation notes further in the code explaining the query + * generation phase in detail. + * + * Also, please see usage examples in + * {@link org.apache.pinot.tools.admin.command.AnonymizeDataCommand} to learn + * how this tool can be invoked from command line. + * + * Limitations: + * (1) Add support for multi-value columns + * (2) Add support for BYTES type + * (3) Add support for partitioning (where dataset is hash partitioned on column) + * (4) Add support for ORDER BY in query generator + * (5) Potential memory explosion for extreme high cardinality global dictionary columns + */ +public class PinotDataAndQueryAnonymizer { + private static final Logger LOGGER = LoggerFactory.getLogger(PinotDataAndQueryAnonymizer.class); + + private final static int INT_BASE_VALUE = 1000; + private final static long LONG_BASE_VALUE = 100000; + private static final float FLOAT_BASE_VALUE = 100.23f; + private static final double DOUBLE_BASE_VALUE = 1000.2375; + private static final String DICT_FILE_EXTENSION = ".dict"; + private static final String COLUMN_MAPPING_FILE_KEY = "columns.mapping"; + private static final String COLUMN_MAPPING_SEPARATOR = ":"; + + private final String _outputDir; + private int _numFilesToGenerate; + private final String _segmentDir; + private final String _filePrefix; + // dictionaries used to generate data with same cardinality and data distribution + // as in source table segments + private final Map<String, OrigAndDerivedValueHolder> _origToDerivedValueGlobalDictionary; + // used to map the original column name to a generated column name + private final Map<String, String> _origToDerivedColumnsMap; + + private final Stopwatch _timeToBuildDictionaries = Stopwatch.createUnstarted(); + private final Stopwatch _timeToGenerateAvroFiles = Stopwatch.createUnstarted(); + + private Schema _pinotSchema = null; + private org.apache.avro.Schema _avroSchema = null; + + private final Map<String, FieldSpec.DataType> _columnToDataTypeMap; + // name of columns to build global dictionary for and corresponding total cardinality + private final Map<String, Integer> _globalDictionaryColumns; + // name of time (or time derived) columns for which we will retain data + private final Set<String> _timeColumns; + + private String[] _segmentDirectories; + + /** + * Create an instance of PinotDataGenerator + * @param outputDir parent directory where avro files will be generated + * @param segmentDir directory containing segment + * @param fileNamePrefix generated avro file name prefix + */ + public PinotDataAndQueryAnonymizer( + String segmentDir, + String outputDir, + String fileNamePrefix, + Map<String, Integer> globalDictionaryColumns, + Set<String> timeColumns) { + _outputDir = outputDir; + _segmentDir = segmentDir; + _filePrefix = fileNamePrefix; + _origToDerivedValueGlobalDictionary = new HashMap<>(); + _origToDerivedColumnsMap = new HashMap<>(); + _columnToDataTypeMap = new HashMap<>(); + _globalDictionaryColumns = globalDictionaryColumns; + _timeColumns = timeColumns; + + for (String column : timeColumns) { + // sometime the predicates can also be on columns which the user + // wants to retain the data for as is and thus these columns will be + // part of filter column set as well. + // But since the values are retained for these columns, we + // don't need to build global dictionary for them. So remove + // these columns from the filter column set. + _globalDictionaryColumns.remove(column); + } + + LOGGER.info("Columns to retain data for"); + for (String column : _timeColumns) { + LOGGER.info("Column name: " + column); + } + + LOGGER.info("Columns to build global dictionary for"); + for (Map.Entry<String, Integer> entry : _globalDictionaryColumns.entrySet()) { + LOGGER.info("Column name: " + entry.getKey() + " cardinality: " + entry.getValue()); + } + } + + /***************************************************** + * * + * Global Dictionary Builder * + * * + *****************************************************/ + + /* + * Global Dictionary Implementation Notes + * + * We build global dictionary for a specific set of columns Review comment: You can reduce the number of lines in these comments by making lines wider? ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
