[31294] | 1 | package org.hathitrust.extractedfeatures;
|
---|
| 2 |
|
---|
| 3 | import java.io.BufferedInputStream;
|
---|
| 4 | import java.io.FileInputStream;
|
---|
| 5 | import java.io.FileNotFoundException;
|
---|
| 6 | import java.io.IOException;
|
---|
| 7 | import java.io.Serializable;
|
---|
| 8 | import org.apache.commons.cli.*;
|
---|
[31361] | 9 | import org.apache.hadoop.io.Text;
|
---|
[31294] | 10 | import org.apache.spark.api.java.*;
|
---|
| 11 | import org.apache.spark.api.java.function.Function2;
|
---|
| 12 | import org.apache.spark.api.java.function.PairFunction;
|
---|
| 13 | import org.apache.spark.storage.StorageLevel;
|
---|
| 14 | import org.apache.spark.util.DoubleAccumulator;
|
---|
| 15 | import scala.Tuple2;
|
---|
| 16 |
|
---|
| 17 | import org.apache.spark.SparkConf;
|
---|
| 18 |
|
---|
| 19 | public class ProcessForCatalogLangCount implements Serializable
|
---|
| 20 | {
|
---|
| 21 | private static final long serialVersionUID = 1L;
|
---|
| 22 |
|
---|
| 23 | protected static final int DEFAULT_FILES_PER_PARTITION = 3000;
|
---|
| 24 |
|
---|
| 25 | protected String _input_dir;
|
---|
| 26 | protected String _json_list_filename;
|
---|
| 27 |
|
---|
| 28 | protected int _verbosity;
|
---|
| 29 |
|
---|
| 30 | public ProcessForCatalogLangCount(String input_dir, String json_list_filename, int verbosity)
|
---|
| 31 | {
|
---|
| 32 | _input_dir = input_dir;
|
---|
| 33 | _json_list_filename = (json_list_filename != null) ? json_list_filename : input_dir;
|
---|
| 34 |
|
---|
| 35 | _verbosity = verbosity;
|
---|
| 36 | }
|
---|
| 37 |
|
---|
| 38 | protected String generateSparkAppName(String exec_mode)
|
---|
| 39 | {
|
---|
| 40 | String spark_app_name = "[" + exec_mode + "] Extracted Features: Process for Catalog Language Labels";
|
---|
| 41 | spark_app_name += " [" + _json_list_filename + "]";
|
---|
| 42 |
|
---|
| 43 | return spark_app_name;
|
---|
| 44 | }
|
---|
| 45 |
|
---|
[31359] | 46 | public void execCatalogLangCountSparkDirect()
|
---|
[31294] | 47 | {
|
---|
[31359] | 48 | String spark_app_name = generateSparkAppName("Spark-Direct + Per Volume");
|
---|
[31294] | 49 |
|
---|
| 50 | SparkConf conf = new SparkConf().setAppName(spark_app_name);
|
---|
| 51 | JavaSparkContext jsc = new JavaSparkContext(conf);
|
---|
[31359] | 52 |
|
---|
[31294] | 53 | String filename_root = _json_list_filename.replaceAll(".*/","").replaceAll("\\..*$","");
|
---|
| 54 | String output_directory = "catalog-lang-" + filename_root + "-out";
|
---|
| 55 | if (ClusterFileIO.exists(output_directory))
|
---|
| 56 | {
|
---|
| 57 | System.err.println("Error: " + output_directory + " already exists. Spark unable to write output data");
|
---|
| 58 | jsc.close();
|
---|
| 59 | System.exit(1);
|
---|
| 60 | }
|
---|
| 61 |
|
---|
| 62 | //int num_partitions = Integer.getInteger("wcsa-ef-ingest.num-partitions", DEFAULT_NUM_PARTITIONS);
|
---|
| 63 | int files_per_partition = Integer.getInteger("wcsa-ef-ingest.files-per-partition", DEFAULT_FILES_PER_PARTITION);
|
---|
| 64 |
|
---|
| 65 | JavaRDD<String> json_list_data = jsc.textFile(_json_list_filename).cache();
|
---|
| 66 | json_list_data.setName("JSON-file-list");
|
---|
| 67 |
|
---|
| 68 | long num_volumes = json_list_data.count();
|
---|
| 69 | double per_vol = 100.0/(double)num_volumes;
|
---|
| 70 |
|
---|
| 71 | int num_partitions = (int)(num_volumes/files_per_partition)+1;
|
---|
| 72 |
|
---|
| 73 | JavaRDD<String> json_list_data_rp = json_list_data.repartition(num_partitions);
|
---|
| 74 | json_list_data_rp.setName("JSON-file-list--repartitioned");
|
---|
| 75 |
|
---|
| 76 | DoubleAccumulator per_vol_progress_accum = jsc.sc().doubleAccumulator("Per Volume Progress Percent");
|
---|
| 77 |
|
---|
| 78 | boolean strict_file_io = Boolean.getBoolean("wcsa-ef-ingest.strict-file-io");
|
---|
| 79 |
|
---|
| 80 | PerVolumeCatalogLangStreamFlatmap volume_catalog_langfreq_flatmap
|
---|
| 81 | = new PerVolumeCatalogLangStreamFlatmap(_input_dir,_verbosity,
|
---|
| 82 | per_vol_progress_accum,per_vol,
|
---|
| 83 | strict_file_io);
|
---|
| 84 | JavaRDD<String> catalog_lang_list = json_list_data_rp.flatMap(volume_catalog_langfreq_flatmap);
|
---|
| 85 | catalog_lang_list.persist(StorageLevel.MEMORY_AND_DISK());
|
---|
| 86 | catalog_lang_list.setName("catalog-lang-stream");
|
---|
| 87 |
|
---|
| 88 |
|
---|
| 89 | JavaPairRDD<String, Long> catalog_lang_pairs = catalog_lang_list.mapToPair(s -> new Tuple2<String, Long>(s, 1L));
|
---|
| 90 | catalog_lang_pairs.setName("single-catalog-lang-count");
|
---|
| 91 |
|
---|
| 92 | JavaPairRDD<String, Long> catalog_lang_counts = catalog_lang_pairs.reduceByKey((a, b) -> a + b);
|
---|
| 93 | catalog_lang_counts.setName("catalog-lang-frequency");
|
---|
| 94 |
|
---|
| 95 | JavaPairRDD<Long, String> catalog_lang_counts_swapped_pair
|
---|
| 96 | = catalog_lang_counts.mapToPair(item -> item.swap());
|
---|
| 97 | catalog_lang_counts_swapped_pair.setName("frequency-catalog-lang-swap");
|
---|
| 98 |
|
---|
| 99 | JavaPairRDD<Long, String> catalog_lang_counts_swapped_pair_sorted
|
---|
| 100 | = catalog_lang_counts_swapped_pair.sortByKey(false, num_partitions);
|
---|
| 101 | catalog_lang_counts_swapped_pair_sorted.setName("descending-sorted-frequency-catalog-lang");
|
---|
| 102 |
|
---|
| 103 | JavaPairRDD<String, Long> catalog_lang_count_sorted
|
---|
| 104 | = catalog_lang_counts_swapped_pair_sorted.mapToPair(item -> item.swap());
|
---|
| 105 | catalog_lang_count_sorted.setName("descending-catalog-lang-frequency");
|
---|
| 106 |
|
---|
| 107 |
|
---|
| 108 | catalog_lang_count_sorted.saveAsTextFile(output_directory);
|
---|
| 109 | jsc.close();
|
---|
| 110 | }
|
---|
| 111 |
|
---|
[31359] | 112 | public void execCatalogLangCount()
|
---|
| 113 | {
|
---|
| 114 |
|
---|
| 115 |
|
---|
| 116 | String spark_app_name = generateSparkAppName("YARN Cluster + Per Volume");
|
---|
| 117 |
|
---|
| 118 | SparkConf conf = new SparkConf().setAppName(spark_app_name);
|
---|
| 119 | JavaSparkContext jsc = new JavaSparkContext(conf);
|
---|
| 120 | jsc.hadoopConfiguration().set("io.compression.codec.bzip2.library", "java-builtin");
|
---|
| 121 |
|
---|
| 122 | String packed_sequence_path = "hdfs:///user/capitanu/data/packed-ef";
|
---|
| 123 |
|
---|
[31361] | 124 | JavaPairRDD<String, Text> inputRdd = jsc.sequenceFile(packed_sequence_path, String.class, Text.class);
|
---|
| 125 | JavaRDD<Text> jsonTextRdd = inputRdd.map(Tuple2::_2);
|
---|
[31359] | 126 |
|
---|
| 127 | /*
|
---|
| 128 | jsonTextRdd.map(
|
---|
| 129 | jsonText -> "" // parse JSON text and do whatever ...
|
---|
| 130 | );
|
---|
| 131 | */
|
---|
| 132 |
|
---|
| 133 | boolean strict_file_io = Boolean.getBoolean("wcsa-ef-ingest.strict-file-io");
|
---|
| 134 |
|
---|
| 135 | PerVolumeCatalogLangSequenceFileMap volume_catalog_langfreq_map
|
---|
| 136 | = new PerVolumeCatalogLangSequenceFileMap(_input_dir,_verbosity,strict_file_io);
|
---|
| 137 | JavaRDD<String> catalog_lang_list = jsonTextRdd.map(volume_catalog_langfreq_map);
|
---|
| 138 | //catalog_lang_list.persist(StorageLevel.MEMORY_AND_DISK());
|
---|
| 139 | catalog_lang_list.setName("catalog-lang-stream");
|
---|
[31294] | 140 |
|
---|
[31359] | 141 |
|
---|
| 142 | JavaPairRDD<String, Long> catalog_lang_pairs = catalog_lang_list.mapToPair(s -> new Tuple2<String, Long>(s, 1L));
|
---|
| 143 | catalog_lang_pairs.setName("single-catalog-lang-count");
|
---|
| 144 |
|
---|
| 145 | JavaPairRDD<String, Long> catalog_lang_counts = catalog_lang_pairs.reduceByKey((a, b) -> a + b);
|
---|
| 146 | catalog_lang_counts.setName("catalog-lang-frequency");
|
---|
| 147 |
|
---|
| 148 | JavaPairRDD<Long, String> catalog_lang_counts_swapped_pair
|
---|
| 149 | = catalog_lang_counts.mapToPair(item -> item.swap());
|
---|
| 150 | catalog_lang_counts_swapped_pair.setName("frequency-catalog-lang-swap");
|
---|
| 151 |
|
---|
| 152 | JavaPairRDD<Long, String> catalog_lang_counts_swapped_pair_sorted
|
---|
| 153 | = catalog_lang_counts_swapped_pair.sortByKey(false);
|
---|
| 154 | catalog_lang_counts_swapped_pair_sorted.setName("descending-sorted-frequency-catalog-lang");
|
---|
| 155 |
|
---|
| 156 | JavaPairRDD<String, Long> catalog_lang_count_sorted
|
---|
| 157 | = catalog_lang_counts_swapped_pair_sorted.mapToPair(item -> item.swap());
|
---|
| 158 | catalog_lang_count_sorted.setName("descending-catalog-lang-frequency");
|
---|
| 159 |
|
---|
| 160 | String filename_root = _json_list_filename.replaceAll(".*/","").replaceAll("\\..*$","");
|
---|
| 161 | String output_directory = "catalog-lang-" + filename_root + "-out";
|
---|
| 162 | catalog_lang_count_sorted.saveAsTextFile(output_directory);
|
---|
| 163 | jsc.close();
|
---|
| 164 | }
|
---|
| 165 |
|
---|
| 166 |
|
---|
| 167 |
|
---|
[31294] | 168 | public static void print_usage(HelpFormatter formatter, Options options)
|
---|
| 169 | {
|
---|
| 170 | formatter.printHelp("RUN.bash [options] input-dir json-filelist.txt", options);
|
---|
| 171 | }
|
---|
| 172 |
|
---|
| 173 | public static void main(String[] args) {
|
---|
| 174 | Options options = new Options();
|
---|
| 175 |
|
---|
| 176 | Option verbosity_opt = new Option("v", "verbosity", true,
|
---|
| 177 | "Set to control the level of debugging output [0=none, 1=some, 2=lots]");
|
---|
| 178 | verbosity_opt.setRequired(false);
|
---|
| 179 | options.addOption(verbosity_opt);
|
---|
| 180 |
|
---|
| 181 | Option properties_opt = new Option("p", "properties", true,
|
---|
| 182 | "Read in the specified Java properties file");
|
---|
| 183 | properties_opt.setRequired(false);
|
---|
| 184 | options.addOption(properties_opt);
|
---|
| 185 |
|
---|
| 186 | // Need to work with CLI v1.2 as this is the JAR that is bundled with Hadoop/Spark
|
---|
| 187 | CommandLineParser parser = new GnuParser();
|
---|
| 188 | //CommandLineParser parser = new DefaultParser(); // if working with CLI v1.3 and above
|
---|
| 189 |
|
---|
| 190 | HelpFormatter formatter = new HelpFormatter();
|
---|
| 191 | CommandLine cmd = null;
|
---|
| 192 |
|
---|
| 193 | try {
|
---|
| 194 | cmd = parser.parse(options, args);
|
---|
| 195 | }
|
---|
| 196 | catch (ParseException e) {
|
---|
| 197 | System.err.println(e.getMessage());
|
---|
| 198 | print_usage(formatter,options);
|
---|
| 199 | System.exit(1);
|
---|
| 200 | }
|
---|
| 201 |
|
---|
| 202 | String verbosity_str = cmd.getOptionValue("verbosity","1");
|
---|
| 203 | int verbosity = Integer.parseInt(verbosity_str);
|
---|
| 204 |
|
---|
| 205 | String property_filename = cmd.getOptionValue("properties",null);
|
---|
| 206 |
|
---|
| 207 | String[] filtered_args = cmd.getArgs();
|
---|
| 208 |
|
---|
| 209 | if (filtered_args.length != 2) {
|
---|
| 210 | print_usage(formatter,options);
|
---|
| 211 | System.exit(1);
|
---|
| 212 | }
|
---|
| 213 |
|
---|
| 214 | if (property_filename != null) {
|
---|
| 215 | try {
|
---|
| 216 | FileInputStream fis = new FileInputStream(property_filename);
|
---|
| 217 | BufferedInputStream bis = new BufferedInputStream(fis);
|
---|
| 218 |
|
---|
| 219 | System.getProperties().load(bis);
|
---|
| 220 | }
|
---|
| 221 | catch (FileNotFoundException e) {
|
---|
| 222 | e.printStackTrace();
|
---|
| 223 | System.err.println("File not found: '" + property_filename + "'. Skipping property file read");
|
---|
| 224 | }
|
---|
| 225 | catch (IOException e) {
|
---|
| 226 | System.err.println("IO Exception for: '" + property_filename + "'. Malformed syntax? Skipping property file read");
|
---|
| 227 | }
|
---|
| 228 | }
|
---|
| 229 |
|
---|
| 230 | String input_dir = filtered_args[0];
|
---|
| 231 | String json_list_filename = filtered_args[1];
|
---|
| 232 |
|
---|
| 233 | ProcessForCatalogLangCount prep_for_lang
|
---|
| 234 | = new ProcessForCatalogLangCount(input_dir,json_list_filename,verbosity);
|
---|
| 235 |
|
---|
| 236 | prep_for_lang.execCatalogLangCount();
|
---|
| 237 |
|
---|
| 238 | }
|
---|
| 239 | }
|
---|