source: other-projects/hathitrust/wcsa/extracted-features-solr/trunk/solr-ingest/src/main/java/org/hathitrust/extractedfeatures/ProcessForLangCount.java@ 31270

package org.hathitrust.extractedfeatures;

import java.io.BufferedInputStream;
import java.io.FileInputStream;
import java.io.FileNotFoundException;
import java.io.IOException;
import java.io.Serializable;
import org.apache.commons.cli.*;

import org.apache.spark.api.java.*;
import org.apache.spark.api.java.function.Function2;
import org.apache.spark.api.java.function.PairFunction;
import org.apache.spark.util.DoubleAccumulator;
import scala.Tuple2;

import org.apache.spark.SparkConf;

public class ProcessForLangCount implements Serializable
{
    private static final long serialVersionUID = 1L;

    // Following details on number of partitions to use given in 
    //  "Parallelized collections" section of:
    //   https://spark.apache.org/docs/2.0.1/programming-guide.html
    //
    // For a more detailed discussion see:
    //   http://blog.cloudera.com/blog/2015/03/how-to-tune-your-apache-spark-jobs-part-2/
    
    //protected static final int DEFAULT_NUM_CORES = 6;
    //protected static final int DEFAULT_NUM_PARTITIONS = 3*DEFAULT_NUM_CORES; 
    protected static final int DEFAULT_FILES_PER_PARTITION = 3000;
    
    protected String _input_dir;
    protected String _json_list_filename;
    
    protected int    _verbosity;

    public ProcessForLangCount(String input_dir, String json_list_filename, int verbosity)
    {
        _input_dir = input_dir;
        _json_list_filename = (json_list_filename != null) ? json_list_filename : input_dir;

        _verbosity  = verbosity;
    }

    protected String generateSparkAppName(String exec_mode)
    {
        String spark_app_name = "[" + exec_mode + "] Extracted Features: Process for Language Labels";
        spark_app_name += " [" + _json_list_filename + "]";
    
        return spark_app_name;
    }
        
    public void execLangCount()
    {   
        String spark_app_name = generateSparkAppName("Per Volume");     
        
        SparkConf conf = new SparkConf().setAppName(spark_app_name);
        JavaSparkContext jsc = new JavaSparkContext(conf);

        String filename_root = _json_list_filename.replaceAll(".*/","").replaceAll("\\..*$","");
        String output_directory = "lang-" + filename_root + "-out";
        if (ClusterFileIO.exists(output_directory))
        {
            System.err.println("Error: " + output_directory + " already exists.  Spark unable to write output data");
            jsc.close();
            System.exit(1);
        }
        
        //int num_partitions = Integer.getInteger("wcsa-ef-ingest.num-partitions", DEFAULT_NUM_PARTITIONS);
        int files_per_partition = Integer.getInteger("wcsa-ef-ingest.num-partitions", DEFAULT_FILES_PER_PARTITION);
        
        
        JavaRDD<String> json_list_data = jsc.textFile(_json_list_filename,files_per_partition).cache();
        json_list_data.setName("JSON-file-list");
        
        long num_volumes = json_list_data.count();
        double per_vol = 100.0/(double)num_volumes;

        int num_partitions = (int)(num_volumes/files_per_partition)+1;
        
        JavaRDD<String> json_list_data_rp = json_list_data.repartition(num_partitions);
        
        DoubleAccumulator per_vol_progress_accum = jsc.sc().doubleAccumulator("Per Volume Progress Percent");
        
        boolean strict_file_io = Boolean.getBoolean("wcsa-ef-ingest.strict-file-io");
        
        PerVolumeLangStreamFlatmap paged_solr_langfreq_flatmap 
            = new PerVolumeLangStreamFlatmap(_input_dir,_verbosity, 
                                     per_vol_progress_accum,per_vol,
                                     strict_file_io);
        JavaRDD<String> lang_list = json_list_data_rp.flatMap(paged_solr_langfreq_flatmap); 
        lang_list.setName("lang-stream");
    
        
        JavaPairRDD<String, Long> lang_pairs = lang_list.mapToPair(s -> new Tuple2<String, Long>(s, 1L));
        lang_pairs.setName("single-lang-count");
        
        JavaPairRDD<String, Long> lang_counts = lang_pairs.reduceByKey((a, b) -> a + b);
        lang_counts.setName("lang-frequency");
        
        JavaPairRDD<Long, String> lang_counts_swapped_pair
            = lang_counts.mapToPair(item -> item.swap());
        lang_counts_swapped_pair.setName("frequency-lang-swap");
        
        JavaPairRDD<Long, String> lang_counts_swapped_pair_sorted 
            = lang_counts_swapped_pair.sortByKey(false, num_partitions);
        lang_counts_swapped_pair_sorted.setName("descending-sorted-frequency-lang");
        
        JavaPairRDD<String, Long> lang_count_sorted 
            = lang_counts_swapped_pair_sorted.mapToPair(item -> item.swap());
        lang_count_sorted.setName("descending-lang-frequency");
        
        
        lang_count_sorted.saveAsTextFile(output_directory);
        jsc.close();
    }

    
    public static void print_usage(HelpFormatter formatter, Options options)
    {
        formatter.printHelp("RUN.bash [options] input-dir json-filelist.txt", options);
    }
    
    public static void main(String[] args) {
        Options options = new Options();

        Option verbosity_opt = new Option("v", "verbosity", true, 
                "Set to control the level of debugging output [0=none, 1=some, 2=lots]");
        verbosity_opt.setRequired(false);
        options.addOption(verbosity_opt);
        
        Option properties_opt = new Option("p", "properties", true, 
                "Read in the specified Java properties file");
        properties_opt.setRequired(false);
        options.addOption(properties_opt);
        
        // Need to work with CLI v1.2 as this is the JAR that is bundled with Hadoop/Spark 
        CommandLineParser parser = new GnuParser(); 
        //CommandLineParser parser = new DefaultParser(); // if working with CLI v1.3 and above
        
        HelpFormatter formatter = new HelpFormatter();
        CommandLine cmd = null;

        try {
            cmd = parser.parse(options, args);
        }
        catch (ParseException e) {
            System.err.println(e.getMessage());
            print_usage(formatter,options);
            System.exit(1);
        }

        String verbosity_str = cmd.getOptionValue("verbosity","1");
        int verbosity = Integer.parseInt(verbosity_str);

        String property_filename = cmd.getOptionValue("properties",null);
        
        String[] filtered_args = cmd.getArgs();

        if (filtered_args.length != 2) {
            print_usage(formatter,options);
            System.exit(1);
        }
    
        if (property_filename != null) {
            try {
                FileInputStream fis = new FileInputStream(property_filename);
                BufferedInputStream bis = new BufferedInputStream(fis);
                
                System.getProperties().load(bis);
            } 
            catch (FileNotFoundException e) {
                e.printStackTrace();
                System.err.println("File not found: '" + property_filename + "'. Skipping property file read");
            }
            catch (IOException e) {
                System.err.println("IO Exception for: '" + property_filename + "'. Malformed syntax? Skipping property file read");
            }
        }
                
        String input_dir  = filtered_args[0];
        String json_list_filename = filtered_args[1];
        
        ProcessForLangCount prep_for_lang 
            = new ProcessForLangCount(input_dir,json_list_filename,verbosity);
        
        prep_for_lang.execLangCount();
    
    }
}