| 1 | package org.hathitrust.extractedfeatures;
|
|---|
| 2 |
|
|---|
| 3 | import java.io.IOException;
|
|---|
| 4 | import java.util.ArrayList;
|
|---|
| 5 | import java.util.Iterator;
|
|---|
| 6 |
|
|---|
| 7 | import org.apache.spark.api.java.function.FlatMapFunction;
|
|---|
| 8 | import org.apache.spark.util.DoubleAccumulator;
|
|---|
| 9 | import org.json.JSONArray;
|
|---|
| 10 | import org.json.JSONObject;
|
|---|
| 11 |
|
|---|
| 12 | class PerVolumeCatalogLangStreamFlatmap implements FlatMapFunction<String, String>
|
|---|
| 13 | {
|
|---|
| 14 | private static final long serialVersionUID = 1L;
|
|---|
| 15 |
|
|---|
| 16 | protected String _input_dir;
|
|---|
| 17 | protected int _verbosity;
|
|---|
| 18 |
|
|---|
| 19 | protected DoubleAccumulator _progress_accum;
|
|---|
| 20 | protected double _progress_step;
|
|---|
| 21 |
|
|---|
| 22 | boolean _strict_file_io;
|
|---|
| 23 |
|
|---|
| 24 | public PerVolumeCatalogLangStreamFlatmap(String input_dir, int verbosity,
|
|---|
| 25 | DoubleAccumulator progress_accum, double progress_step,
|
|---|
| 26 | boolean strict_file_io)
|
|---|
| 27 | {
|
|---|
| 28 | _input_dir = input_dir;
|
|---|
| 29 | _verbosity = verbosity;
|
|---|
| 30 |
|
|---|
| 31 | _progress_accum = progress_accum;
|
|---|
| 32 | _progress_step = progress_step;
|
|---|
| 33 |
|
|---|
| 34 | _strict_file_io = strict_file_io;
|
|---|
| 35 | }
|
|---|
| 36 |
|
|---|
| 37 | public Iterator<String> call(String json_file_in) throws IOException
|
|---|
| 38 | {
|
|---|
| 39 |
|
|---|
| 40 | String full_json_file_in = _input_dir + "/" + json_file_in;
|
|---|
| 41 | JSONObject extracted_feature_record = JSONClusterFileIO.readJSONFile(full_json_file_in);
|
|---|
| 42 |
|
|---|
| 43 | ArrayList<String> catalog_lang_list = new ArrayList<String>();
|
|---|
| 44 |
|
|---|
| 45 | if (extracted_feature_record != null) {
|
|---|
| 46 | String volume_id = extracted_feature_record.getString("id");
|
|---|
| 47 |
|
|---|
| 48 | JSONObject ef_metadata = extracted_feature_record.getJSONObject("metadata");
|
|---|
| 49 |
|
|---|
| 50 | if (_verbosity >= 1) {
|
|---|
| 51 | System.out.println("Processing: " + json_file_in);
|
|---|
| 52 | }
|
|---|
| 53 |
|
|---|
| 54 | if (ef_metadata != null) {
|
|---|
| 55 | String ef_catalog_language = ef_metadata.getString("language");
|
|---|
| 56 | if (!ef_catalog_language.equals("")) {
|
|---|
| 57 |
|
|---|
| 58 | catalog_lang_list.add(ef_catalog_language);
|
|---|
| 59 | }
|
|---|
| 60 | else {
|
|---|
| 61 | System.err.println("No catalog 'language' metadata => Skipping id: " + volume_id);
|
|---|
| 62 | }
|
|---|
| 63 | }
|
|---|
| 64 | else {
|
|---|
| 65 | System.err.println("No 'metadata' section in JSON file => Skipping id: " + volume_id);
|
|---|
| 66 | }
|
|---|
| 67 |
|
|---|
| 68 | }
|
|---|
| 69 | else {
|
|---|
| 70 | // File did not exist, or could not be parsed
|
|---|
| 71 | String mess = "Failed to read in bzipped JSON file '" + full_json_file_in + "'";
|
|---|
| 72 | if (_strict_file_io) {
|
|---|
| 73 | throw new IOException(mess);
|
|---|
| 74 | }
|
|---|
| 75 | else {
|
|---|
| 76 | System.err.println("Warning: " + mess);
|
|---|
| 77 | System.out.println("Warning: " + mess);
|
|---|
| 78 | }
|
|---|
| 79 | }
|
|---|
| 80 |
|
|---|
| 81 | _progress_accum.add(_progress_step);
|
|---|
| 82 |
|
|---|
| 83 | return catalog_lang_list.iterator();
|
|---|
| 84 | }
|
|---|
| 85 |
|
|---|
| 86 |
|
|---|
| 87 | }
|
|---|
| 88 |
|
|---|
