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 PerVolumeWordStreamFlatmap 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 PerVolumeWordStreamFlatmap(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> all_word_list = new ArrayList<String>();
|
---|
44 |
|
---|
45 | if (extracted_feature_record != null) {
|
---|
46 | String volume_id = extracted_feature_record.getString("id");
|
---|
47 |
|
---|
48 | JSONObject ef_features = extracted_feature_record.getJSONObject("features");
|
---|
49 |
|
---|
50 | int ef_page_count = ef_features.getInt("pageCount");
|
---|
51 |
|
---|
52 | if (_verbosity >= 1) {
|
---|
53 | System.out.println("Processing: " + json_file_in);
|
---|
54 | System.out.println(" pageCount = " + ef_page_count);
|
---|
55 | }
|
---|
56 |
|
---|
57 | JSONArray ef_pages = ef_features.getJSONArray("pages");
|
---|
58 | int ef_num_pages = ef_pages.length();
|
---|
59 | if (ef_num_pages != ef_page_count) {
|
---|
60 | System.err.println("Warning: number of page elements in JSON (" + ef_num_pages + ")"
|
---|
61 | +" does not match 'pageCount' metadata (" + ef_page_count + ")");
|
---|
62 | }
|
---|
63 |
|
---|
64 | if (_verbosity >= 2) {
|
---|
65 | System.out.print(" Pages: ");
|
---|
66 | }
|
---|
67 |
|
---|
68 | for (int i = 0; i < ef_page_count; i++) {
|
---|
69 | String formatted_i = String.format("page-%06d", i);
|
---|
70 | String page_id = volume_id + "." + formatted_i;
|
---|
71 |
|
---|
72 | if (_verbosity >= 2) {
|
---|
73 | if (i>0) {
|
---|
74 | System.out.print(", ");
|
---|
75 | }
|
---|
76 | System.out.print(page_id);
|
---|
77 | }
|
---|
78 |
|
---|
79 | if (i==(ef_page_count-1)) {
|
---|
80 | if (_verbosity >= 2) {
|
---|
81 | System.out.println();
|
---|
82 | }
|
---|
83 | }
|
---|
84 |
|
---|
85 | JSONObject ef_page = ef_pages.getJSONObject(i);
|
---|
86 |
|
---|
87 | if (ef_page != null) {
|
---|
88 |
|
---|
89 | ArrayList<String> page_word_list = SolrDocJSON.generateTokenPosCountText(volume_id, page_id, ef_page);
|
---|
90 | all_word_list.addAll(page_word_list);
|
---|
91 | }
|
---|
92 | else {
|
---|
93 | System.err.println("Skipping: " + page_id);
|
---|
94 | }
|
---|
95 | }
|
---|
96 | }
|
---|
97 | else {
|
---|
98 | // File did not exist, or could not be parsed
|
---|
99 | String mess = "Failed to read in bzipped JSON file '" + full_json_file_in + "'";
|
---|
100 | if (_strict_file_io) {
|
---|
101 | throw new IOException(mess);
|
---|
102 | }
|
---|
103 | else {
|
---|
104 | System.err.println("Warning: " + mess);
|
---|
105 | System.out.println("Warning: " + mess);
|
---|
106 | }
|
---|
107 | }
|
---|
108 |
|
---|
109 | _progress_accum.add(_progress_step);
|
---|
110 |
|
---|
111 | return all_word_list.iterator();
|
---|
112 | }
|
---|
113 |
|
---|
114 |
|
---|
115 | }
|
---|
116 |
|
---|