source: other-projects/hathitrust/wcsa/extracted-features-solr/trunk/solr-ingest/src/main/java/org/hathitrust/extractedfeatures/ProcessForSolrIngest.java@ 31026

Last change on this file since 31026 was 31026, checked in by davidb, 7 years ago

Corrected flag setting
  • Property svn:executable set to *
File size: 9.2 KB
Line 
1package org.hathitrust.extractedfeatures;
2
3import java.io.BufferedInputStream;
4import java.io.FileInputStream;
5import java.io.FileNotFoundException;
6import java.io.IOException;
7import java.io.Serializable;
8import org.apache.commons.cli.*;
9
10import org.apache.spark.api.java.*;
11import org.apache.spark.util.DoubleAccumulator;
12import org.hathitrust.extractedfeatures.PerPageJSONFlatmap;
13import org.json.JSONObject;
14import org.apache.spark.SparkConf;
15
16public class ProcessForSolrIngest implements Serializable
17{
18 private static final long serialVersionUID = 1L;
19
20 // Following details on number of partitions to use given in
21 // "Parallelized collections" section of:
22 // https://spark.apache.org/docs/2.0.1/programming-guide.html
23 //
24 // For a more detailed discussion see:
25 // http://blog.cloudera.com/blog/2015/03/how-to-tune-your-apache-spark-jobs-part-2/
26
27 public static final int NUM_CORES = 6;
28 public static final int NUM_PARTITIONS = 2*NUM_CORES; // default would appear to be 2
29
30 protected String _input_dir;
31 protected String _json_list_filename;
32 protected String _solr_url;
33 protected String _output_dir;
34
35 protected int _verbosity;
36
37 public ProcessForSolrIngest(String input_dir, String json_list_filename,
38 String solr_url, String output_dir, int verbosity)
39 {
40 _input_dir = input_dir;
41 _json_list_filename = (json_list_filename != null) ? json_list_filename : input_dir;
42
43 _solr_url = solr_url;
44 _output_dir = output_dir;
45 _verbosity = verbosity;
46 }
47
48 protected String generateSparkAppName(String exec_mode)
49 {
50 String spark_app_name = "[" + exec_mode + "] Extract Features: Process for Solr Ingest";
51 spark_app_name += " [" + _json_list_filename + "]";
52
53 if (_solr_url != null) {
54 spark_app_name += " solr_url=" + _solr_url;
55 }
56
57 if (_output_dir != null) {
58 spark_app_name += " output_dir=" + _output_dir;
59 }
60
61 return spark_app_name;
62 }
63
64 public void execPerVolume()
65 {
66 String spark_app_name = generateSparkAppName("Per Volume");
67
68 SparkConf conf = new SparkConf().setAppName(spark_app_name);
69 JavaSparkContext jsc = new JavaSparkContext(conf);
70
71 if (_verbosity >= 2) {
72 System.out.println("Default Minimum Partions: " + jsc.defaultMinPartitions());
73 System.out.println("Default Parallelism: " + jsc.defaultParallelism());
74 }
75
76 JavaRDD<String> json_list_data = jsc.textFile(_json_list_filename,NUM_PARTITIONS).cache();
77
78 long num_volumes = json_list_data.count();
79 double per_vol = 100.0/(double)num_volumes;
80
81 DoubleAccumulator progress_accum = jsc.sc().doubleAccumulator("Progress Percent");
82
83 System.err.println();
84 System.err.println();
85 System.err.println();
86 System.err.println("****##### _input_dir = " + _input_dir);
87 System.err.println();
88 System.err.println();
89 System.err.println();
90
91 PerVolumeJSON per_vol_json = new PerVolumeJSON(_input_dir,_solr_url,_output_dir,_verbosity, progress_accum,per_vol);
92
93 json_list_data.foreach(per_vol_json);
94
95 long num_ids = num_volumes;
96
97 System.out.println("");
98 System.out.println("############");
99 System.out.println("# Number of volume ids: " + num_ids);
100 System.out.println("############");
101 System.out.println("");
102
103 jsc.close();
104 }
105
106 public void execPerPage()
107 {
108 String spark_app_name = generateSparkAppName("Per Page");
109
110 SparkConf conf = new SparkConf().setAppName(spark_app_name);
111 JavaSparkContext jsc = new JavaSparkContext(conf);
112
113 if (_verbosity >= 2) {
114 System.out.println("Default Minimum Partions: " + jsc.defaultMinPartitions());
115 System.out.println("Default Parallelism: " + jsc.defaultParallelism());
116 }
117
118 JavaRDD<String> json_list_data = jsc.textFile(_json_list_filename,NUM_PARTITIONS).cache();
119
120 long num_volumes = json_list_data.count();
121 double per_vol = 100.0/(double)num_volumes;
122
123 DoubleAccumulator per_vol_progress_accum = jsc.sc().doubleAccumulator("Per Volume Progress Percent");
124
125 PerPageJSONFlatmap paged_solr_json_flatmap
126 = new PerPageJSONFlatmap(_input_dir,_solr_url,_output_dir,_verbosity, per_vol_progress_accum,per_vol);
127 JavaRDD<JSONObject> per_page_jsonobjects = json_list_data.flatMap(paged_solr_json_flatmap).cache();
128
129 //long num_page_ids = per_page_jsonobjects.count(); // trigger lazy eval of: flatmap:per-vol
130
131 DoubleAccumulator per_page_progress_accum = jsc.sc().doubleAccumulator("Pages Processed");
132
133 PerPageJSONMap paged_json_id_map
134 = new PerPageJSONMap(_input_dir,_solr_url,_output_dir,_verbosity, per_page_progress_accum,1.0);
135 JavaRDD<String> per_page_ids = per_page_jsonobjects.map(paged_json_id_map);
136
137/*
138 System.out.println("");
139 System.out.println("############");
140 System.out.println("# Progress Accumulator: " + progress_accum.value());
141 System.out.println("############");
142 System.out.println("");
143*/
144
145 long num_page_ids = per_page_ids.count(); // trigger lazy eval of: flatmap:per-vol -> map:per-page
146
147 System.out.println("");
148 System.out.println("############");
149 System.out.println("# Number of page ids: " + num_page_ids);
150 System.out.println("############");
151 System.out.println("");
152
153 /*
154 if (_output_dir != null) {
155 String rdd_save_file = "rdd-solr-json-page-files";
156 json_ids.saveAsTextFile(rdd_save_file);
157 System.out.println("############");
158 System.out.println("# Saved RDD of Solr JSON page files, top-level, as:");
159 System.out.println("# " + rdd_save_file);
160 System.out.println("############");
161 System.out.println("");
162 }
163 */
164
165 jsc.close();
166 }
167
168
169
170
171 public static void print_usage(HelpFormatter formatter, Options options)
172 {
173 formatter.printHelp("RUN.bash [options] input-dir json-filelist.txt", options);
174 }
175
176 public static void main(String[] args) {
177 Options options = new Options();
178
179 Option verbosity_opt = new Option("v", "verbosity", true,
180 "Set to control the level of debugging output [0=none, 1=some, 2=lots]");
181 verbosity_opt.setRequired(false);
182 options.addOption(verbosity_opt);
183
184 Option properties_opt = new Option("p", "properties", true,
185 "Read in the specified Java properties file");
186 properties_opt.setRequired(false);
187 options.addOption(properties_opt);
188
189 Option output_dir_opt = new Option("o", "output-dir", true,
190 "If specified, save BZipped Solr JSON files to this directory");
191 output_dir_opt.setRequired(false);
192 options.addOption(output_dir_opt);
193
194 Option solr_url_opt = new Option("u", "solr-url", true,
195 "If specified, the URL to post the Solr JSON data to");
196 solr_url_opt.setRequired(false);
197 options.addOption(solr_url_opt);
198
199 Option read_only_opt = new Option("r", "read-only", false,
200 "Used to initiate a run where the files are all read in, but nothing is ingested/saved");
201 read_only_opt.setRequired(false);
202 options.addOption(read_only_opt);
203
204 // Need to work with CLI v1.2 as this is the JAR that is bundled with Hadoop/Spark
205 CommandLineParser parser = new GnuParser();
206 //CommandLineParser parser = new DefaultParser(); // if working with CLI v1.3 and above
207
208 HelpFormatter formatter = new HelpFormatter();
209 CommandLine cmd = null;
210
211 try {
212 cmd = parser.parse(options, args);
213 }
214 catch (ParseException e) {
215 System.err.println(e.getMessage());
216 print_usage(formatter,options);
217 System.exit(1);
218 }
219
220
221 String verbosity_str = cmd.getOptionValue("verbosity","0");
222 int verbosity = Integer.parseInt(verbosity_str);
223
224 String property_filename = cmd.getOptionValue("properties",null);
225
226 String output_dir = cmd.getOptionValue("output-dir",null);
227 String solr_url = cmd.getOptionValue("solr-url",null);
228 boolean read_only = cmd.hasOption("read-only");
229
230 String[] filtered_args = cmd.getArgs();
231
232 if (filtered_args.length != 2) {
233 print_usage(formatter,options);
234 System.exit(1);
235 }
236
237 if (property_filename != null) {
238 try {
239 FileInputStream fis = new FileInputStream(property_filename);
240 BufferedInputStream bis = new BufferedInputStream(fis);
241
242 System.getProperties().load(bis);
243 }
244 catch (FileNotFoundException e) {
245 // TODO Auto-generated catch block
246 e.printStackTrace();
247 System.err.println("File not found: '" + property_filename + "'. Skipping property file read");
248 }
249 catch (IOException e) {
250 System.err.println("IO Exception for: '" + property_filename + "'. Malformed syntax? Skipping property file read");
251 }
252 }
253
254 if (!read_only && ((output_dir == null) && (solr_url==null))) {
255 System.err.println("Need to specify either --solr-url or --output-dir otherwise generated files are not ingested/saved");
256 print_usage(formatter,options);
257 System.exit(1);
258 }
259 if (read_only) {
260 // For this case, need to ensure solr-url and output-dir are null
261 output_dir = null;
262 solr_url = null;
263 }
264
265 String input_dir = filtered_args[0];
266 String json_list_filename = filtered_args[1];
267
268 ProcessForSolrIngest prep_for_ingest
269 = new ProcessForSolrIngest(input_dir,json_list_filename,solr_url,output_dir,verbosity);
270
271 String process_json_mode = System.getProperty("solr-ingest.process-json-mode","per-page");
272 if (process_json_mode.equals("per-volume")) {
273 prep_for_ingest.execPerVolume();
274 }
275 else {
276 prep_for_ingest.execPerPage();
277 }
278 }
279}
Note: See TracBrowser for help on using the repository browser.