source: other-projects/hathitrust/wcsa/extracted-features-solr/trunk/solr-ingest/src/main/java/org/hathitrust/extractedfeatures/ProcessForSolrIngest.java@ 31045

Last change on this file since 31045 was 31045, checked in by davidb, 7 years ago

More careful treatment of what to do when a JSON file isn't there
  • Property svn:executable set to *
File size: 10.2 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 java.util.ArrayList;
9
10import org.apache.commons.cli.*;
11
12import org.apache.spark.api.java.*;
13import org.apache.spark.util.DoubleAccumulator;
14import org.apache.spark.util.LongAccumulator;
15import org.hathitrust.extractedfeatures.PerPageJSONFlatmap;
16import org.json.JSONObject;
17import org.apache.spark.SparkConf;
18
19public class ProcessForSolrIngest implements Serializable
20{
21 private static final long serialVersionUID = 1L;
22
23 // Following details on number of partitions to use given in
24 // "Parallelized collections" section of:
25 // https://spark.apache.org/docs/2.0.1/programming-guide.html
26 //
27 // For a more detailed discussion see:
28 // http://blog.cloudera.com/blog/2015/03/how-to-tune-your-apache-spark-jobs-part-2/
29
30 public static final int NUM_CORES = 6;
31 public static final int NUM_PARTITIONS = 2*NUM_CORES; // default would appear to be 2
32
33 protected String _input_dir;
34 protected String _json_list_filename;
35 protected String _solr_url;
36 protected String _output_dir;
37
38 protected int _verbosity;
39
40 public ProcessForSolrIngest(String input_dir, String json_list_filename,
41 String solr_url, String output_dir, int verbosity)
42 {
43 _input_dir = input_dir;
44 _json_list_filename = (json_list_filename != null) ? json_list_filename : input_dir;
45
46 _solr_url = solr_url;
47 _output_dir = output_dir;
48 _verbosity = verbosity;
49 }
50
51 protected String generateSparkAppName(String exec_mode)
52 {
53 String spark_app_name = "[" + exec_mode + "] Extract Features: Process for Solr Ingest";
54 spark_app_name += " [" + _json_list_filename + "]";
55
56 if (_solr_url != null) {
57 spark_app_name += " solr_url=" + _solr_url;
58 }
59
60 if (_output_dir != null) {
61 spark_app_name += " output_dir=" + _output_dir;
62 }
63
64 return spark_app_name;
65 }
66
67 public ArrayList<String> extrapolateSolrEndpoints()
68 {
69 ArrayList<String> solr_endpoints = new ArrayList<String>();
70
71 if (_solr_url != null) {
72 String solr_endpoint_ips = System.getProperty("wcsa-ef-ingest.solr-endpoint-ips",null);
73 if (solr_endpoint_ips != null) {
74 String [] ips = solr_endpoint_ips.split(",");
75 for (String ip : ips) {
76 String solr_endpoint = _solr_url.replaceFirst("//.*?:", "//"+ip+":");
77 solr_endpoints.add(solr_endpoint);
78 }
79 }
80 else {
81 solr_endpoints.add(_solr_url);
82 }
83 }
84
85 return solr_endpoints;
86 }
87
88 public void execPerVolume()
89 {
90 String spark_app_name = generateSparkAppName("Per Volume");
91
92 SparkConf conf = new SparkConf().setAppName(spark_app_name);
93 JavaSparkContext jsc = new JavaSparkContext(conf);
94
95 if (_verbosity >= 2) {
96 System.out.println("Default Minimum Partions: " + jsc.defaultMinPartitions());
97 System.out.println("Default Parallelism: " + jsc.defaultParallelism());
98 }
99
100 JavaRDD<String> json_list_data = jsc.textFile(_json_list_filename,NUM_PARTITIONS).cache();
101
102 long num_volumes = json_list_data.count();
103 double per_vol = 100.0/(double)num_volumes;
104
105 DoubleAccumulator progress_accum = jsc.sc().doubleAccumulator("Progress Percent");
106
107 System.err.println();
108 System.err.println();
109 System.err.println();
110 System.err.println("****##### _input_dir = " + _input_dir);
111 System.err.println();
112 System.err.println();
113 System.err.println();
114
115 PerVolumeJSON per_vol_json = new PerVolumeJSON(_input_dir,_solr_url,_output_dir,_verbosity, progress_accum,per_vol);
116
117 json_list_data.foreach(per_vol_json);
118
119 long num_ids = num_volumes;
120
121 System.out.println("");
122 System.out.println("############");
123 System.out.println("# Number of volume ids: " + num_ids);
124 System.out.println("############");
125 System.out.println("");
126
127 jsc.close();
128 }
129
130
131
132 public void execPerPage()
133 {
134 String spark_app_name = generateSparkAppName("Per Page");
135
136 SparkConf conf = new SparkConf().setAppName(spark_app_name);
137 JavaSparkContext jsc = new JavaSparkContext(conf);
138
139 if (_verbosity >= 2) {
140 System.out.println("Default Minimum Partions: " + jsc.defaultMinPartitions());
141 System.out.println("Default Parallelism: " + jsc.defaultParallelism());
142 }
143
144 JavaRDD<String> json_list_data = jsc.textFile(_json_list_filename,NUM_PARTITIONS).cache();
145
146 long num_volumes = json_list_data.count();
147 double per_vol = 100.0/(double)num_volumes;
148
149 DoubleAccumulator per_vol_progress_accum = jsc.sc().doubleAccumulator("Per Volume Progress Percent");
150
151 //String strict_file_io_str = System.getProperty("wcsa-ef-ingest.strict-file-io","true");
152 boolean strict_file_io = Boolean.getBoolean("wcsa-ef-ingest.strict-file-io");
153
154 PerPageJSONFlatmap paged_solr_json_flatmap
155 = new PerPageJSONFlatmap(_input_dir,_solr_url,_output_dir,_verbosity,
156 per_vol_progress_accum,per_vol,
157 strict_file_io);
158 JavaRDD<JSONObject> per_page_jsonobjects = json_list_data.flatMap(paged_solr_json_flatmap).cache();
159
160 //long num_page_ids = per_page_jsonobjects.count(); // trigger lazy eval of: flatmap:per-vol
161
162 LongAccumulator per_page_progress_accum = jsc.sc().longAccumulator("Pages Processed");
163 ArrayList<String> solr_endpoints = extrapolateSolrEndpoints();
164
165
166 PerPageJSONMap paged_json_id_map
167 = new PerPageJSONMap(_input_dir,solr_endpoints,_output_dir,_verbosity,
168 per_page_progress_accum,1);
169 JavaRDD<String> per_page_ids = per_page_jsonobjects.map(paged_json_id_map);
170
171/*
172 System.out.println("");
173 System.out.println("############");
174 System.out.println("# Progress Accumulator: " + progress_accum.value());
175 System.out.println("############");
176 System.out.println("");
177*/
178
179 long num_page_ids = per_page_ids.count(); // trigger lazy eval of: flatmap:per-vol -> map:per-page
180
181 System.out.println("");
182 System.out.println("############");
183 System.out.println("# Number of page ids: " + num_page_ids);
184 System.out.println("############");
185 System.out.println("");
186
187 /*
188 if (_output_dir != null) {
189 String rdd_save_file = "rdd-solr-json-page-files";
190 json_ids.saveAsTextFile(rdd_save_file);
191 System.out.println("############");
192 System.out.println("# Saved RDD of Solr JSON page files, top-level, as:");
193 System.out.println("# " + rdd_save_file);
194 System.out.println("############");
195 System.out.println("");
196 }
197 */
198
199 jsc.close();
200 }
201
202
203
204
205 public static void print_usage(HelpFormatter formatter, Options options)
206 {
207 formatter.printHelp("RUN.bash [options] input-dir json-filelist.txt", options);
208 }
209
210 public static void main(String[] args) {
211 Options options = new Options();
212
213 Option verbosity_opt = new Option("v", "verbosity", true,
214 "Set to control the level of debugging output [0=none, 1=some, 2=lots]");
215 verbosity_opt.setRequired(false);
216 options.addOption(verbosity_opt);
217
218 Option properties_opt = new Option("p", "properties", true,
219 "Read in the specified Java properties file");
220 properties_opt.setRequired(false);
221 options.addOption(properties_opt);
222
223 Option output_dir_opt = new Option("o", "output-dir", true,
224 "If specified, save BZipped Solr JSON files to this directory");
225 output_dir_opt.setRequired(false);
226 options.addOption(output_dir_opt);
227
228 Option solr_url_opt = new Option("u", "solr-url", true,
229 "If specified, the URL to post the Solr JSON data to");
230 solr_url_opt.setRequired(false);
231 options.addOption(solr_url_opt);
232
233 Option read_only_opt = new Option("r", "read-only", false,
234 "Used to initiate a run where the files are all read in, but nothing is ingested/saved");
235 read_only_opt.setRequired(false);
236 options.addOption(read_only_opt);
237
238 // Need to work with CLI v1.2 as this is the JAR that is bundled with Hadoop/Spark
239 CommandLineParser parser = new GnuParser();
240 //CommandLineParser parser = new DefaultParser(); // if working with CLI v1.3 and above
241
242 HelpFormatter formatter = new HelpFormatter();
243 CommandLine cmd = null;
244
245 try {
246 cmd = parser.parse(options, args);
247 }
248 catch (ParseException e) {
249 System.err.println(e.getMessage());
250 print_usage(formatter,options);
251 System.exit(1);
252 }
253
254
255 String verbosity_str = cmd.getOptionValue("verbosity","1");
256 int verbosity = Integer.parseInt(verbosity_str);
257
258 String property_filename = cmd.getOptionValue("properties",null);
259
260 String output_dir = cmd.getOptionValue("output-dir",null);
261 String solr_url = cmd.getOptionValue("solr-url",null);
262 boolean read_only = cmd.hasOption("read-only");
263
264 String[] filtered_args = cmd.getArgs();
265
266 if (filtered_args.length != 2) {
267 print_usage(formatter,options);
268 System.exit(1);
269 }
270
271 if (property_filename != null) {
272 try {
273 FileInputStream fis = new FileInputStream(property_filename);
274 BufferedInputStream bis = new BufferedInputStream(fis);
275
276 System.getProperties().load(bis);
277 }
278 catch (FileNotFoundException e) {
279 // TODO Auto-generated catch block
280 e.printStackTrace();
281 System.err.println("File not found: '" + property_filename + "'. Skipping property file read");
282 }
283 catch (IOException e) {
284 System.err.println("IO Exception for: '" + property_filename + "'. Malformed syntax? Skipping property file read");
285 }
286 }
287
288 if (!read_only && ((output_dir == null) && (solr_url==null))) {
289 System.err.println("Need to specify either --solr-url or --output-dir otherwise generated files are not ingested/saved");
290 print_usage(formatter,options);
291 System.exit(1);
292 }
293 if (read_only) {
294 // For this case, need to ensure solr-url and output-dir are null
295 output_dir = null;
296 solr_url = null;
297 }
298
299 String input_dir = filtered_args[0];
300 String json_list_filename = filtered_args[1];
301
302 ProcessForSolrIngest prep_for_ingest
303 = new ProcessForSolrIngest(input_dir,json_list_filename,solr_url,output_dir,verbosity);
304
305 String process_ef_json_mode = System.getProperty("wcsa-ef-ingest.process-ef-json-mode","per-page");
306 if (process_ef_json_mode.equals("per-volume")) {
307 prep_for_ingest.execPerVolume();
308 }
309 else {
310 prep_for_ingest.execPerPage();
311 }
312 }
313}
Note: See TracBrowser for help on using the repository browser.