source: other-projects/hathitrust/wcsa/extracted-features-solr/trunk/solr-ingest/src/main/java/org/hathitrust/extractedfeatures/ProcessForSolrIngest.java@ 31277

Last change on this file since 31277 was 31277, checked in by davidb, 7 years ago

Tweak to minimum value
  • Property svn:executable set to *
File size: 11.4 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 protected static final int DEFAULT_NUM_CORES = 10;
31 protected static final int MINIMUM_NUM_PARTITIONS = 10*DEFAULT_NUM_CORES;
32
33 protected static final int DEFAULT_FILES_PER_PARTITION = 3000;
34
35 protected String _input_dir;
36 protected String _json_list_filename;
37 protected String _whitelist_filename;
38 protected String _solr_url;
39 protected String _output_dir;
40
41 protected int _verbosity;
42
43 public ProcessForSolrIngest(String input_dir, String json_list_filename,
44 String solr_url, String output_dir, int verbosity)
45 {
46 _input_dir = input_dir;
47 _json_list_filename = (json_list_filename != null) ? json_list_filename : input_dir;
48
49 boolean use_whitelist = Boolean.getBoolean("wcsa-ef-ingest.use-whitelist");
50 _whitelist_filename = (use_whitelist) ? System.getProperty("wcsa-ef-ingest.whitelist-filename") : null;
51
52 _solr_url = solr_url;
53 _output_dir = output_dir;
54 _verbosity = verbosity;
55 }
56
57 protected String generateSparkAppName(String exec_mode)
58 {
59 String spark_app_name = "[" + exec_mode + "] Extract Features: Process for Solr Ingest";
60 spark_app_name += " [" + _json_list_filename + "]";
61
62 if (_solr_url != null) {
63 spark_app_name += " solr_url=" + _solr_url;
64 }
65
66 if (_output_dir != null) {
67 spark_app_name += " output_dir=" + _output_dir;
68 }
69
70 return spark_app_name;
71 }
72
73 public ArrayList<String> extrapolateSolrEndpoints()
74 {
75 ArrayList<String> solr_endpoints = new ArrayList<String>();
76
77 if (_solr_url != null) {
78 String solr_cloud_nodes = System.getProperty("wcsa-ef-ingest.solr-cloud-nodes",null);
79 if (solr_cloud_nodes != null) {
80 String [] cloud_nodes = solr_cloud_nodes.split(",");
81 for (String cn : cloud_nodes) {
82 String solr_endpoint = _solr_url.replaceFirst("//.*?:\\d+/", "//"+cn+"/");
83 solr_endpoints.add(solr_endpoint);
84 }
85 }
86 else {
87 solr_endpoints.add(_solr_url);
88 }
89 }
90
91 return solr_endpoints;
92 }
93
94 public void execPerVolume()
95 {
96 String spark_app_name = generateSparkAppName("Per Volume");
97
98 SparkConf conf = new SparkConf().setAppName(spark_app_name);
99 JavaSparkContext jsc = new JavaSparkContext(conf);
100
101 //int num_partitions = Integer.getInteger("wcsa-ef-ingest.num-partitions", DEFAULT_NUM_PARTITIONS);
102 int files_per_partition = Integer.getInteger("wcsa-ef-ingest.files-per-partition", DEFAULT_FILES_PER_PARTITION);
103
104 JavaRDD<String> json_list_data = jsc.textFile(_json_list_filename).cache();
105
106 long num_volumes = json_list_data.count();
107 double per_vol = 100.0/(double)num_volumes;
108
109 int num_partitions = (int)(num_volumes/files_per_partition)+1;
110 if (num_partitions < MINIMUM_NUM_PARTITIONS) {
111 num_partitions = MINIMUM_NUM_PARTITIONS;
112 }
113 JavaRDD<String> json_list_data_rp = json_list_data.repartition(num_partitions);
114
115 DoubleAccumulator progress_accum = jsc.sc().doubleAccumulator("Progress Percent");
116
117 boolean icu_tokenize = Boolean.getBoolean("wcsa-ef-ingest.icu-tokenize");
118 boolean strict_file_io = Boolean.getBoolean("wcsa-ef-ingest.strict-file-io");
119
120 PerVolumeJSON per_vol_json = new PerVolumeJSON(_input_dir,_whitelist_filename,
121 _solr_url,_output_dir,_verbosity, progress_accum,per_vol,
122 icu_tokenize,strict_file_io);
123
124 //json_list_data_rp.foreach(per_vol_json);
125 JavaRDD<String> per_page_ids = json_list_data_rp.flatMap(per_vol_json);
126 long num_page_ids = per_page_ids.count(); // trigger lazy eval of: flatmap:per-vol
127
128 //long num_ids = num_volumes;
129
130 System.out.println("");
131 System.out.println("############");
132 System.out.println("# Number of page ids: " + num_page_ids);
133 System.out.println("############");
134 System.out.println("");
135
136 jsc.close();
137 }
138
139
140
141 public void execPerPage()
142 {
143 String spark_app_name = generateSparkAppName("Per Page");
144
145 SparkConf conf = new SparkConf().setAppName(spark_app_name);
146 JavaSparkContext jsc = new JavaSparkContext(conf);
147
148 /*
149 if (_verbosity >= 2) {
150 System.out.println("Default Minimum Partions: " + jsc.defaultMinPartitions());
151 System.out.println("Default Parallelism: " + jsc.defaultParallelism());
152 }
153 */
154
155 //int num_partitions = Integer.getInteger("wcsa-ef-ingest.num-partitions", DEFAULT_NUM_PARTITIONS);
156 int files_per_partition = Integer.getInteger("wcsa-ef-ingest.num-partitions", DEFAULT_FILES_PER_PARTITION);
157
158 JavaRDD<String> json_list_data = jsc.textFile(_json_list_filename).cache();
159
160 long num_volumes = json_list_data.count();
161 double per_vol = 100.0/(double)num_volumes;
162
163 int num_partitions = (int)(num_volumes/files_per_partition)+1;
164 JavaRDD<String> json_list_data_rp = json_list_data.repartition(num_partitions);
165
166 DoubleAccumulator per_vol_progress_accum = jsc.sc().doubleAccumulator("Per Volume Progress Percent");
167
168 boolean icu_tokenize = Boolean.getBoolean("wcsa-ef-ingest.icu-tokenize");
169 boolean strict_file_io = Boolean.getBoolean("wcsa-ef-ingest.strict-file-io");
170
171 PerPageJSONFlatmap paged_solr_json_flatmap
172 = new PerPageJSONFlatmap(_input_dir,_whitelist_filename,
173 _solr_url,_output_dir,_verbosity,
174 per_vol_progress_accum,per_vol,
175 icu_tokenize,strict_file_io);
176 //JavaRDD<JSONObject> per_page_jsonobjects = json_list_data_rp.flatMap(paged_solr_json_flatmap).cache();
177 JavaRDD<JSONObject> per_page_jsonobjects = json_list_data_rp.flatMap(paged_solr_json_flatmap);
178
179 //long num_page_ids = per_page_jsonobjects.count(); // trigger lazy eval of: flatmap:per-vol
180
181 LongAccumulator per_page_progress_accum = jsc.sc().longAccumulator("Pages Processed");
182 ArrayList<String> solr_endpoints = extrapolateSolrEndpoints();
183
184
185 PerPageJSONMap paged_json_id_map
186 = new PerPageJSONMap(_input_dir,solr_endpoints,_output_dir,_verbosity,
187 per_page_progress_accum,1);
188 JavaRDD<String> per_page_ids = per_page_jsonobjects.map(paged_json_id_map);
189
190/*
191 System.out.println("");
192 System.out.println("############");
193 System.out.println("# Progress Accumulator: " + progress_accum.value());
194 System.out.println("############");
195 System.out.println("");
196*/
197
198 long num_page_ids = per_page_ids.count(); // trigger lazy eval of: flatmap:per-vol -> map:per-page
199
200 System.out.println("");
201 System.out.println("############");
202 System.out.println("# Number of page ids: " + num_page_ids);
203 System.out.println("############");
204 System.out.println("");
205
206 /*
207 if (_output_dir != null) {
208 String rdd_save_file = "rdd-solr-json-page-files";
209 json_ids.saveAsTextFile(rdd_save_file);
210 System.out.println("############");
211 System.out.println("# Saved RDD of Solr JSON page files, top-level, as:");
212 System.out.println("# " + rdd_save_file);
213 System.out.println("############");
214 System.out.println("");
215 }
216 */
217
218 jsc.close();
219 }
220
221
222
223
224 public static void print_usage(HelpFormatter formatter, Options options)
225 {
226 formatter.printHelp("RUN.bash [options] input-dir json-filelist.txt", options);
227 }
228
229 public static void main(String[] args) {
230 Options options = new Options();
231
232 Option verbosity_opt = new Option("v", "verbosity", true,
233 "Set to control the level of debugging output [0=none, 1=some, 2=lots]");
234 verbosity_opt.setRequired(false);
235 options.addOption(verbosity_opt);
236
237 Option properties_opt = new Option("p", "properties", true,
238 "Read in the specified Java properties file");
239 properties_opt.setRequired(false);
240 options.addOption(properties_opt);
241
242 Option output_dir_opt = new Option("o", "output-dir", true,
243 "If specified, save BZipped Solr JSON files to this directory");
244 output_dir_opt.setRequired(false);
245 options.addOption(output_dir_opt);
246
247 Option solr_url_opt = new Option("u", "solr-url", true,
248 "If specified, the URL to post the Solr JSON data to");
249 solr_url_opt.setRequired(false);
250 options.addOption(solr_url_opt);
251
252 Option read_only_opt = new Option("r", "read-only", false,
253 "Used to initiate a run where the files are all read in, but nothing is ingested/saved");
254 read_only_opt.setRequired(false);
255 options.addOption(read_only_opt);
256
257 // Need to work with CLI v1.2 as this is the JAR that is bundled with Hadoop/Spark
258 CommandLineParser parser = new GnuParser();
259 //CommandLineParser parser = new DefaultParser(); // if working with CLI v1.3 and above
260
261 HelpFormatter formatter = new HelpFormatter();
262 CommandLine cmd = null;
263
264 try {
265 cmd = parser.parse(options, args);
266 }
267 catch (ParseException e) {
268 System.err.println(e.getMessage());
269 print_usage(formatter,options);
270 System.exit(1);
271 }
272
273
274 String verbosity_str = cmd.getOptionValue("verbosity","1");
275 int verbosity = Integer.parseInt(verbosity_str);
276
277 String property_filename = cmd.getOptionValue("properties",null);
278
279 String output_dir = cmd.getOptionValue("output-dir",null);
280 String solr_url = cmd.getOptionValue("solr-url",null);
281 boolean read_only = cmd.hasOption("read-only");
282
283 String[] filtered_args = cmd.getArgs();
284
285 if (filtered_args.length != 2) {
286 print_usage(formatter,options);
287 System.exit(1);
288 }
289
290 if (property_filename != null) {
291 try {
292 FileInputStream fis = new FileInputStream(property_filename);
293 BufferedInputStream bis = new BufferedInputStream(fis);
294
295 System.getProperties().load(bis);
296 }
297 catch (FileNotFoundException e) {
298 // TODO Auto-generated catch block
299 e.printStackTrace();
300 System.err.println("File not found: '" + property_filename + "'. Skipping property file read");
301 }
302 catch (IOException e) {
303 System.err.println("IO Exception for: '" + property_filename + "'. Malformed syntax? Skipping property file read");
304 }
305 }
306
307 if (!read_only && ((output_dir == null) && (solr_url==null))) {
308 System.err.println("Need to specify either --solr-url or --output-dir otherwise generated files are not ingested/saved");
309 print_usage(formatter,options);
310 System.exit(1);
311 }
312 if (read_only) {
313 // For this case, need to ensure solr-url and output-dir are null
314 output_dir = null;
315 solr_url = null;
316 }
317
318 String input_dir = filtered_args[0];
319 String json_list_filename = filtered_args[1];
320
321 ProcessForSolrIngest prep_for_ingest
322 = new ProcessForSolrIngest(input_dir,json_list_filename,solr_url,output_dir,verbosity);
323
324 String process_ef_json_mode = System.getProperty("wcsa-ef-ingest.process-ef-json-mode","per-page");
325 if (process_ef_json_mode.equals("per-volume")) {
326 prep_for_ingest.execPerVolume();
327 }
328 else {
329 prep_for_ingest.execPerPage();
330 }
331 }
332}
Note: See TracBrowser for help on using the repository browser.