source: other-projects/hathitrust/wcsa/extracted-features-solr/trunk/solr-ingest/src/main/java/org/hathitrust/extractedfeatures/ProcessForSolrIngest.java@ 31597

Last change on this file since 31597 was 31597, checked in by davidb, 7 years ago

Additional _s and _ss fields to help with faceting. Temporarily commented out the full-text page part.
  • Property svn:executable set to *
File size: 13.2 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.*;
11import org.apache.hadoop.io.Text;
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 protected static final int DEFAULT_NUM_CORES = 10;
24 protected static final int MINIMUM_NUM_PARTITIONS = 10*DEFAULT_NUM_CORES;
25
26 protected static final int DEFAULT_FILES_PER_PARTITION = 3000;
27
28 protected String _input_dir;
29 protected String _solr_base_url;
30 protected String _solr_collection;
31
32 protected String _whitelist_filename;
33 protected String _langmap_directory;
34
35 //protected String _solr_url;
36 protected String _output_dir;
37
38 protected int _verbosity;
39
40 public ProcessForSolrIngest(String input_dir, String solr_collection,
41 String solr_base_url, String output_dir, int verbosity)
42 {
43 _input_dir = input_dir;
44 _solr_collection = solr_collection;
45
46 boolean use_whitelist = Boolean.getBoolean("wcsa-ef-ingest.use-whitelist");
47 _whitelist_filename = (use_whitelist) ? System.getProperty("wcsa-ef-ingest.whitelist-filename") : null;
48
49 boolean use_langmap = Boolean.getBoolean("wcsa-ef-ingest.use-langmap");
50 _langmap_directory = (use_langmap) ? System.getProperty("wcsa-ef-ingest.langmap-directory") : null;
51
52
53 _solr_base_url = solr_base_url;
54 _output_dir = output_dir;
55 _verbosity = verbosity;
56 }
57
58 protected String generateSparkAppName(String exec_mode)
59 {
60 String spark_app_name = "[" + exec_mode + "] Extract Features: Process for Solr Ingest";
61 spark_app_name += " [" + _solr_collection + "]";
62
63 if (_solr_base_url != null) {
64 spark_app_name += " solr_base_url=" + _solr_base_url;
65 }
66
67 if (_output_dir != null) {
68 spark_app_name += " output_dir=" + _output_dir;
69 }
70
71 return spark_app_name;
72 }
73
74 public ArrayList<String> extrapolateSolrEndpoints(String solr_collection)
75 {
76 ArrayList<String> solr_endpoints = new ArrayList<String>();
77
78 if (_solr_base_url != null) {
79 String solr_url = _solr_base_url + "/" + solr_collection + "/update";
80
81 String solr_cloud_nodes = System.getProperty("wcsa-ef-ingest.solr-cloud-nodes",null);
82 if (solr_cloud_nodes != null) {
83 String [] cloud_nodes = solr_cloud_nodes.split(",");
84 for (String cn : cloud_nodes) {
85 String solr_endpoint = solr_url.replaceFirst("//.*?:\\d+/", "//"+cn+"/");
86 solr_endpoints.add(solr_endpoint);
87 }
88 }
89 else {
90 solr_endpoints.add(solr_url);
91 }
92 }
93
94 return solr_endpoints;
95 }
96
97 public void execPerVolumeSequenceFile()
98 {
99 String spark_app_name = generateSparkAppName("Per Volume");
100
101 SparkConf conf = new SparkConf().setAppName(spark_app_name);
102 JavaSparkContext jsc = new JavaSparkContext(conf);
103 jsc.hadoopConfiguration().set("io.compression.codec.bzip2.library", "java-builtin");
104
105 //String packed_sequence_path = "hdfs:///user/capitanu/data/packed-ef";
106 String packed_sequence_path = _input_dir;
107
108 JavaPairRDD<Text, Text> input_pair_rdd = jsc.sequenceFile(packed_sequence_path, Text.class, Text.class);
109 //JavaPairRDD<String, String> input_pair_rdd = jsc.wholeTextFiles(packed_sequence_path);
110
111 //JavaPairRDD<Text, Text> input_pair_sampled_rdd = input_pair_rdd.sample(false,0.5,42);
112
113 //JavaRDD<Text> json_text_rdd = input_pair_sampled_rdd.map(item -> item._2);
114 //JavaRDD<Text> json_text_rdd = input_pair_rdd.map(item -> new Text(item._2));
115 JavaRDD<Text> json_text_rdd = input_pair_rdd.map(item -> item._2);
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 ArrayList<String> solr_endpoints = extrapolateSolrEndpoints(_solr_collection);
121
122 System.out.println("*** away to create PerVolumeJSON class, _langmap_directory = " + _langmap_directory);
123 PerVolumeJSON per_vol_json = new PerVolumeJSON(_input_dir,_whitelist_filename, _langmap_directory,
124 solr_endpoints,_output_dir,_verbosity,
125 icu_tokenize,strict_file_io);
126
127 JavaRDD<Integer> per_volume_page_count = json_text_rdd.map(per_vol_json);
128
129 //Integer num_page_ids = per_volume_page_count.reduce((a, b) -> a + b);
130 long num_vol_ids = per_volume_page_count.count();
131
132 System.out.println("");
133 System.out.println("############");
134 //System.out.println("# Number of page ids: " + num_page_ids);
135 System.out.println("# Number of volume ids: " + num_vol_ids);
136 System.out.println("############");
137 System.out.println("");
138
139 jsc.close();
140
141 }
142
143 /*
144 public void execPerVolume()
145 {
146 String spark_app_name = generateSparkAppName("Per Volume");
147
148 SparkConf conf = new SparkConf().setAppName(spark_app_name);
149 JavaSparkContext jsc = new JavaSparkContext(conf);
150
151 //int num_partitions = Integer.getInteger("wcsa-ef-ingest.num-partitions", DEFAULT_NUM_PARTITIONS);
152 int files_per_partition = Integer.getInteger("wcsa-ef-ingest.files-per-partition", DEFAULT_FILES_PER_PARTITION);
153
154 JavaRDD<String> json_list_data = jsc.textFile(_json_list_filename).cache();
155
156 long num_volumes = json_list_data.count();
157 double per_vol = 100.0/(double)num_volumes;
158
159 int num_partitions = (int)(num_volumes/files_per_partition)+1;
160 if (num_partitions < MINIMUM_NUM_PARTITIONS) {
161 num_partitions = MINIMUM_NUM_PARTITIONS;
162 }
163 JavaRDD<String> json_list_data_rp = json_list_data.repartition(num_partitions);
164
165 DoubleAccumulator progress_accum = jsc.sc().doubleAccumulator("Progress Percent");
166
167 boolean icu_tokenize = Boolean.getBoolean("wcsa-ef-ingest.icu-tokenize");
168 boolean strict_file_io = Boolean.getBoolean("wcsa-ef-ingest.strict-file-io");
169
170 PerVolumeJSON per_vol_json = new PerVolumeJSON(_input_dir,_whitelist_filename,
171 _solr_url,_output_dir,_verbosity, progress_accum,per_vol,
172 icu_tokenize,strict_file_io);
173
174 //json_list_data_rp.foreach(per_vol_json);
175 JavaRDD<String> per_page_ids = json_list_data_rp.flatMap(per_vol_json);
176 long num_page_ids = per_page_ids.count(); // trigger lazy eval of: flatmap:per-vol
177
178 //long num_ids = num_volumes;
179
180 System.out.println("");
181 System.out.println("############");
182 System.out.println("# Number of page ids: " + num_page_ids);
183 System.out.println("############");
184 System.out.println("");
185
186 jsc.close();
187 }
188 */
189
190 /*
191 public void execPerPage()
192 {
193 String spark_app_name = generateSparkAppName("Per Page");
194
195 SparkConf conf = new SparkConf().setAppName(spark_app_name);
196 JavaSparkContext jsc = new JavaSparkContext(conf);
197
198
199
200 //int num_partitions = Integer.getInteger("wcsa-ef-ingest.num-partitions", DEFAULT_NUM_PARTITIONS);
201 int files_per_partition = Integer.getInteger("wcsa-ef-ingest.num-partitions", DEFAULT_FILES_PER_PARTITION);
202
203 JavaRDD<String> json_list_data = jsc.textFile(_json_list_filename).cache();
204
205 long num_volumes = json_list_data.count();
206 double per_vol = 100.0/(double)num_volumes;
207
208 int num_partitions = (int)(num_volumes/files_per_partition)+1;
209 JavaRDD<String> json_list_data_rp = json_list_data.repartition(num_partitions);
210
211 DoubleAccumulator per_vol_progress_accum = jsc.sc().doubleAccumulator("Per Volume Progress Percent");
212
213 boolean icu_tokenize = Boolean.getBoolean("wcsa-ef-ingest.icu-tokenize");
214 boolean strict_file_io = Boolean.getBoolean("wcsa-ef-ingest.strict-file-io");
215
216 PerPageJSONFlatmap paged_solr_json_flatmap
217 = new PerPageJSONFlatmap(_input_dir,_whitelist_filename,
218 _solr_url,_output_dir,_verbosity,
219 per_vol_progress_accum,per_vol,
220 icu_tokenize,strict_file_io);
221 //JavaRDD<JSONObject> per_page_jsonobjects = json_list_data_rp.flatMap(paged_solr_json_flatmap).cache();
222 JavaRDD<JSONObject> per_page_jsonobjects = json_list_data_rp.flatMap(paged_solr_json_flatmap);
223
224 //long num_page_ids = per_page_jsonobjects.count(); // trigger lazy eval of: flatmap:per-vol
225
226 LongAccumulator per_page_progress_accum = jsc.sc().longAccumulator("Pages Processed");
227 ArrayList<String> solr_endpoints = extrapolateSolrEndpoints();
228
229
230 PerPageJSONMap paged_json_id_map
231 = new PerPageJSONMap(_input_dir,solr_endpoints,_output_dir,_verbosity,
232 per_page_progress_accum,1);
233 JavaRDD<String> per_page_ids = per_page_jsonobjects.map(paged_json_id_map);
234
235
236 long num_page_ids = per_page_ids.count(); // trigger lazy eval of: flatmap:per-vol -> map:per-page
237
238 System.out.println("");
239 System.out.println("############");
240 System.out.println("# Number of page ids: " + num_page_ids);
241 System.out.println("############");
242 System.out.println("");
243
244
245 //if (_output_dir != null) {
246 //String rdd_save_file = "rdd-solr-json-page-files";
247 //json_ids.saveAsTextFile(rdd_save_file);
248 //System.out.println("############");
249 //System.out.println("# Saved RDD of Solr JSON page files, top-level, as:");
250 //System.out.println("# " + rdd_save_file);
251 //System.out.println("############");
252 //System.out.println("");
253 //}
254
255
256 jsc.close();
257 }
258*/
259
260
261
262 public static void print_usage(HelpFormatter formatter, Options options)
263 {
264 formatter.printHelp("RUN.bash [options] input-dir solr-collection", options);
265 }
266
267 public static void main(String[] args) {
268 Options options = new Options();
269
270 Option verbosity_opt = new Option("v", "verbosity", true,
271 "Set to control the level of debugging output [0=none, 1=some, 2=lots]");
272 verbosity_opt.setRequired(false);
273 options.addOption(verbosity_opt);
274
275 Option properties_opt = new Option("p", "properties", true,
276 "Read in the specified Java properties file");
277 properties_opt.setRequired(false);
278 options.addOption(properties_opt);
279
280 Option output_dir_opt = new Option("o", "output-dir", true,
281 "If specified, save BZipped Solr JSON files to this directory");
282 output_dir_opt.setRequired(false);
283 options.addOption(output_dir_opt);
284
285 Option solr_base_url_opt = new Option("u", "solr-base-url", true,
286 "If specified, the base URL to post the Solr JSON data to");
287 solr_base_url_opt.setRequired(false);
288 options.addOption(solr_base_url_opt);
289
290 Option read_only_opt = new Option("r", "read-only", false,
291 "Used to initiate a run where the files are all read in, but nothing is ingested/saved");
292 read_only_opt.setRequired(false);
293 options.addOption(read_only_opt);
294
295 // Need to work with CLI v1.2 as this is the JAR that is bundled with Hadoop/Spark
296 CommandLineParser parser = new GnuParser();
297 //CommandLineParser parser = new DefaultParser(); // if working with CLI v1.3 and above
298
299 HelpFormatter formatter = new HelpFormatter();
300 CommandLine cmd = null;
301
302 try {
303 cmd = parser.parse(options, args);
304 }
305 catch (ParseException e) {
306 System.err.println(e.getMessage());
307 print_usage(formatter,options);
308 System.exit(1);
309 }
310
311
312 String verbosity_str = cmd.getOptionValue("verbosity","1");
313 int verbosity = Integer.parseInt(verbosity_str);
314
315 String property_filename = cmd.getOptionValue("properties",null);
316
317 String output_dir = cmd.getOptionValue("output-dir",null);
318 String solr_base_url = cmd.getOptionValue("solr-base-url",null);
319 boolean read_only = cmd.hasOption("read-only");
320
321 String[] filtered_args = cmd.getArgs();
322
323 if (filtered_args.length != 2) {
324 print_usage(formatter,options);
325 System.exit(1);
326 }
327
328 if (property_filename != null) {
329 try {
330 FileInputStream fis = new FileInputStream(property_filename);
331 BufferedInputStream bis = new BufferedInputStream(fis);
332
333 System.getProperties().load(bis);
334 }
335 catch (FileNotFoundException e) {
336 // TODO Auto-generated catch block
337 e.printStackTrace();
338 System.err.println("File not found: '" + property_filename + "'. Skipping property file read");
339 }
340 catch (IOException e) {
341 System.err.println("IO Exception for: '" + property_filename + "'. Malformed syntax? Skipping property file read");
342 }
343 }
344
345 if (!read_only && ((output_dir == null) && (solr_base_url==null))) {
346 System.err.println("Need to specify either --solr-base-url or --output-dir otherwise generated files are not ingested/saved");
347 print_usage(formatter,options);
348 System.exit(1);
349 }
350 if (read_only) {
351 // For this case, need to ensure solr-url and output-dir are null
352 output_dir = null;
353 solr_base_url = null;
354 }
355
356 String input_dir = filtered_args[0];
357 String solr_collection = filtered_args[1];
358
359 ProcessForSolrIngest prep_for_ingest
360 = new ProcessForSolrIngest(input_dir,solr_collection,solr_base_url,output_dir,verbosity);
361
362 prep_for_ingest.execPerVolumeSequenceFile();
363
364 /*
365 String process_ef_json_mode = System.getProperty("wcsa-ef-ingest.process-ef-json-mode","per-page");
366 if (process_ef_json_mode.equals("per-volume")) {
367 prep_for_ingest.execPerVolume();
368 }
369 else {
370 prep_for_ingest.execPerPage();
371 }*/
372 }
373}
Note: See TracBrowser for help on using the repository browser.