1 | package org.greenstone.atea;
|
---|
2 |
|
---|
3 |
|
---|
4 | import java.io.*;
|
---|
5 | import java.util.Properties;
|
---|
6 | import java.util.Iterator;
|
---|
7 | import java.util.Set;
|
---|
8 | import java.util.TreeSet;
|
---|
9 |
|
---|
10 | import org.apache.log4j.Logger;
|
---|
11 |
|
---|
12 | /**
|
---|
13 | * A single instance of the WETProcessor class can process a single unzipped warc.wet file.
|
---|
14 | * A WETProcessor take a warc.wet file and goes through all its WET records,
|
---|
15 | * putting each WET record into a file. Each file is put into a keep, discard or greylisted folder
|
---|
16 | * and its url listed written into a keep, discard or greylisted text file, based on:
|
---|
17 | *
|
---|
18 | * 1. whether it's whitelisted, else greylisted else blacklisted
|
---|
19 | * 2. and if explicitly whitelisted or else not greylisted or blacklisted and there's
|
---|
20 | * enough content. Formerly, content-length and number of lines were used to determine if
|
---|
21 | * the content was sufficient. Now it's just word count and number of MAX characters
|
---|
22 | * (not MINIMUM characters) that determine a string is a word.
|
---|
23 | * Explicit whitelisting has precedence over greylisting and which takes precedence
|
---|
24 | * over blacklisting in turn.
|
---|
25 | * However, even explicitly whitelisted urls still need to have sufficient content to end
|
---|
26 | * up in keepURLs.txt.
|
---|
27 | *
|
---|
28 | * See CCWETProcessor.java for compile instructions and how to run.
|
---|
29 | *
|
---|
30 | */
|
---|
31 | public class WETProcessor {
|
---|
32 | private static Logger logger = Logger.getLogger(org.greenstone.atea.WETProcessor.class.getName());
|
---|
33 |
|
---|
34 | // WARC WET header lines and header line prefixes of interest
|
---|
35 | static final String WARC_RECORD_START = "WARC/1.0";
|
---|
36 | static final String WARC_INFO_HEADER = "WARC-Type: warcinfo";
|
---|
37 | static final String WARC_TARGET_URI_HEADER_PREFIX = "WARC-Target-URI:";
|
---|
38 | static final String WARC_CONTENT_LENGTH_HEADER_PREFIX = "Content-Length:";
|
---|
39 |
|
---|
40 | private final String crawlID;
|
---|
41 | private final int WETFileID;
|
---|
42 | private final File inFile;
|
---|
43 |
|
---|
44 | private int recordCount = 0;
|
---|
45 |
|
---|
46 | /** Handle to a CCWETProcessor that processes a set of WET files
|
---|
47 | * Whereas a WETProcessor instance only processes a single WET file
|
---|
48 | * containing multiple WET records.
|
---|
49 | */
|
---|
50 | private CCWETProcessor batchProcessor;
|
---|
51 |
|
---|
52 | /**
|
---|
53 | * WET processor processes a single warc.wet file containing multiple WET records
|
---|
54 | * containing text identified as primary langcode=mri. Each individual WET record is written
|
---|
55 | * out to a uniquely named file in either the keep or discard folder depending on the WET
|
---|
56 | * record's content length and number of lines of actual content (excluding WARC headers).
|
---|
57 | * @param inFile the warc.wet file whose WET records are to be processed
|
---|
58 | * @param crawlID is the ID of the commoncrawl containing this warc.wet file
|
---|
59 | * and is of the format YYYY-## (of full crawlID CC-MAIN-YYYY-##) which will be used
|
---|
60 | * as prefix to create unique filenames when storing each individual record).
|
---|
61 | */
|
---|
62 | public WETProcessor(File inFile, String crawlID, CCWETProcessor batchProcessor) {
|
---|
63 | this.batchProcessor = batchProcessor;
|
---|
64 |
|
---|
65 | this.inFile = inFile;
|
---|
66 | this.crawlID = crawlID;
|
---|
67 |
|
---|
68 | // We just want a unique recordID prefix, which we get from concatenating
|
---|
69 | // the commoncrawl ID with the wet file name suffix and record count within the file:
|
---|
70 | // inFile name looks something like MAORI-CC-2019-30-20190902100139-000000.warc.wet
|
---|
71 | // the prefix will be everything after the last hyphen and without file extension,
|
---|
72 | // so "000000" in our example. Then converted into a number and padded to 2, e.g. 00.
|
---|
73 | // Then prefix the crawlID and suffix the 4-digit padded recordCount keeping track
|
---|
74 | // of the current WET record to get a unique filename to store each WET record into.
|
---|
75 | // e.g. 2019-30-00-0015 is the 15th WET record in the *00.warc.wet file of the
|
---|
76 | // common crawl CC-MAIN-2019-30 (15th WET record of CC-MAIN-2019-30-*-000000.warc.wet.gz)
|
---|
77 |
|
---|
78 | String fileID = inFile.getName();
|
---|
79 | //System.err.println("*** Processing wetfile: " + fileID);
|
---|
80 | fileID = fileID.substring(fileID.lastIndexOf("0")+1);
|
---|
81 | if(fileID.startsWith(".")) { // took off too many zeroes, as happens with *000000.warc.wet
|
---|
82 | this.WETFileID = 0;
|
---|
83 | } else {
|
---|
84 | fileID = fileID.substring(0, fileID.indexOf("."));
|
---|
85 | this.WETFileID = Integer.parseInt(fileID);
|
---|
86 | }
|
---|
87 | }
|
---|
88 |
|
---|
89 | /**
|
---|
90 | * Processes all the WET records of a single warc.wet file
|
---|
91 | */
|
---|
92 | public int processWETFile() {
|
---|
93 | File keepURLsFile = this.batchProcessor.keepURLsFile;
|
---|
94 | File discardURLsFile = this.batchProcessor.discardURLsFile;
|
---|
95 | File greyListedFile = this.batchProcessor.greyListedFile;
|
---|
96 |
|
---|
97 | StringBuilder record = null;
|
---|
98 | String line = null;
|
---|
99 | boolean readingRecord = false;
|
---|
100 |
|
---|
101 | String WARCtargetURI = "";
|
---|
102 |
|
---|
103 | //int recordCount = 0;
|
---|
104 |
|
---|
105 | int contentLength = -1; // of record
|
---|
106 | int lineCount = -1; // actual number of non-empty lines in record body (i.e. excludes WET/WARC headers)
|
---|
107 |
|
---|
108 | // read from WETfile
|
---|
109 | try (
|
---|
110 | BufferedReader reader = new BufferedReader(new FileReader(this.inFile));
|
---|
111 | BufferedWriter keepURLsWriter = new BufferedWriter(new FileWriter(keepURLsFile, true));
|
---|
112 | BufferedWriter discardURLsWriter = new BufferedWriter(new FileWriter(discardURLsFile, true)); // true to append
|
---|
113 | BufferedWriter greyListedURLsWriter = new BufferedWriter(new FileWriter(greyListedFile, true)); // true to append
|
---|
114 | ) {
|
---|
115 |
|
---|
116 | while((line = reader.readLine()) != null) { // readLine removes newline separator
|
---|
117 |
|
---|
118 | if(line.startsWith(WARC_INFO_HEADER)) { // "WARC-Type: warcinfo"
|
---|
119 | readingRecord = false;
|
---|
120 | record = null; // drop this record, which is just an info record not actual web page's text
|
---|
121 | recordCount--;
|
---|
122 | continue;
|
---|
123 | }
|
---|
124 |
|
---|
125 | if(line.startsWith(WARC_RECORD_START)) { // "WARC/1.0" means finished previous WET record
|
---|
126 | // process any previous record
|
---|
127 | if(record != null) {
|
---|
128 | processWETrecord(keepURLsWriter, discardURLsWriter, greyListedURLsWriter,
|
---|
129 | recordCount, contentLength, lineCount,
|
---|
130 | WARCtargetURI, record.toString());
|
---|
131 | record = null;
|
---|
132 | contentLength = -1;
|
---|
133 | lineCount = -1;
|
---|
134 | }
|
---|
135 |
|
---|
136 | recordCount++;
|
---|
137 | // get ready to start a new record
|
---|
138 | readingRecord = true;
|
---|
139 | record = new StringBuilder();
|
---|
140 | }
|
---|
141 |
|
---|
142 | if(readingRecord) { // append current line to current record
|
---|
143 |
|
---|
144 | if(line.startsWith(WARC_TARGET_URI_HEADER_PREFIX)) { // "WARC-Target-URI:"
|
---|
145 | // get and store the value
|
---|
146 | WARCtargetURI = line.substring(WARC_TARGET_URI_HEADER_PREFIX.length()).trim();
|
---|
147 | }
|
---|
148 |
|
---|
149 | record.append(line + "\n"); // add back (unix style) line ending
|
---|
150 |
|
---|
151 | // if the line is non-empty
|
---|
152 | // AND if we've started counting lines, which happens only when the current
|
---|
153 | // line is past WARC/WET headers and we're into the actual body portion
|
---|
154 | // of the WET record,
|
---|
155 | // start incrementing the line counter.
|
---|
156 | if(lineCount >= 0 && !line.trim().equals("")) {
|
---|
157 | lineCount++;
|
---|
158 | }
|
---|
159 | else if(line.startsWith(WARC_CONTENT_LENGTH_HEADER_PREFIX)) { // "Content-Length:"
|
---|
160 | String contentLengthStr = line.substring(WARC_CONTENT_LENGTH_HEADER_PREFIX.length()).trim();
|
---|
161 | contentLength = Integer.parseInt(contentLengthStr);
|
---|
162 | lineCount = 0;
|
---|
163 | }
|
---|
164 |
|
---|
165 | }
|
---|
166 |
|
---|
167 | }
|
---|
168 |
|
---|
169 | // flush the last record. If it was a warcinfo record, record would be null here
|
---|
170 | if(record != null) {
|
---|
171 | processWETrecord(keepURLsWriter, discardURLsWriter, greyListedURLsWriter,
|
---|
172 | recordCount, contentLength, lineCount,
|
---|
173 | WARCtargetURI, record.toString());
|
---|
174 | record = null;
|
---|
175 | }
|
---|
176 |
|
---|
177 | } catch(IOException ioe) {
|
---|
178 | ioe.printStackTrace();
|
---|
179 | }
|
---|
180 |
|
---|
181 | return recordCount;
|
---|
182 | }
|
---|
183 |
|
---|
184 | public int getRecordCount() { return this.recordCount; }
|
---|
185 |
|
---|
186 | /**
|
---|
187 | * Determines if a WET record belongs in the keep or discard pile depending on if it
|
---|
188 | * contains enough text, based on contentLength and line count of the record body.
|
---|
189 | * Then writes out the WET record to a uniquely named file in the keep or discard folder,
|
---|
190 | * and writes out the WET record's URL to the keepURLs.txt file or discardURLs.txt file.
|
---|
191 | */
|
---|
192 | private void processWETrecord(BufferedWriter keepURLsWriter, BufferedWriter discardURLsWriter,
|
---|
193 | BufferedWriter greyListedURLsWriter,
|
---|
194 | int recordID, int contentLength, int lineCount,
|
---|
195 | String recordURI, String record)
|
---|
196 | {
|
---|
197 | System.err.println("CrawlID: CC-MAIN-" + this.crawlID
|
---|
198 | + " WET #" + this.WETFileID
|
---|
199 | + " record #" + recordID
|
---|
200 | + " - contentLength: " + contentLength
|
---|
201 | + " - lineCount: " + lineCount);
|
---|
202 | System.err.println("URI: " + recordURI);
|
---|
203 | //System.err.println(record);
|
---|
204 | //System.err.println("--------------------------");
|
---|
205 |
|
---|
206 | File parentFolder = null;
|
---|
207 |
|
---|
208 | if(batchProcessor.isBlacklisted(recordURI)) {
|
---|
209 |
|
---|
210 | // explicit whitelisting overrides blacklisting
|
---|
211 | if(batchProcessor.isWhitelisted(recordURI)) {
|
---|
212 | parentFolder = batchProcessor.keepFolder; //tentative
|
---|
213 | }
|
---|
214 | // if not whitelisted, then greylisting still overrides blacklisting
|
---|
215 | else if(batchProcessor.isGreylisted(recordURI)) {
|
---|
216 | parentFolder = batchProcessor.greyListedFolder;
|
---|
217 | System.err.println("@@@GREYLISTED");
|
---|
218 | }
|
---|
219 | else { // url was only blacklisted
|
---|
220 | parentFolder = batchProcessor.discardFolder;
|
---|
221 | System.err.println("@@@DISCARDING - blacklisted");
|
---|
222 | }
|
---|
223 | }
|
---|
224 | else if(batchProcessor.isGreylisted(recordURI)) { // e.g. products sites
|
---|
225 | // explicit whitelisting overrides greylisting
|
---|
226 | if(batchProcessor.isWhitelisted(recordURI)) {
|
---|
227 | parentFolder = batchProcessor.keepFolder; // tentative
|
---|
228 | }
|
---|
229 | else {
|
---|
230 | parentFolder = batchProcessor.greyListedFolder;
|
---|
231 | System.err.println("@@@GREYLISTED");
|
---|
232 | }
|
---|
233 | }
|
---|
234 |
|
---|
235 | // If URL was not blacklisted/greylisted, or was even explicitly whitelisted,
|
---|
236 | // it still can't be in the keep list as it needs further inspection:
|
---|
237 | // it needs sufficient content for language analysis.
|
---|
238 | // We don't care about the combination of number of lines and content-length,
|
---|
239 | // we just care about the number of "valid words" as defined by us.
|
---|
240 | if(parentFolder != batchProcessor.greyListedFolder && parentFolder != batchProcessor.discardFolder) { // i.e. parentFolder == keepFolder if whiteListed || parentFolder == null
|
---|
241 |
|
---|
242 | // If a web page's WET record contains a certain minimum number of words,
|
---|
243 | // we will think it's a meaningful web page and has sufficient content for text analysis
|
---|
244 | // to have been successful. Cut off values at present are:
|
---|
245 | // - a minimum of 20 words
|
---|
246 | // - a word consists of 1 to 15 chars demarcated by spaces. Any more chars may point
|
---|
247 | // to words having been glued together. This is used by irrelevant sites and moreover
|
---|
248 | // can't be analysed for language, so may not be actually MRI.
|
---|
249 |
|
---|
250 | // Though StringTokenizer still in use, as seen in discussion at
|
---|
251 | // https://stackoverflow.com/questions/6983856/why-is-stringtokenizer-deprecated
|
---|
252 | // String.split(regex) seems better for splitting on general whitespace
|
---|
253 | String[] allWords = record.split("\\s");
|
---|
254 | int validWordCount = 0;
|
---|
255 | int numCamelCaseWords = 0;
|
---|
256 | for(int i = 0; i < allWords.length; i++) {
|
---|
257 | String word = allWords[i];
|
---|
258 |
|
---|
259 | // throw away if n words contain camelcase, which is another case of words glued together
|
---|
260 | // For now, we'll only skip camelcased words in our count of valid words
|
---|
261 | if(word.matches(".*[a-z][A-Z].*") && word.length() >= 5) {
|
---|
262 | numCamelCaseWords++;
|
---|
263 | }
|
---|
264 | // In Maori, word length of 1 is not uncommon
|
---|
265 | // but let's skip camelcased words when counting valid words
|
---|
266 | else if(word.length() >= 1 && word.length() <= batchProcessor.MAX_WORD_LENGTH) {
|
---|
267 | validWordCount++;
|
---|
268 | }
|
---|
269 | }
|
---|
270 |
|
---|
271 |
|
---|
272 | /*
|
---|
273 | // dump if too many camelcase words (ideally keep no WET record of that kind?)
|
---|
274 | if(numCamelCaseWords >= batchProcessor.MAX_WORDS_CAMELCASE) {
|
---|
275 | parentFolder = batchProcessor.discardFolder;
|
---|
276 | System.err.println("@@@DISCARDING - CAMELCASED CONTENTS");
|
---|
277 | }
|
---|
278 | else*/
|
---|
279 | // For now, don't discount content with too many camelcased words
|
---|
280 | // Just focus on whether there are a sufficient number of valid words
|
---|
281 | // (camelcased words are however still ignored in our count of valid words)
|
---|
282 | if(validWordCount >= batchProcessor.MIN_NUM_WORDS) { // otherwise, keep anything with a sufficient number of valid words
|
---|
283 | parentFolder = batchProcessor.keepFolder;
|
---|
284 | System.err.println("@@@KEEPING");
|
---|
285 | }
|
---|
286 | }
|
---|
287 | // if parentFolder still not set, it means that the content length/num words or lines
|
---|
288 | // were insufficient, so meant to be discarded
|
---|
289 | if(parentFolder == null) {
|
---|
290 | parentFolder = batchProcessor.discardFolder;
|
---|
291 | System.err.println("@@@DISCARDING");
|
---|
292 | }
|
---|
293 |
|
---|
294 | try {
|
---|
295 | if (parentFolder == batchProcessor.keepFolder) {
|
---|
296 | keepURLsWriter.write(recordURI + "\n");
|
---|
297 | } else if (parentFolder == batchProcessor.greyListedFolder) {
|
---|
298 | greyListedURLsWriter.write(recordURI + "\n");
|
---|
299 | } else {
|
---|
300 | discardURLsWriter.write(recordURI + "\n");
|
---|
301 | }
|
---|
302 | } catch(Exception e) {
|
---|
303 | System.err.println("Unable to write URL");
|
---|
304 | e.printStackTrace();
|
---|
305 | }
|
---|
306 |
|
---|
307 | System.err.println("--------------------------");
|
---|
308 |
|
---|
309 | // outFilename will look something like YYYY-##-####
|
---|
310 | String outFilename = String.format("%s-%02d-%04d", this.crawlID, this.WETFileID, recordID);
|
---|
311 | //= this.crawlID + "-" + String.format("%02d", this.WETFileID) + "-" + String.format("%04d.txt", recordID);
|
---|
312 | File outFile = new File(parentFolder, outFilename);
|
---|
313 |
|
---|
314 | try (BufferedWriter writer = new BufferedWriter(new FileWriter(outFile))) {
|
---|
315 | writer.write(record);
|
---|
316 | // Try-with-resources examples don't call close() explicitly:
|
---|
317 | // https://docs.oracle.com/javase/tutorial/essential/exceptions/tryResourceClose.html
|
---|
318 | //writer.close();
|
---|
319 | } catch(IOException ioe) {
|
---|
320 | ioe.printStackTrace();
|
---|
321 | System.err.println("\n@@@@@@@@@ Error writing to file " + outFile);
|
---|
322 | }
|
---|
323 | }
|
---|
324 | }
|
---|