1 | package org.greenstone.atea;
|
---|
2 |
|
---|
3 | import java.util.*;
|
---|
4 | import java.io.*;
|
---|
5 |
|
---|
6 | import org.apache.log4j.Logger;
|
---|
7 |
|
---|
8 | /**
|
---|
9 | * TO COMPILE OR RUN, FIRST DO:
|
---|
10 | * cd maori-lang-detection/apache-opennlp-1.9.1
|
---|
11 | * export OPENNLP_HOME=`pwd`
|
---|
12 | * cd maori-lang-detection/src
|
---|
13 | *
|
---|
14 | * TO COMPILE:
|
---|
15 | * maori-lang-detection/src$
|
---|
16 | * javac -cp ".:../conf:../lib/*" org/greenstone/atea/WebPageURLsListing.java
|
---|
17 | *
|
---|
18 | * TO RUN:
|
---|
19 | * maori-lang-detection/src$
|
---|
20 | * java -cp ".:../conf:../lib/*" org/greenstone/atea/WebPageURLsListing ../mongodb-data/domainsNZ_IsMRI.txt 255
|
---|
21 | *
|
---|
22 | */
|
---|
23 | public class WebPageURLsListing {
|
---|
24 | static Logger logger = Logger.getLogger(org.greenstone.atea.WebPageURLsListing.class.getName());
|
---|
25 |
|
---|
26 | private final MongoDBAccess mongodbAccess;
|
---|
27 | private int numURLs;
|
---|
28 | private File domainsFile;
|
---|
29 |
|
---|
30 |
|
---|
31 |
|
---|
32 | public WebPageURLsListing(MongoDBAccess mongodbAccess,
|
---|
33 | File domainsFile)
|
---|
34 | {
|
---|
35 | this.mongodbAccess = mongodbAccess;
|
---|
36 | this.domainsFile = domainsFile;
|
---|
37 | }
|
---|
38 |
|
---|
39 | public WebPageURLsListing(MongoDBAccess mongodbAccess,
|
---|
40 | File domainsFile,
|
---|
41 | int numURLs)
|
---|
42 | {
|
---|
43 | this(mongodbAccess, domainsFile);
|
---|
44 | this.numURLs = numURLs;
|
---|
45 | }
|
---|
46 |
|
---|
47 | public String produceURLsForPagesInMRI() {
|
---|
48 | return writeFile(MongoDBAccess.IS_MRI);
|
---|
49 | }
|
---|
50 |
|
---|
51 | public String produceURLsForPagesContainingMRI() {
|
---|
52 | return writeFile(MongoDBAccess.CONTAINS_MRI);
|
---|
53 | }
|
---|
54 |
|
---|
55 |
|
---|
56 | public String writeFile(int filterType) {
|
---|
57 |
|
---|
58 | ArrayList<String> urlsList = new ArrayList<String>();
|
---|
59 |
|
---|
60 | // 1. read each url from the domainsFile
|
---|
61 | // 1a. do the query
|
---|
62 | // 1b. add the arraylist result to urls
|
---|
63 |
|
---|
64 | try (
|
---|
65 | BufferedReader reader = new BufferedReader(new FileReader(domainsFile));
|
---|
66 | ) {
|
---|
67 |
|
---|
68 |
|
---|
69 | String domain;
|
---|
70 |
|
---|
71 | while((domain = reader.readLine()) != null) {
|
---|
72 | domain = domain.trim();
|
---|
73 | if(!domain.equals("")) {
|
---|
74 | ArrayList<String> moreURLs = mongodbAccess.queryAllMatchingURLsFilteredBy(domain, filterType);
|
---|
75 | urlsList.addAll(moreURLs);
|
---|
76 | }
|
---|
77 | }
|
---|
78 | } catch(Exception e) {
|
---|
79 | logger.error("Unable to read URLs from file " + domainsFile.getAbsolutePath());
|
---|
80 | logger.error(e.getMessage(), e);
|
---|
81 | }
|
---|
82 |
|
---|
83 | // Shuffle the urlsList, then write out the first numURLs into a file.
|
---|
84 | // https://stackoverflow.com/questions/5505927/how-to-generate-a-random-permutation-in-java
|
---|
85 | File parentFolder = domainsFile.getParentFile();
|
---|
86 | //File outFile = new File(parentFolder, "random"+numURLs+"_"+domainsFile.getName());
|
---|
87 | String fileName = (filterType == MongoDBAccess.IS_MRI) ? "isMRI_" : "containsMRI_";
|
---|
88 | File outFile = new File(parentFolder, fileName+domainsFile.getName());
|
---|
89 |
|
---|
90 | // write out ALL the URLs
|
---|
91 | try (
|
---|
92 | Writer writer = new BufferedWriter(new FileWriter(outFile));
|
---|
93 | ) {
|
---|
94 |
|
---|
95 | for (int i=0; i < urlsList.size(); i++) {
|
---|
96 | String url = urlsList.get(i);
|
---|
97 | //System.out.println(list.get(i));
|
---|
98 | writer.write(url + "\n");
|
---|
99 | }
|
---|
100 | } catch(Exception e) {
|
---|
101 | logger.error("Unable to write to file " + outFile.getAbsolutePath());
|
---|
102 | logger.error(e.getMessage(), e);
|
---|
103 | }
|
---|
104 |
|
---|
105 | /*
|
---|
106 | // shuffle list and take the first n - write to file
|
---|
107 | try (
|
---|
108 | Writer writer = new BufferedWriter(new FileWriter(outFile));
|
---|
109 | ) {
|
---|
110 | Collections.shuffle(urlsList);
|
---|
111 | for (int i=0; i<numURLs; i++) {
|
---|
112 | String url = urlsList.get(i);
|
---|
113 | //System.out.println(list.get(i));
|
---|
114 | writer.write(url + "\n");
|
---|
115 | }
|
---|
116 | } catch(Exception e) {
|
---|
117 | logger.error("Unable to write to file " + outFile.getAbsolutePath());
|
---|
118 | logger.error(e.getMessage(), e);
|
---|
119 | }
|
---|
120 | */
|
---|
121 |
|
---|
122 | return outFile.getAbsolutePath();
|
---|
123 | }
|
---|
124 |
|
---|
125 | /* ---------------------------------------- */
|
---|
126 |
|
---|
127 | /**
|
---|
128 | * Create the file
|
---|
129 | * @return full path of file generated
|
---|
130 | */
|
---|
131 | public String writeTentativeNonAutotranslatedSites() {
|
---|
132 | File outFolder = new File("../mongodb-data/").getAbsoluteFile();
|
---|
133 | File outFile = new File(outFolder, "5counts_tentativeNonAutotranslatedSites.json");
|
---|
134 |
|
---|
135 | String filename = outFile.getAbsolutePath();
|
---|
136 |
|
---|
137 | try (
|
---|
138 | Writer writer = new BufferedWriter(new FileWriter(outFile));
|
---|
139 | ) {
|
---|
140 | // first write out NZ sites and .nz TLD count and domains
|
---|
141 | mongodbAccess.aggregateContainsMRIForNZ(writer);
|
---|
142 | // next write out all overseas sites and .nz TLD count and domains
|
---|
143 | mongodbAccess.aggregateContainsMRIForOverseas(writer);
|
---|
144 |
|
---|
145 | filename = outFile.getCanonicalPath();
|
---|
146 | } catch(Exception e) {
|
---|
147 | logger.error("Unable to write to file " + outFile.getAbsolutePath());
|
---|
148 | logger.error(e.getMessage(), e);
|
---|
149 | }
|
---|
150 |
|
---|
151 | return filename;
|
---|
152 | }
|
---|
153 |
|
---|
154 |
|
---|
155 | public static void printUsage() {
|
---|
156 | System.err.println("WebPageURLsListing [domains.txt]");
|
---|
157 | }
|
---|
158 |
|
---|
159 | // Depending on args, generates isMRI and containsMRI file listings for:
|
---|
160 | // 1. UNFILTERED: all (NZ + overseas) AND takes manually curated domain list file for overseas and adds all NZ
|
---|
161 |
|
---|
162 |
|
---|
163 | public static void main(String args[]) {
|
---|
164 | if(args.length >= 2) {
|
---|
165 | printUsage();
|
---|
166 | System.exit(-1);
|
---|
167 | }
|
---|
168 |
|
---|
169 | try (
|
---|
170 | MongoDBAccess mongodb = new MongoDBAccess();
|
---|
171 | ) {
|
---|
172 | File domainsFile = new File(args[0]);
|
---|
173 | if(!domainsFile.exists()) {
|
---|
174 | System.err.println("File " + domainsFile + " does not exist");
|
---|
175 | System.exit(-1);
|
---|
176 | }
|
---|
177 |
|
---|
178 | //int genNumURLs = Integer.parseInt(args[1]);
|
---|
179 |
|
---|
180 | mongodb.connectToDB();
|
---|
181 |
|
---|
182 | WebPageURLsListing listing = new WebPageURLsListing(mongodb, domainsFile);
|
---|
183 | //String isMRIFile = listing.produceURLsForPagesInMRI();
|
---|
184 | //String containsMRIFile = listing.produceURLsForPagesContainingMRI();
|
---|
185 | String filename = listing.writeTentativeNonAutotranslatedSites();
|
---|
186 | System.err.println("Check file: " + filename);
|
---|
187 |
|
---|
188 |
|
---|
189 | } catch(Exception e) {
|
---|
190 | logger.error(e.getMessage(), e);
|
---|
191 | }
|
---|
192 | }
|
---|
193 | }
|
---|