1 | package org.hathitrust.extractedfeatures;
|
---|
2 |
|
---|
3 | import java.io.BufferedInputStream;
|
---|
4 | import java.io.BufferedOutputStream;
|
---|
5 | import java.io.FileInputStream;
|
---|
6 | import java.io.FileNotFoundException;
|
---|
7 | import java.io.IOException;
|
---|
8 | import java.io.InputStream;
|
---|
9 | import java.nio.charset.StandardCharsets;
|
---|
10 | import java.nio.file.Files;
|
---|
11 | import java.nio.file.Paths;
|
---|
12 | import java.util.stream.Stream;
|
---|
13 |
|
---|
14 | import com.google.common.hash.BloomFilter;
|
---|
15 | import com.google.common.hash.Funnel;
|
---|
16 | import com.google.common.hash.Funnels;
|
---|
17 |
|
---|
18 | public class WhitelistBloomFilter {
|
---|
19 |
|
---|
20 | protected BloomFilter<CharSequence> _bloomFilter;
|
---|
21 | protected static final String SERIALIZED_SUFFIX = "-serialized";
|
---|
22 | protected static final double FALSE_POSITIVE_PERCENTAGE = 0.01;
|
---|
23 |
|
---|
24 | // http://stackoverflow.com/questions/453018/number-of-lines-in-a-file-in-java
|
---|
25 | protected static int countLines(String filename) throws IOException
|
---|
26 | {
|
---|
27 | InputStream is = new BufferedInputStream(new FileInputStream(filename));
|
---|
28 |
|
---|
29 | try {
|
---|
30 | byte[] c = new byte[1024];
|
---|
31 | int count = 0;
|
---|
32 | int readChars = 0;
|
---|
33 | boolean empty = true;
|
---|
34 | while ((readChars = is.read(c)) != -1) {
|
---|
35 | empty = false;
|
---|
36 | for (int i = 0; i < readChars; ++i) {
|
---|
37 | if (c[i] == '\n') {
|
---|
38 | ++count;
|
---|
39 | }
|
---|
40 | }
|
---|
41 | }
|
---|
42 | return (count == 0 && !empty) ? 1 : count;
|
---|
43 | } finally {
|
---|
44 | is.close();
|
---|
45 | }
|
---|
46 | }
|
---|
47 |
|
---|
48 |
|
---|
49 | public WhitelistBloomFilter(String dictionary_filename, boolean serialize) {
|
---|
50 | System.out.println("Constructing: WhitelistBloomFilter");
|
---|
51 |
|
---|
52 | String ser_dictionary_filename = dictionary_filename + SERIALIZED_SUFFIX;
|
---|
53 |
|
---|
54 | if (ClusterFileIO.exists(ser_dictionary_filename)) {
|
---|
55 | System.out.println("Loading Serialized Bloom filter ...");
|
---|
56 | _bloomFilter = serializeIn(ser_dictionary_filename);
|
---|
57 | System.out.println("... done");
|
---|
58 | }
|
---|
59 | else {
|
---|
60 | // Need to generate the Bloom filter from the given raw text file
|
---|
61 |
|
---|
62 | System.out.println("Counting lines in: " + dictionary_filename);
|
---|
63 | int num_lines = -1;
|
---|
64 | try {
|
---|
65 | num_lines = countLines(dictionary_filename);
|
---|
66 |
|
---|
67 | Funnel<CharSequence> string_funnel = Funnels.stringFunnel(StandardCharsets.UTF_8);
|
---|
68 | _bloomFilter = BloomFilter.create(string_funnel, num_lines,FALSE_POSITIVE_PERCENTAGE);
|
---|
69 | }
|
---|
70 | catch (IOException e) {
|
---|
71 | e.printStackTrace();
|
---|
72 | }
|
---|
73 | System.out.println("Number of lines: " + num_lines);
|
---|
74 |
|
---|
75 | storeEntries(dictionary_filename,serialize);
|
---|
76 | }
|
---|
77 |
|
---|
78 | }
|
---|
79 |
|
---|
80 | protected void storeEntries(String filename, boolean serialize)
|
---|
81 | {
|
---|
82 | System.out.println("Building Bloom filter ...");
|
---|
83 |
|
---|
84 | //read file into stream, try-with-resources
|
---|
85 | try (Stream<String> stream = Files.lines(Paths.get(filename))) {
|
---|
86 | stream.forEach(word -> {_bloomFilter.put(word);});
|
---|
87 | } catch (IOException e) {
|
---|
88 | e.printStackTrace();
|
---|
89 | }
|
---|
90 |
|
---|
91 | System.out.println("... done");
|
---|
92 |
|
---|
93 | if (serialize) {
|
---|
94 | System.out.println("Serializing Bloom filter ...");
|
---|
95 |
|
---|
96 | String ser_filename = filename + SERIALIZED_SUFFIX;
|
---|
97 | serializeOut(ser_filename);
|
---|
98 |
|
---|
99 | System.out.println("... done");
|
---|
100 | }
|
---|
101 |
|
---|
102 | }
|
---|
103 |
|
---|
104 | public boolean contains(String key)
|
---|
105 | {
|
---|
106 | return _bloomFilter.mightContain(key);
|
---|
107 | }
|
---|
108 |
|
---|
109 | protected void serializeOut(String ser_filename)
|
---|
110 | {
|
---|
111 | try {
|
---|
112 | BufferedOutputStream bos = ClusterFileIO.getBufferedOutputStream(ser_filename);
|
---|
113 | _bloomFilter.writeTo(bos);
|
---|
114 | bos.close();
|
---|
115 | }
|
---|
116 | catch (FileNotFoundException e) {
|
---|
117 | System.err.println("Unable to open Bloom file:" + ser_filename);
|
---|
118 | //e.printStackTrace();
|
---|
119 | } catch (IOException e) {
|
---|
120 | System.err.println("Error reading in Bloom file:" + ser_filename);
|
---|
121 | //e.printStackTrace();
|
---|
122 | }
|
---|
123 | }
|
---|
124 |
|
---|
125 | protected static BloomFilter<CharSequence> serializeIn(String ser_filename)
|
---|
126 | {
|
---|
127 | BloomFilter<CharSequence> bloomFilter = null;
|
---|
128 |
|
---|
129 | try {
|
---|
130 | BufferedInputStream bis = ClusterFileIO.getBufferedInputStream(ser_filename);
|
---|
131 |
|
---|
132 | Funnel<CharSequence> string_funnel = Funnels.stringFunnel(StandardCharsets.UTF_8);
|
---|
133 | bloomFilter = BloomFilter.readFrom(bis,string_funnel);
|
---|
134 |
|
---|
135 | bis.close();
|
---|
136 | }
|
---|
137 | catch (FileNotFoundException e) {
|
---|
138 | System.err.println("Unable to open Bloom file:" + ser_filename);
|
---|
139 | //e.printStackTrace();
|
---|
140 | } catch (IOException e) {
|
---|
141 | System.err.println("Error writing out Bloom file:" + ser_filename);
|
---|
142 | //e.printStackTrace();
|
---|
143 | }
|
---|
144 | return bloomFilter;
|
---|
145 | }
|
---|
146 | } |
---|