1 | package org.greenstone.atea;
|
---|
2 |
|
---|
3 | import java.io.*;
|
---|
4 | import java.util.HashMap;
|
---|
5 | import java.util.Map;
|
---|
6 |
|
---|
7 | import org.apache.log4j.Logger;
|
---|
8 |
|
---|
9 |
|
---|
10 | public class TextDumpPage {
|
---|
11 | private static Logger logger = Logger.getLogger(org.greenstone.atea.TextDumpPage.class.getName());
|
---|
12 |
|
---|
13 | public static final String TEXT_START_MARKER="text:start:";
|
---|
14 | public static final String TEXT_END_MARKER="text:end:";
|
---|
15 |
|
---|
16 | private Map<String, String> tuples;
|
---|
17 |
|
---|
18 | public TextDumpPage(String siteID, String unparsedPageDump) {
|
---|
19 | tuples = new HashMap<String, String>();
|
---|
20 |
|
---|
21 | try (
|
---|
22 | BufferedReader reader = new BufferedReader(new StringReader(unparsedPageDump));
|
---|
23 | ) {
|
---|
24 |
|
---|
25 | String line = reader.readLine(); // should have at least first line
|
---|
26 |
|
---|
27 | // first line always has a "key:" somewhere after the pageURL
|
---|
28 | int endIndex = line.indexOf("key:");
|
---|
29 |
|
---|
30 | //String pageURL = line.substring(0, endIndex);
|
---|
31 | String pageURL = (endIndex == -1) ? line : line.substring(0, endIndex);
|
---|
32 |
|
---|
33 |
|
---|
34 | tuples.put("pageURL", pageURL.trim());
|
---|
35 |
|
---|
36 | if(endIndex != -1) {
|
---|
37 | String key = line.substring(endIndex);
|
---|
38 | tuples.put("key", key.trim());
|
---|
39 | } else {
|
---|
40 | debug("@@@@ no key for pageURL: " + pageURL);
|
---|
41 | }
|
---|
42 |
|
---|
43 | if(pageURL.contains(TEXT_END_MARKER)) {
|
---|
44 | debug("@@@@ TEXT_END_MARKER assigned to pageURL for page: ");
|
---|
45 | debug("+++++++++");
|
---|
46 | debug(unparsedPageDump);
|
---|
47 | debug("+++++++++");
|
---|
48 | }
|
---|
49 |
|
---|
50 | boolean readingPageText = false;
|
---|
51 | StringBuilder pageText = null;
|
---|
52 |
|
---|
53 | // continue reading all other tuples for this page, if any
|
---|
54 | while((line = reader.readLine()) != null) {
|
---|
55 |
|
---|
56 | if(!readingPageText) {
|
---|
57 | // check if we're dealing with metadata or start/end of pagetext
|
---|
58 | endIndex = line.indexOf(":");
|
---|
59 | if(endIndex != -1) { // dealing with the rest of the page dump's metadata
|
---|
60 | String k = line.substring(0, endIndex);
|
---|
61 | String v = line.substring(endIndex+1);
|
---|
62 | tuples.put(k.trim(), v.trim());
|
---|
63 | }
|
---|
64 |
|
---|
65 | else if(line.equals(TEXT_START_MARKER)) { // dealing with the page body text
|
---|
66 | pageText = new StringBuilder();
|
---|
67 | readingPageText = true;
|
---|
68 | }
|
---|
69 | }
|
---|
70 |
|
---|
71 | else { // we're reading in the page text
|
---|
72 |
|
---|
73 | if(line.equals(TEXT_END_MARKER)) {
|
---|
74 | // finished with a page body
|
---|
75 | // remove any FINAL artificial newline we introduced
|
---|
76 | tuples.put("pageText", pageText.toString().trim());
|
---|
77 | readingPageText = false;
|
---|
78 | pageText = null;
|
---|
79 | }
|
---|
80 | else {
|
---|
81 | pageText.append(line);
|
---|
82 | pageText.append("\n"); // there are no newlines within pageText
|
---|
83 | // but if there were newlines, add them back here as readLine() removes them
|
---|
84 | }
|
---|
85 |
|
---|
86 | }
|
---|
87 | }
|
---|
88 |
|
---|
89 | // If the page had no pageText, add a "pageText" -> "" mapping
|
---|
90 | if(!tuples.containsKey("pageText")) {
|
---|
91 | tuples.put("pageText", "");
|
---|
92 | }
|
---|
93 |
|
---|
94 |
|
---|
95 | } catch (IOException ioe) {
|
---|
96 | error("@@@@@@@@@ Error reading in txtdump of a page.", ioe);
|
---|
97 | }
|
---|
98 |
|
---|
99 | /*
|
---|
100 | // START DEBUG
|
---|
101 | debug("__________________________________________");
|
---|
102 | for(Map.Entry<String, String> entry : tuples.entrySet()) {
|
---|
103 | String key = entry.getKey();
|
---|
104 | String value = entry.getValue();
|
---|
105 | debug(key + " - " + value);
|
---|
106 | }
|
---|
107 | debug("__________________________________________");
|
---|
108 | // END DEBUG
|
---|
109 | */
|
---|
110 | }
|
---|
111 |
|
---|
112 |
|
---|
113 | public String getPageURL() {
|
---|
114 | return tuples.get("url");
|
---|
115 | }
|
---|
116 |
|
---|
117 | public String getPageText() {
|
---|
118 | return tuples.get("pageText");
|
---|
119 | }
|
---|
120 |
|
---|
121 | public String get(String key) {
|
---|
122 | return tuples.get(key);
|
---|
123 | }
|
---|
124 |
|
---|
125 | // --------------- STATIC METHODS AND INNER CLASSED USED BY MAIN -------------- //
|
---|
126 | public static void info(String msg) {
|
---|
127 | System.err.println(msg);
|
---|
128 | logger.info(msg);
|
---|
129 | }
|
---|
130 | public static void debug(String msg) {
|
---|
131 | System.err.println(msg);
|
---|
132 | logger.debug(msg);
|
---|
133 | }
|
---|
134 | public static void warn(String msg) {
|
---|
135 | System.err.println(msg);
|
---|
136 | logger.warn(msg);
|
---|
137 | }
|
---|
138 | public static void error(String msg) {
|
---|
139 | System.err.println(msg);
|
---|
140 | logger.error(msg);
|
---|
141 | }
|
---|
142 | public static void error(String msg, Exception e) {
|
---|
143 | logger.error(msg, e);
|
---|
144 | System.err.println("\n"+msg);
|
---|
145 | e.printStackTrace();
|
---|
146 | }
|
---|
147 |
|
---|
148 | }
|
---|