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 |
|
---|
51 | boolean readingPageText = false;
|
---|
52 | StringBuilder pageText = null;
|
---|
53 |
|
---|
54 | // continue reading all other tuples for this page, if any
|
---|
55 | while((line = reader.readLine()) != null) {
|
---|
56 | line = line.trim();
|
---|
57 |
|
---|
58 | // check if we're dealing with metadata or start/end of page's text body
|
---|
59 | // or actual text body
|
---|
60 |
|
---|
61 | if(line.equals(TEXT_START_MARKER)) { // dealing with the page body text
|
---|
62 | pageText = new StringBuilder();
|
---|
63 | readingPageText = true;
|
---|
64 | }
|
---|
65 | else if(line.equals(TEXT_END_MARKER)) {
|
---|
66 | // finished with a page body
|
---|
67 | // Remove any FINAL artificial newline we introduced to a page's body text
|
---|
68 | tuples.put("pageText", pageText.toString().trim());
|
---|
69 | readingPageText = false;
|
---|
70 | pageText = null;
|
---|
71 | }
|
---|
72 | else {
|
---|
73 | if(readingPageText) { // So we're reading in the page text
|
---|
74 | pageText.append(line);
|
---|
75 | pageText.append("\n"); // there are no newlines within pageText
|
---|
76 | // but if there were newlines, add them back here as readLine() removes them
|
---|
77 | }
|
---|
78 | else { // dealing with the rest of the page dump's metadata
|
---|
79 | endIndex = line.indexOf(":");
|
---|
80 | if(endIndex != -1) {
|
---|
81 | String k = line.substring(0, endIndex);
|
---|
82 | String v = line.substring(endIndex+1);
|
---|
83 | tuples.put(k.trim(), v.trim());
|
---|
84 | } else {
|
---|
85 | error("No meta key for meta: " + line);
|
---|
86 | }
|
---|
87 | }
|
---|
88 | }
|
---|
89 |
|
---|
90 | }
|
---|
91 |
|
---|
92 | // If the page had no pageText, add a "pageText" -> "" mapping
|
---|
93 | if(!tuples.containsKey("pageText")) {
|
---|
94 | tuples.put("pageText", "");
|
---|
95 | }
|
---|
96 |
|
---|
97 |
|
---|
98 | } catch (IOException ioe) {
|
---|
99 | error("@@@@@@@@@ Error reading in txtdump of a page.", ioe);
|
---|
100 | }
|
---|
101 |
|
---|
102 |
|
---|
103 | // START DEBUG
|
---|
104 | debug("__________________________________________");
|
---|
105 | for(Map.Entry<String, String> entry : tuples.entrySet()) {
|
---|
106 | String key = entry.getKey();
|
---|
107 | String value = entry.getValue();
|
---|
108 | debug(key + " - " + value);
|
---|
109 | }
|
---|
110 | debug("__________________________________________");
|
---|
111 | // END DEBUG
|
---|
112 |
|
---|
113 | }
|
---|
114 |
|
---|
115 |
|
---|
116 | public String getPageURL() {
|
---|
117 | return tuples.get("url");
|
---|
118 | }
|
---|
119 |
|
---|
120 | public String getPageText() {
|
---|
121 | return tuples.get("pageText");
|
---|
122 | }
|
---|
123 |
|
---|
124 | public String get(String key) {
|
---|
125 | return tuples.get(key);
|
---|
126 | }
|
---|
127 |
|
---|
128 | // --------------- STATIC METHODS AND INNER CLASSED USED BY MAIN -------------- //
|
---|
129 | public static void info(String msg) {
|
---|
130 | System.err.println(msg);
|
---|
131 | logger.info(msg);
|
---|
132 | }
|
---|
133 | public static void debug(String msg) {
|
---|
134 | System.err.println(msg);
|
---|
135 | logger.debug(msg);
|
---|
136 | }
|
---|
137 | public static void warn(String msg) {
|
---|
138 | System.err.println(msg);
|
---|
139 | logger.warn(msg);
|
---|
140 | }
|
---|
141 | public static void error(String msg) {
|
---|
142 | System.err.println(msg);
|
---|
143 | logger.error(msg);
|
---|
144 | }
|
---|
145 | public static void error(String msg, Exception e) {
|
---|
146 | logger.error(msg, e);
|
---|
147 | System.err.println("\n"+msg);
|
---|
148 | e.printStackTrace();
|
---|
149 | }
|
---|
150 |
|
---|
151 | }
|
---|