package com.hughes.android.dictionary;\r
\r
-import java.io.DataOutputStream;\r
-import java.io.File;\r
-import java.io.FileOutputStream;\r
+import java.io.FileNotFoundException;\r
import java.io.IOException;\r
import java.io.RandomAccessFile;\r
import java.io.Serializable;\r
import java.util.ArrayList;\r
-import java.util.LinkedHashSet;\r
+import java.util.Collections;\r
+import java.util.LinkedHashMap;\r
import java.util.List;\r
import java.util.Map;\r
-import java.util.Set;\r
import java.util.TreeMap;\r
-import java.util.concurrent.atomic.AtomicInteger;\r
import java.util.regex.Pattern;\r
\r
-import com.hughes.android.dictionary.Index.Node;\r
import com.hughes.util.FileUtil;\r
\r
public class IndexBuilder {\r
System.err.println("No input file.");\r
return;\r
}\r
+ final String dictionaryFileName = args[0];\r
+ createIndex(dictionaryFileName, Entry.LANG1);\r
+ createIndex(dictionaryFileName, Entry.LANG2);\r
+ }\r
\r
- final String file = args[0];\r
- final byte lang = Entry.LANG1;\r
+ private static void createIndex(final String dictionaryFileName,\r
+ final byte lang) throws IOException, FileNotFoundException,\r
+ ClassNotFoundException {\r
Node rootBuilder;\r
-// rootBuilder = createIndex(file, lang);\r
-// FileUtil.write(rootBuilder, String.format("%s_builder_%d.serialized", file, lang));\r
- rootBuilder = (Node) FileUtil.read(String.format("%s_builder_%d.serialized", file, lang));\r
- \r
- final AtomicInteger c = new AtomicInteger();\r
+ rootBuilder = processDictionaryLines(dictionaryFileName, lang);\r
+ FileUtil.write(rootBuilder, String.format("%s_builder_%d.serialized", dictionaryFileName, lang));\r
+ rootBuilder = (Node) FileUtil.read(String.format("%s_builder_%d.serialized", dictionaryFileName, lang));\r
+\r
rootBuilder.forEachNode(new Function<Node>() {\r
@Override\r
- public void invoke(Node t) {\r
- if (t.offsetsList.size() > 200) {\r
- System.out.println(t);\r
- c.incrementAndGet();\r
+ public void invoke(final Node node) {\r
+ for (final List<EntryDescriptor> entryDescriptors : node.entryDescriptorsMap.values()) {\r
+ Collections.sort(entryDescriptors);\r
}\r
}});\r
- System.out.println(c);\r
\r
- rootBuilder.recursiveSetDescendantOffsetCount();\r
- rootBuilder.packDescendants(128);\r
-\r
- final DataOutputStream os = new DataOutputStream(new FileOutputStream(\r
- String.format("%s_index_%d", file, lang)));\r
- final Index.Node root = rootBuilder.toIndexNode();\r
- root.write(os);\r
- os.close();\r
- \r
- FileUtil.write(root, String.format("%s_index_%d.serialized", file, lang));\r
- \r
- Object o = FileUtil.read(String.format("%s_index_%d.serialized", file, lang));\r
-\r
-\r
+ // Dump twice to get accurate file locations.\r
+ for (int i = 0; i < 2; ++i) {\r
+ final RandomAccessFile raf = new RandomAccessFile(String.format(Dictionary.INDEX_FORMAT, dictionaryFileName, lang), "rw"); \r
+ rootBuilder.dump(raf);\r
+ raf.close();\r
+ }\r
}\r
\r
// ----------------------------------------------------------------\r
-\r
- static final class Node implements Serializable {\r
- private static final long serialVersionUID = -5423134653901704956L;\r
- \r
- final TreeMap<String, Node> childrenMap = new TreeMap<String, Node>();\r
- final List<Integer> offsetsList = new ArrayList<Integer>();\r
- final String sequence;\r
- int descendantOffsetCount = 0;\r
-\r
- public Node(String sequence) {\r
- if (sequence.length() == 0) {\r
- System.out.println("Created root.");\r
- }\r
- this.sequence = sequence.intern();\r
- }\r
-\r
- public Node getIndexNode(final String word, final int pos,\r
- final boolean create) {\r
- assert this.sequence.equals(word.substring(0, pos));\r
-\r
- if (pos == word.length()) {\r
- assert sequence.equals(word);\r
- return this;\r
- }\r
-\r
- final String rest = word.substring(pos);\r
- assert rest.length() > 0;\r
-\r
- final Map.Entry<String, Node> lcsEntry;\r
- final String lcs;\r
- {\r
- final Map.Entry<String, Node> floorEntry = childrenMap.floorEntry(rest);\r
- final Map.Entry<String, Node> ceilingEntry = childrenMap\r
- .ceilingEntry(rest);\r
- final String floorLcs = floorEntry == null ? "" : StringUtil\r
- .longestCommonSubstring(rest, floorEntry.getKey());\r
- final String ceilingLcs = ceilingEntry == null ? "" : StringUtil\r
- .longestCommonSubstring(rest, ceilingEntry.getKey());\r
- if (floorLcs.length() > ceilingLcs.length()) {\r
- lcsEntry = floorEntry;\r
- lcs = floorLcs;\r
- } else {\r
- lcsEntry = ceilingEntry;\r
- lcs = ceilingLcs;\r
- }\r
- }\r
-\r
- // No LCS, have to add everything.\r
- if (lcs.length() == 0) {\r
- if (!create) {\r
- return null;\r
- }\r
- final Node result = new Node(word);\r
- final Object old = childrenMap.put(rest.intern(), result);\r
- assert old == null;\r
- // System.out.println(" Adding final chunk: " + rest);\r
- return result;\r
- }\r
-\r
- assert lcsEntry != null;\r
-\r
- // The map already contained the LCS.\r
- if (lcs.length() == lcsEntry.getKey().length()) {\r
- assert lcs.equals(lcsEntry.getKey());\r
- final Node result = lcsEntry.getValue().getIndexNode(word,\r
- pos + lcs.length(), create);\r
- assert result.sequence.equals(word);\r
- return result;\r
- }\r
-\r
- if (!create) {\r
- return null;\r
- }\r
-\r
- // Have to split, inserting the LCS.\r
- // System.out.println(" Splitting " + lcsEntry + "/" + word + " @ " +\r
- // lcs);\r
- final Node newChild = new Node(word.substring(0, pos + lcs.length()));\r
- final Object old = childrenMap.put(lcs.intern(), newChild);\r
- assert old == null;\r
- childrenMap.remove(lcsEntry.getKey());\r
- newChild.childrenMap.put(lcsEntry.getKey().substring(lcs.length())\r
- .intern(), lcsEntry.getValue());\r
-\r
- if (lcs.equals(rest)) {\r
- return newChild;\r
- }\r
- final Node result = new Node(word);\r
- final Object old2 = newChild.childrenMap.put(rest.substring(lcs.length())\r
- .intern(), result);\r
- assert old2 == null;\r
- // System.out.println(" newChildrenMap=" + newChild.childrenMap);\r
-\r
- return result;\r
- }\r
-\r
- Index.Node toIndexNode() {\r
- final Index.Node result = new Index.Node(childrenMap.size(), offsetsList\r
- .size());\r
- int i = 0;\r
- for (final Map.Entry<String, Node> entry : childrenMap.entrySet()) {\r
- result.chars[i] = entry.getKey();\r
- result.children[i] = entry.getValue().toIndexNode();\r
- i++;\r
- }\r
- return result;\r
- }\r
-\r
- void forEachNode(final Function<Node> f) {\r
- f.invoke(this);\r
- for (final Node child : childrenMap.values()) {\r
- child.forEachNode(f);\r
- }\r
- }\r
-\r
- int descendantCount() {\r
- int count = 1;\r
- for (final Node child : childrenMap.values()) {\r
- count += child.descendantCount();\r
- }\r
- return count;\r
- }\r
-\r
- void recursiveSetDescendantOffsetCount() {\r
- descendantOffsetCount = offsetsList.size();\r
- for (final Node child : childrenMap.values()) {\r
- child.recursiveSetDescendantOffsetCount();\r
- descendantOffsetCount += child.descendantOffsetCount;\r
- }\r
+ \r
+ static final class EntryDescriptor implements Comparable<EntryDescriptor>, Serializable {\r
+ final int offset;\r
+ final int numTokens;\r
+ public EntryDescriptor(int offset, int numTokens) {\r
+ this.offset = offset;\r
+ this.numTokens = numTokens;\r
}\r
-\r
- public void packDescendants(final int maxDescendants) {\r
- if (descendantOffsetCount <= maxDescendants) {\r
- final Set<Integer> descendantOffsets = new LinkedHashSet<Integer>();\r
- recursiveAddDescendants(descendantOffsets);\r
- assert descendantOffsets.size() <= maxDescendants;\r
- offsetsList.clear();\r
- offsetsList.addAll(descendantOffsets);\r
- childrenMap.clear();\r
- } else {\r
- for (final Node child : childrenMap.values()) {\r
- child.packDescendants(maxDescendants);\r
- }\r
- }\r
+ @Override\r
+ public boolean equals(Object obj) {\r
+ final EntryDescriptor that = (EntryDescriptor) obj;\r
+ return this.offset == that.offset;\r
}\r
-\r
- private void recursiveAddDescendants(final Set<Integer> descendantOffsets) {\r
- descendantOffsets.addAll(this.offsetsList);\r
- for (final Node child : childrenMap.values()) {\r
- child.recursiveAddDescendants(descendantOffsets);\r
- }\r
+ @Override\r
+ public int hashCode() {\r
+ return offset;\r
}\r
-\r
-\r
@Override\r
- public String toString() {\r
- return sequence + ":" + offsetsList.size();\r
+ public int compareTo(EntryDescriptor o) {\r
+ return this.numTokens < o.numTokens ? -1 : this.numTokens == o.numTokens ? 0 : 1;\r
}\r
-\r
}\r
\r
+\r
// ----------------------------------------------------------------\r
\r
- static Node createIndex(final String file, final byte lang) throws IOException {\r
+ static Node processDictionaryLines(final String dictionaryFileName, final byte lang) throws IOException {\r
final Node root = new Node("");\r
- final RandomAccessFile raf = new RandomAccessFile(file, "r");\r
+ final RandomAccessFile dictionaryFile = new RandomAccessFile(dictionaryFileName, "r");\r
String line;\r
final Entry entry = new Entry();\r
int lineCount = 0;\r
- while ((line = raf.readLine()) != null) {\r
- final long fileLocation = raf.getFilePointer();\r
+ long fileLocation = 0;\r
+ while ((line = dictionaryFile.readLine()) != null) {\r
assert ((int) fileLocation) == fileLocation;\r
\r
line = line.trim();\r
}\r
final String text = entry.getIndexableText(Entry.LANG1);\r
final String[] tokens = WHITESPACE.split(text);\r
- final Set<String> tokenSet = new LinkedHashSet<String>();\r
+ final Map<String,String> tokenToNormalizedMap = new LinkedHashMap<String,String>();\r
for (String token : tokens) {\r
if (token.length() <= 1 || !Character.isLetter(token.charAt(0))) {\r
continue;\r
}\r
- tokenSet.add(entry.normalizeToken(token, lang));\r
+ tokenToNormalizedMap.put(token, EntryFactory.entryFactory.normalizeToken(token));\r
}\r
- for (final String normalized : tokenSet) {\r
- // System.out.println("Inserting: " + normalized);\r
- if ("die".equals(normalized) || "eine".equals(normalized)) {\r
- // System.out.println("hello");\r
- }\r
- final Node node = root.getIndexNode(normalized, 0, true);\r
- node.offsetsList.add((int) fileLocation);\r
- assert node == root.getIndexNode(normalized, 0, false);\r
- assert normalized\r
- .equals(root.getIndexNode(normalized, 0, false).sequence);\r
+ for (final Map.Entry<String, String> tokenToNormalized : tokenToNormalizedMap.entrySet()) {\r
+ final String normalizedToken = tokenToNormalized.getValue();\r
+ final Node node = root.getNode(normalizedToken, 0, true);\r
+ node.addToken(tokenToNormalized.getKey(), new EntryDescriptor((int) fileLocation, tokens.length));\r
+ assert node == root.getNode(normalizedToken, 0, false);\r
+ assert normalizedToken\r
+ .equals(root.getNode(normalizedToken, 0, false).normalizedToken);\r
}\r
\r
if (lineCount % 10000 == 0) {\r
System.out.println("IndexBuilder: " + "lineCount=" + lineCount);\r
}\r
+ \r
lineCount++;\r
+ fileLocation = dictionaryFile.getFilePointer();\r
}\r
- raf.close();\r
+ dictionaryFile.close();\r
+ \r
+ root.recursiveSetDescendantCounts();\r
+ \r
return root;\r
}\r
\r
projects / DictionaryPC.git / blobdiff