// OpenGeoDB.java // (C) 2009 by Michael Peter Christen; mc@yacy.net, Frankfurt a. M., Germany // first published 04.10.2009 on http://yacy.net // // This is a part of YaCy // // $LastChangedDate$ // $LastChangedRevision$ // $LastChangedBy$ // // LICENSE // // This program is free software; you can redistribute it and/or modify // it under the terms of the GNU General Public License as published by // the Free Software Foundation; either version 2 of the License, or // (at your option) any later version. // // This program is distributed in the hope that it will be useful, // but WITHOUT ANY WARRANTY; without even the implied warranty of // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the // GNU General Public License for more details. // // You should have received a copy of the GNU General Public License // along with this program; if not, write to the Free Software // Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA package de.anomic.data; import java.io.BufferedReader; import java.io.File; import java.io.FileInputStream; import java.io.IOException; import java.io.InputStream; import java.io.InputStreamReader; import java.text.Collator; import java.util.ArrayList; import java.util.HashMap; import java.util.HashSet; import java.util.List; import java.util.Locale; import java.util.Map; import java.util.Set; import java.util.SortedMap; import java.util.TreeMap; import java.util.zip.GZIPInputStream; /** * this class loads and parses database dumps from the OpenGeoDB project * files can be loaded from http://sourceforge.net/projects/opengeodb/files/ * this class is used by the LibraryProvider, which expects input files inside * DATA\DICTIONARIES\source * * ATTENTION: * if this class is used, expect an extra memory usage of more than 100MB! * * This class will provide a super-fast access to the OpenGeoDB, * since all request are evaluated using data in the RAM. */ public class OpenGeoDB { // use a collator to relax when distinguishing between lowercase und uppercase letters private static final Collator insensitiveCollator = Collator.getInstance(Locale.US); static { insensitiveCollator.setStrength(Collator.SECONDARY); insensitiveCollator.setDecomposition(Collator.NO_DECOMPOSITION); } private HashMap locTypeHash2locType; private HashMap id2loc; private HashMap id2locTypeHash; private TreeMap> locationName2ids; private TreeMap> kfz2ids; private HashMap> predial2ids; private HashMap zip2id; public OpenGeoDB(final File file) { this.locTypeHash2locType = new HashMap(); this.id2loc = new HashMap(); this.id2locTypeHash = new HashMap(); this.locationName2ids = new TreeMap>(insensitiveCollator); this.kfz2ids = new TreeMap>(insensitiveCollator); this.predial2ids = new HashMap>(); this.zip2id = new HashMap(); if (file == null || !file.exists()) return; BufferedReader reader = null; try { InputStream is = new FileInputStream(file); if (file.getName().endsWith(".gz")) is = new GZIPInputStream(is); reader = new BufferedReader(new InputStreamReader(is, "UTF-8")); String line; // read lines String[] v; Integer id; String h; while ((line = reader.readLine()) != null) { line = line.trim(); if (!line.startsWith("INSERT INTO ")) continue; if (!line.endsWith(");")) continue; line = line.substring(12); //p = line.indexOf(' '); if (p < 0) continue; if (line.startsWith("geodb_coordinates ")) { line = line.substring(18 + 7);v = line.split(","); v = line.split(","); id2loc.put(Integer.parseInt(v[0]), new Location(Double.parseDouble(v[2]), Double.parseDouble(v[3]))); } if (line.startsWith("geodb_textdata ")) { line = line.substring(15 + 7); v = line.split(","); if (v[1].equals("500100000")) { // Ortsname id = Integer.parseInt(v[0]); h = removeQuotes(v[2]); List l = this.locationName2ids.get(h); if (l == null) l = new ArrayList(1); l.add(id); this.locationName2ids.put(h, l); Location loc = this.id2loc.get(id); if (loc != null) loc.setName(h); } else if (v[1].equals("500400000")) { // Vorwahl id = Integer.parseInt(v[0]); h = removeQuotes(v[2]); List l = this.predial2ids.get(h); if (l == null) l = new ArrayList(1); l.add(id); this.predial2ids.put(h, l); } else if (v[1].equals("400300000")) { // Ortstyp id = Integer.parseInt(v[0]); h = removeQuotes(v[2]); Integer hc = h.hashCode(); String t = this.locTypeHash2locType.get(hc); if (t == null) this.locTypeHash2locType.put(hc, h); this.id2locTypeHash.put(id, hc); } else if (v[1].equals("500300000")) { // PLZ this.zip2id.put(removeQuotes(v[2]), Integer.parseInt(v[0])); } else if (v[1].equals("500500000")) { // KFZ-Kennzeichen id = Integer.parseInt(v[0]); h = removeQuotes(v[2]); List l = this.kfz2ids.get(h); if (l == null) l = new ArrayList(1); l.add(id); this.kfz2ids.put(h, l); } } continue; } reader.close(); } catch (final IOException e) { e.printStackTrace(); } finally { if (reader != null) try { reader.close(); } catch (final Exception e) {} } } private static final String removeQuotes(String s) { if (s.charAt(0) != '\'') return s; if (s.charAt(s.length() - 1) != '\'') return s; s = s.substring(1, s.length() - 1); return s; } /** * check database tables against occurrences of this entity * the anyname - String may be one of: * - name of a town, villa, region etc * - zip code * - telephone prefix * - kfz sign * @param anyname * @return */ public HashSet find(String anyname) { HashSet r = new HashSet(); SortedMap> cities = this.locationName2ids.tailMap(anyname); for (Map.Entry> e: cities.entrySet()) { if (e.getKey().toLowerCase().startsWith(anyname.toLowerCase())) r.addAll(e.getValue()); else break; } List c; c = this.kfz2ids.get(anyname); if (c != null) r.addAll(c); c = this.predial2ids.get(anyname); if (c != null) r.addAll(c); Integer i = this.zip2id.get(anyname); if (i != null) r.add(i); HashSet a = new HashSet(); for (Integer e: r) { Location w = this.id2loc.get(e); if (w != null) a.add(w); } return a; } /** * read the dictionary and construct a set of recommendations to a given string * @param s input value that is used to match recommendations * @return a set that contains all words that start with the input value */ public Set recommend(String s) { Set a = new HashSet(); s = s.trim().toLowerCase(); SortedMap> t = this.locationName2ids.tailMap(s); for (String r: t.keySet()) { if (r.startsWith(s)) a.add(r); else break; } return a; } }