|
| 1 | +/* |
| 2 | + * Copyright Elasticsearch B.V. and/or licensed to Elasticsearch B.V. under one |
| 3 | + * or more contributor license agreements. Licensed under the "Elastic License |
| 4 | + * 2.0", the "GNU Affero General Public License v3.0 only", and the "Server Side |
| 5 | + * Public License v 1"; you may not use this file except in compliance with, at |
| 6 | + * your election, the "Elastic License 2.0", the "GNU Affero General Public |
| 7 | + * License v3.0 only", or the "Server Side Public License, v 1". |
| 8 | + */ |
| 9 | + |
| 10 | +package org.elasticsearch.index.mapper.blockloader; |
| 11 | + |
| 12 | +import org.apache.lucene.util.BytesRef; |
| 13 | +import org.elasticsearch.common.geo.GeoPoint; |
| 14 | +import org.elasticsearch.geometry.Point; |
| 15 | +import org.elasticsearch.geometry.utils.WellKnownBinary; |
| 16 | +import org.elasticsearch.index.mapper.BlockLoaderTestCase; |
| 17 | +import org.elasticsearch.index.mapper.MappedFieldType; |
| 18 | + |
| 19 | +import java.nio.ByteOrder; |
| 20 | +import java.util.ArrayList; |
| 21 | +import java.util.Comparator; |
| 22 | +import java.util.List; |
| 23 | +import java.util.Map; |
| 24 | +import java.util.Objects; |
| 25 | + |
| 26 | +public class GeoPointFieldBlockLoaderTests extends BlockLoaderTestCase { |
| 27 | + public GeoPointFieldBlockLoaderTests(BlockLoaderTestCase.Params params) { |
| 28 | + super("geo_point", params); |
| 29 | + } |
| 30 | + |
| 31 | + @Override |
| 32 | + @SuppressWarnings("unchecked") |
| 33 | + protected Object expected(Map<String, Object> fieldMapping, Object value, TestContext testContext) { |
| 34 | + var extractedFieldValues = (ExtractedFieldValues) value; |
| 35 | + var values = extractedFieldValues.values(); |
| 36 | + |
| 37 | + var rawNullValue = fieldMapping.get("null_value"); |
| 38 | + |
| 39 | + GeoPoint nullValue; |
| 40 | + if (rawNullValue == null) { |
| 41 | + nullValue = null; |
| 42 | + } else if (rawNullValue instanceof String s) { |
| 43 | + nullValue = convert(s, null); |
| 44 | + } else { |
| 45 | + throw new IllegalStateException("Unexpected null_value format"); |
| 46 | + } |
| 47 | + |
| 48 | + if (params.preference() == MappedFieldType.FieldExtractPreference.DOC_VALUES && hasDocValues(fieldMapping, true)) { |
| 49 | + if (values instanceof List<?> == false) { |
| 50 | + var point = convert(values, nullValue); |
| 51 | + return point != null ? point.getEncoded() : null; |
| 52 | + } |
| 53 | + |
| 54 | + var resultList = ((List<Object>) values).stream() |
| 55 | + .map(v -> convert(v, nullValue)) |
| 56 | + .filter(Objects::nonNull) |
| 57 | + .map(GeoPoint::getEncoded) |
| 58 | + .sorted() |
| 59 | + .toList(); |
| 60 | + return maybeFoldList(resultList); |
| 61 | + } |
| 62 | + |
| 63 | + if (params.syntheticSource() == false) { |
| 64 | + return exactValuesFromSource(values, nullValue); |
| 65 | + } |
| 66 | + |
| 67 | + // Usually implementation of block loader from source adjusts values read from source |
| 68 | + // so that they look the same as doc_values would (like reducing precision). |
| 69 | + // geo_point does not do that and because of that we need to handle all these cases below. |
| 70 | + // If we are reading from stored source or fallback synthetic source we get the same exact data as source. |
| 71 | + // But if we are using "normal" synthetic source we get lesser precision data from doc_values. |
| 72 | + // That is unless "synthetic_source_keep" forces fallback synthetic source again. |
| 73 | + |
| 74 | + if (testContext.forceFallbackSyntheticSource()) { |
| 75 | + return exactValuesFromSource(values, nullValue); |
| 76 | + } |
| 77 | + |
| 78 | + String syntheticSourceKeep = (String) fieldMapping.getOrDefault("synthetic_source_keep", "none"); |
| 79 | + if (syntheticSourceKeep.equals("all")) { |
| 80 | + return exactValuesFromSource(values, nullValue); |
| 81 | + } |
| 82 | + if (syntheticSourceKeep.equals("arrays") && extractedFieldValues.documentHasObjectArrays()) { |
| 83 | + return exactValuesFromSource(values, nullValue); |
| 84 | + } |
| 85 | + |
| 86 | + // synthetic source and doc_values are present |
| 87 | + if (hasDocValues(fieldMapping, true)) { |
| 88 | + if (values instanceof List<?> == false) { |
| 89 | + return toWKB(normalize(convert(values, nullValue))); |
| 90 | + } |
| 91 | + |
| 92 | + var resultList = ((List<Object>) values).stream() |
| 93 | + .map(v -> convert(v, nullValue)) |
| 94 | + .filter(Objects::nonNull) |
| 95 | + .sorted(Comparator.comparingLong(GeoPoint::getEncoded)) |
| 96 | + .map(p -> toWKB(normalize(p))) |
| 97 | + .toList(); |
| 98 | + return maybeFoldList(resultList); |
| 99 | + } |
| 100 | + |
| 101 | + // synthetic source but no doc_values so using fallback synthetic source |
| 102 | + return exactValuesFromSource(values, nullValue); |
| 103 | + } |
| 104 | + |
| 105 | + @SuppressWarnings("unchecked") |
| 106 | + private Object exactValuesFromSource(Object value, GeoPoint nullValue) { |
| 107 | + if (value instanceof List<?> == false) { |
| 108 | + return toWKB(convert(value, nullValue)); |
| 109 | + } |
| 110 | + |
| 111 | + var resultList = ((List<Object>) value).stream().map(v -> convert(v, nullValue)).filter(Objects::nonNull).map(this::toWKB).toList(); |
| 112 | + return maybeFoldList(resultList); |
| 113 | + } |
| 114 | + |
| 115 | + private record ExtractedFieldValues(Object values, boolean documentHasObjectArrays) {} |
| 116 | + |
| 117 | + @Override |
| 118 | + protected Object getFieldValue(Map<String, Object> document, String fieldName) { |
| 119 | + var extracted = new ArrayList<>(); |
| 120 | + var documentHasObjectArrays = processLevel(document, fieldName, extracted, false); |
| 121 | + |
| 122 | + if (extracted.size() == 1) { |
| 123 | + return new ExtractedFieldValues(extracted.get(0), documentHasObjectArrays); |
| 124 | + } |
| 125 | + |
| 126 | + return new ExtractedFieldValues(extracted, documentHasObjectArrays); |
| 127 | + } |
| 128 | + |
| 129 | + @SuppressWarnings("unchecked") |
| 130 | + private boolean processLevel(Map<String, Object> level, String field, ArrayList<Object> extracted, boolean documentHasObjectArrays) { |
| 131 | + if (field.contains(".") == false) { |
| 132 | + var value = level.get(field); |
| 133 | + processLeafLevel(value, extracted); |
| 134 | + return documentHasObjectArrays; |
| 135 | + } |
| 136 | + |
| 137 | + var nameInLevel = field.split("\\.")[0]; |
| 138 | + var entry = level.get(nameInLevel); |
| 139 | + if (entry instanceof Map<?, ?> m) { |
| 140 | + return processLevel((Map<String, Object>) m, field.substring(field.indexOf('.') + 1), extracted, documentHasObjectArrays); |
| 141 | + } |
| 142 | + if (entry instanceof List<?> l) { |
| 143 | + for (var object : l) { |
| 144 | + processLevel((Map<String, Object>) object, field.substring(field.indexOf('.') + 1), extracted, true); |
| 145 | + } |
| 146 | + return true; |
| 147 | + } |
| 148 | + |
| 149 | + assert false : "unexpected document structure"; |
| 150 | + return false; |
| 151 | + } |
| 152 | + |
| 153 | + private void processLeafLevel(Object value, ArrayList<Object> extracted) { |
| 154 | + if (value instanceof List<?> l) { |
| 155 | + if (l.size() > 0 && l.get(0) instanceof Double) { |
| 156 | + // this must be a single point in array form |
| 157 | + // we'll put it into a different form here to make our lives a bit easier while implementing `expected` |
| 158 | + extracted.add(Map.of("type", "point", "coordinates", l)); |
| 159 | + } else { |
| 160 | + // this is actually an array of points but there could still be points in array form inside |
| 161 | + for (var arrayValue : l) { |
| 162 | + processLeafLevel(arrayValue, extracted); |
| 163 | + } |
| 164 | + } |
| 165 | + } else { |
| 166 | + extracted.add(value); |
| 167 | + } |
| 168 | + } |
| 169 | + |
| 170 | + @SuppressWarnings("unchecked") |
| 171 | + private GeoPoint convert(Object value, GeoPoint nullValue) { |
| 172 | + if (value == null) { |
| 173 | + return nullValue; |
| 174 | + } |
| 175 | + |
| 176 | + if (value instanceof String s) { |
| 177 | + try { |
| 178 | + return new GeoPoint(s); |
| 179 | + } catch (Exception e) { |
| 180 | + return null; |
| 181 | + } |
| 182 | + } |
| 183 | + |
| 184 | + if (value instanceof Map<?, ?> m) { |
| 185 | + if (m.get("type") != null) { |
| 186 | + var coordinates = (List<Double>) m.get("coordinates"); |
| 187 | + // Order is GeoJSON is lon,lat |
| 188 | + return new GeoPoint(coordinates.get(1), coordinates.get(0)); |
| 189 | + } else { |
| 190 | + return new GeoPoint((Double) m.get("lat"), (Double) m.get("lon")); |
| 191 | + } |
| 192 | + } |
| 193 | + |
| 194 | + // Malformed values are excluded |
| 195 | + return null; |
| 196 | + } |
| 197 | + |
| 198 | + private GeoPoint normalize(GeoPoint point) { |
| 199 | + if (point == null) { |
| 200 | + return null; |
| 201 | + } |
| 202 | + return point.resetFromEncoded(point.getEncoded()); |
| 203 | + } |
| 204 | + |
| 205 | + private BytesRef toWKB(GeoPoint point) { |
| 206 | + if (point == null) { |
| 207 | + return null; |
| 208 | + } |
| 209 | + |
| 210 | + return new BytesRef(WellKnownBinary.toWKB(new Point(point.getX(), point.getY()), ByteOrder.LITTLE_ENDIAN)); |
| 211 | + } |
| 212 | +} |
0 commit comments