All Downloads are FREE. Search and download functionalities are using the official Maven repository.
Please wait. This can take some minutes ...
Many resources are needed to download a project. Please understand that we have to compensate our server costs. Thank you in advance.
Project price only 1 $
You can buy this project and download/modify it how often you want.
com.yahoo.schema.parser.ParsedField Maven / Gradle / Ivy
// Copyright Vespa.ai. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
package com.yahoo.schema.parser;
import com.yahoo.schema.document.Stemming;
import java.util.ArrayList;
import java.util.Collections;
import java.util.LinkedHashMap;
import java.util.List;
import java.util.Map;
import java.util.Optional;
/**
* This class holds the extracted information after parsing a "field"
* block, using simple data structures as far as possible. Do not put
* advanced logic here!
* @author arnej27959
**/
class ParsedField extends ParsedBlock {
private ParsedType type;
private boolean hasBolding = false;
private boolean isFilter = false;
private int overrideId = 0;
private boolean isLiteral = false;
private boolean isNormal = false;
private Integer weight;
private String normalizing = null;
private final ParsedMatchSettings matchInfo = new ParsedMatchSettings();
private Stemming stemming = null;
private ParsedIndexingOp indexingOp = null;
private ParsedSorting sortSettings = null;
private final Map attributes = new LinkedHashMap<>();
private final Map fieldIndexes = new LinkedHashMap<>();
private final Map aliases = new LinkedHashMap<>();
private final Map rankTypes = new LinkedHashMap<>();
private final Map structFields = new LinkedHashMap<>();
private final Map summaryFields = new LinkedHashMap<>();
private final List dictionaryOptions = new ArrayList<>();
private final List queryCommands = new ArrayList<>();
ParsedField(String name, ParsedType type) {
super(name, "field");
this.type = type;
}
ParsedType getType() { return this.type; }
boolean hasBolding() { return this.hasBolding; }
boolean hasFilter() { return this.isFilter; }
boolean hasLiteral() { return this.isLiteral; }
boolean hasNormal() { return this.isNormal; }
boolean hasIdOverride() { return overrideId != 0; }
int idOverride() { return overrideId; }
List getDictionaryOptions() { return List.copyOf(dictionaryOptions); }
List getAttributes() { return List.copyOf(attributes.values()); }
List getIndexes() { return List.copyOf(fieldIndexes.values()); }
List getSummaryFields() { return List.copyOf(summaryFields.values()); }
List getStructFields() { return List.copyOf(structFields.values()); }
List getAliases() { return List.copyOf(aliases.keySet()); }
List getQueryCommands() { return List.copyOf(queryCommands); }
String lookupAliasedFrom(String alias) { return aliases.get(alias); }
ParsedMatchSettings matchSettings() { return this.matchInfo; }
Optional getWeight() { return Optional.ofNullable(weight); }
Optional getStemming() { return Optional.ofNullable(stemming); }
Optional getNormalizing() { return Optional.ofNullable(normalizing); }
Optional getIndexing() { return Optional.ofNullable(indexingOp); }
Optional getSorting() { return Optional.ofNullable(sortSettings); }
Map getRankTypes() { return Collections.unmodifiableMap(rankTypes); }
/** get an existing summary field for modification, or create it */
ParsedSummaryField summaryFieldFor(String name) {
if (summaryFields.containsKey(name)) {
return summaryFields.get(name);
}
var sf = new ParsedSummaryField(name, getType());
summaryFields.put(name, sf);
return sf;
}
/** get an existing summary field for modification, or create it */
ParsedSummaryField summaryFieldFor(String name, ParsedType type) {
if (summaryFields.containsKey(name)) {
var sf = summaryFields.get(name);
if (sf.getType() == null) {
sf.setType(type);
} else {
// TODO check that types are properly equal here
String oldName = sf.getType().name();
String newName = type.name();
verifyThat(newName.equals(oldName), "type mismatch for summary field", name, ":", oldName, "/", newName);
}
return sf;
}
var sf = new ParsedSummaryField(name, type);
summaryFields.put(name, sf);
return sf;
}
void addAlias(String from, String to) {
verifyThat(! aliases.containsKey(to), "already has alias", to);
aliases.put(to, from);
}
void addIndex(ParsedIndex index) {
String idxName = index.name();
verifyThat(! fieldIndexes.containsKey(idxName), "already has index", idxName);
fieldIndexes.put(idxName, index);
}
void addRankType(String index, String rankType) {
rankTypes.put(index, rankType);
}
void dictionary(DictionaryOption option) {
dictionaryOptions.add(option);
}
void setBolding(boolean value) { this.hasBolding = value; }
void setFilter(boolean value) { this.isFilter = value; }
void setId(int id) { this.overrideId = id; }
void setLiteral(boolean value) { this.isLiteral = value; }
void setNormal(boolean value) { this.isNormal = value; }
void setNormalizing(String value) { this.normalizing = value; }
void setStemming(Stemming stemming) { this.stemming = stemming; }
void setWeight(int weight) { this.weight = weight; }
ParsedAttribute attributeFor(String attrName) {
return attributes.computeIfAbsent(attrName, n -> new ParsedAttribute(n));
}
void setIndexingOperation(ParsedIndexingOp idxOp) {
verifyThat(indexingOp == null, "already has indexing");
indexingOp = idxOp;
}
ParsedSorting sortInfo() {
if (sortSettings == null) sortSettings = new ParsedSorting(name(), "field.sorting");
return this.sortSettings;
}
void addQueryCommand(String command) {
queryCommands.add(command);
}
void addStructField(ParsedField structField) {
String fieldName = structField.name();
verifyThat(! structFields.containsKey(fieldName), "already has struct-field", fieldName);
structFields.put(fieldName, structField);
}
void addSummaryField(ParsedSummaryField summaryField) {
String fieldName = summaryField.name();
verifyThat(! summaryFields.containsKey(fieldName), "already has summary field", fieldName);
if (summaryField.getType() == null) {
summaryField.setType(getType());
}
summaryFields.put(fieldName, summaryField);
}
}