AntIndex 2.5.2

AntIndex

Linked data text search engine.

Features

• Fast text search using 2-gramm words index
• Designed to search for entities of different types (up to 250 types)
• Support for data hierarchy and relatedness
• Flexible sorting settings
• MessagePack binary serialization enabled index file saving utilities

Documentation

Normalizing and splitting

The library provides built-in tools for normalizing and splitting a name into words for searching. Use this static objects.

AntIndex.Services.Normalizing.DefaultNormalizer.Instance AntIndex.Services.Splitting.DefaultPhraseSplitter.Instance

Also, you can redefine normalization and phrases splitting on words for use in your cases. Implement: IPhraseSplitter and INormalizer

Defining entities of search

• Your search entity must implement the interface IIndexedEntity
  • GetKey() - Defines an entity keys
    • Use Ant.Key() for create key, entityType must be greater than 0
  • GetNames() - Define the phrases by which the search will be carried out
    • Use the Ant.Phrase() method and its overloads to specify names to search for and to define the name type for flexible sorting and scoring.
    • If you are using the extended phrase configuration Ant.Phrase(string phrase, byte phraseType), use phraseType > 0
  • GetLinks() - Identify the keys that are associated with your entity (when creating an index, the entity will be a child of each element in the Links list)

Build index

Building

• Use Ant.Build() to build Anthill index isntance, passing the normalizer and splitter instances to the method, as well as an enumeration entities for searching
• Also, you can get an instance of the builder using Ant.GetBuilder(INormalizer normalizer, IPhraseSplitter phraseSplitter, HierarchySettings? settings = null)
  • Setup HierarchySettings. Declare type dependencies to containers or parents to use SearchBy and AppendChilds
  • Call builder AddEntity method to add entity (multithreading is not working)
  • Call Build to get Anthill instance

Save

• Use Ant.WriteIndex(AntHill index, string filePath) for saving index file
• Use Ant.ReadIndex(string filePath) for read index file

Search

*Create index searcher

Implement abstract AntSearcherBase

• Implement AntSearcherBase.Request property (array of requests to search) to configure your search request. The query options are presented below (use in the same order for proper operation)
  • Search - search current type entities
  • SearchBy - search current type entities in parents hierarchy (parents must be found in the Search block above)
  • Select - performs forced addition of entities of the target type based on the passed ids
  • AppendChilds - use to force adding entities by parent
• Override GetLinkedEntityMatchMiltipler(byte entityType, byte linkedType) for flexible scoring mathes of linked entities
• Override GetPhraseTypeMultipler(byte phraseType) for flexible scoring by phrase types
• Override OnLinkedEntityMatched(Key entityKey, Key linkedKey) to add individual sorting rules if linked entity is match
• Override OnEntityProcessed(EntityMatchesBundle entityMatchesBundle) to add individual sorting rules
• Override ResultVisionFilter(byte type, IEnumerable<EntityMatchesBundle> result) to filter the results that will be output as a result

Search

• Use AntSearcherBase method Search passing the search context
• Or use AntSearcherBase method SearchTypes passing the search context and list of target entities with their count

Optimizations

• If your entity cannot be found if the hierarchy parent is not found, be sure to set the IIndexedEntity.GetContainer method to improve performance.
• If you using overrides OnLinkedEntityMatched or OnEntityProcessed use a static AdditionalRule intances for smaller memory allocations