How to implement a new backend¶

Index¶

Subclass semlix.index.Index.
Indexes must implement the following methods.
Indexes that require/support locking must implement the following methods.
- semlix.index.Index.lock()
- semlix.index.Index.unlock()
Indexes that support deletion must implement the following methods.
- semlix.index.Index.delete_document()
- semlix.index.Index.doc_count_all() – if the backend has delayed deletion.
Indexes that require/support versioning/transactions may implement the following methods.
Index may implement the following methods (the base class’s versions are no-ops).
- semlix.index.Index.optimize()
- semlix.index.Index.close()

Subclass semlix.writing.IndexWriter.
IndexWriters must implement the following methods.
- semlix.writing.IndexWriter.add_document()
- semlix.writing.IndexWriter.add_reader()
Backends that support deletion must implement the following methods.
- semlix.writing.IndexWriter.delete_document()
IndexWriters that work as transactions must implement the following methods.
- semlix.reading.IndexWriter.commit() – Save the additions/deletions done with this IndexWriter to the main index, and release any resources used by the IndexWriter.
- semlix.reading.IndexWriter.cancel() – Throw away any additions/deletions done with this IndexWriter, and release any resources used by the IndexWriter.

Subclass semlix.reading.IndexReader.
IndexReaders must implement the following methods.
- semlix.reading.IndexReader.__contains__()
- semlix.reading.IndexReader.__iter__()
- semlix.reading.IndexReader.iter_from()
- semlix.reading.IndexReader.stored_fields()
- semlix.reading.IndexReader.doc_count_all()
- semlix.reading.IndexReader.doc_count()
- semlix.reading.IndexReader.doc_field_length()
- semlix.reading.IndexReader.field_length()
- semlix.reading.IndexReader.max_field_length()
- semlix.reading.IndexReader.postings()
- semlix.reading.IndexReader.has_vector()
- semlix.reading.IndexReader.vector()
- semlix.reading.IndexReader.doc_frequency()
- semlix.reading.IndexReader.frequency()
Backends that support deleting documents should implement the following methods.
- semlix.reading.IndexReader.has_deletions()
- semlix.reading.IndexReader.is_deleted()
Backends that support versioning should implement the following methods.
- semlix.reading.IndexReader.generation()
If the IndexReader object does not keep the schema in the self.schema attribute, it needs to override the following methods.
- semlix.reading.IndexReader.field()
- semlix.reading.IndexReader.field_names()
- semlix.reading.IndexReader.scorable_names()
- semlix.reading.IndexReader.vector_names()
IndexReaders may implement the following methods.
- semlix.reading.DocReader.close() – closes any open resources associated with the reader.

The semlix.reading.IndexReader.postings() method returns a semlix.matching.Matcher object. You will probably need to implement a custom Matcher class for reading from your posting lists.