Indexes

This article is part of the Query Portal and introduces various index structures, which may speed up many queries by orders of magnitudes.

Index Structures

Currently, the following index structures exist in BaseX:

Tag/Attribute Name Index

All tags and attribute names are automatically indexed and enriched with statistical information.

Path Summary

Unique paths in a document or collection are referenced by the path index, which is applied e.g. to rewrite descendant to more specific child steps.

Text Index

This index speeds up equality tests on text nodes in XPath predicates.

Attribute Index

This index speeds up equality tests on attribute value in XPath predicates.

Full-Text Index

Full-Text indexes are very powerful structures, which are mainly used in information retrieval use cases.

With Template:Mark, a new database option was introduced to support incremental indexing of texts and attributes.

Example Queries

The following queries will be optimized for index access (provided that the relevant index exists in a particular database):

Name/Path Index

• //name may be rewritten to /addressbook/address/name
• /non-existing-name may be rewritten to an empty sequence

Text Index

• //node()[text() = 'Usability']
• //div[p = 'Usability' or p = 'Testing']
• path/to/relevant[text() = 'Usability Testing']/and/so/on

Attribute Index

• //node()[@align = 'right']
• descendant::elem[@id = '1']
• range/query[@id >= 1 and @id <= 5]

Full-Text Index

• //node[text() contains text 'Usability']
• //node[text() contains text 'Usebiliti' using fuzzy]
• //book[chapter contains text ('web' ftor 'WWW' using no stemming) ftand 'diversity' using stemming distance at most 5 words]

Index data structures

Text/Attribute Index

Both the text and attribute index are based on a balanced B-Tree and support exact matches and range queries.

Full-Text Index (Standard)

The standard full-text index is implemented as sorted array structure. It is optimized for simple and fuzzy searches.

Full-Text Index (Wildcards enabled)

A second full-text index is implemented as a compressed trie. It needs slightly more memory than the standard full-text index, but it supports more features, such as full wildcard search.