Storage Layout
Revision as of 03:18, 6 April 2012 by CG (talk | contribs) (→Node Table: {{Mono|tbl}}, {{Mono|tbli}})
This article is part of the Advanced User's Guide. It presents some low-level details on how data is stored in the database files.
Data Types
The following data types are used for specifying the storage layout:
| Type | Description | Example (native → hex integers) |
|---|---|---|
Num
|
Compressed integer (1-5 bytes), specified in Num.java | 15 → 0F; 511 → 41 FF |
Token
|
Length (Num) and bytes of UTF8 byte representation
|
Hello → 05 48 65 6c 6c 6f
|
Double
|
Number, stored as token | 123 → 03 31 32 33
|
Boolean
|
Boolean (1 byte, 00 or 01)
|
true → 01
|
Nums, Tokens, Doubles
|
Arrays of values, introduced with the number of entries | 1,2 → 02 01 31 01 32
|
TokenSet
|
Key array (Tokens), next/bucket/size arrays (3x Nums)
|
Database Files
The following tables illustrate the layout of the BaseX database files. All files are suffixed with .basex.
Meta Data, Name/Path/Doc Indexes: inf
| Description | Format | Method |
|---|---|---|
| 1. Meta Data | 1. Key/value pairs, in no particular order (Token/Token):• Examples: FNAME, TIME, SIZE, ...• PERM → Number of users (Num), and name/password/permission values for each user (Token/Token/Num)2. Empty key as finalizer |
DiskData() MetaData() Users() |
| 2. Main memory indexes | 1. Key/value pairs, in no particular order (Token/Token):• TAGS → Tag Index• ATTS → Attribute Name Index• PATH → Path Index• NS → Namespaces• DOCS → Document Index2. Empty key as finalizer |
DiskData() |
| 2 a) Name Index Tag/attribute names |
1. Token set, storing all names (TokenSet)2. One StatsKey instance per entry: 2.1. Content kind ( Num):2.1.1. Number: min/max ( Doubles)2.1.2. Category: number of entries ( Num), entries (Tokens)2.2. Number of entries ( Num)2.3. Leaf flag ( Boolean)2.4. Maximum text length ( Double; legacy, could be Num)
|
Names() TokenSet.read() StatsKey() |
| 2 b) Path Index | 1. Flag for path definition (Boolean, always true; legacy)2. PathNode: 2.1. Name reference ( Num)2.2. Node kind ( Num)2.3. Number of occurrences ( Num)2.4. Number of children ( Num)2.5. Double; legacy, can be reused or discarded2.6. Recursive generation of child nodes (→ 2) |
PathSummary() PathNode() |
| 2 c) Namespaces | 1. Token set, storing prefixes (TokenSet)2. Token set, storing URIs ( TokenSet)3. NSNode: 3.1. pre value ( Num)3.2. References to prefix/URI pairs ( Nums)3.3. Number of children ( Num)3.4. Recursive generation of child nodes (→ 3) |
Namespaces() NSNode() |
| 2 d) Document Index | Array of integers, representing the distances between all document pre values (Nums)
|
DocIndex() |
Node Table: tbl, tbli
tbl: Main database table, stored in blocks.tbli: Database directory, organizing the database blocks.
Some more information on Node Table Storage is available.
Texts: txt, atv
txt: Heap file for text values (document names, string values of texts, comments and processing instructions)atv: Heap file for attribute values.
Value Indexes: txtl, txtr, atvl, atvr
Text Index:
txtl: Heap file with ID lists.txtr: Index file with references to ID lists.
The Attribute Index is contained in the files atvl and atvr; it uses the same layout.
Full-Text Fuzzy Index: ftxx, ftxy, ftxz
...will soon be reimplemented.
Full-Text Trie Index: ftxa, ftxb, ftxc
...will soon be dismissed.
