Difference between revisions of "Storage Layout"
James Ball (talk | contribs) (Added information regarding pth.basex and idp.basex) |
|||
Line 93: | Line 93: | ||
For a more detailed discussion and examples of these file formats please see [[Index File Structure]]. | For a more detailed discussion and examples of these file formats please see [[Index File Structure]]. | ||
+ | |||
+ | ==Document Path Index: {{Code|pth}}== | ||
+ | |||
+ | Provides an index of all the document paths in the database. For databases with a large number of paths this file can be quite large so it is only generated the first time a function requesting a path lookup is run. For databases where path lookups are never used this file will not exist. | ||
+ | |||
+ | '''Note:''' On Windows/Mac systems this file is case insensitive (all paths are lower case). On UNIX-like systems this file is case sensitive. The behaviour of path look ups will vary between systems. Copying this file between system types may lead to unexpected behaviour. | ||
+ | |||
+ | ==ID/Pre Mapping: {{Code|idp}}== | ||
+ | |||
+ | This file is only created if incremental indexing (UPDINDEX) is enabled for a database. It is used to provide a quick look up of the pre value for a database node id. | ||
==Full-Text Fuzzy Index: {{Code|ftxx}}, {{Code|ftxy}}, {{Code|ftxz}}== | ==Full-Text Fuzzy Index: {{Code|ftxx}}, {{Code|ftxy}}, {{Code|ftxz}}== | ||
...may soon be reimplemented. | ...may soon be reimplemented. |
Revision as of 23:52, 12 December 2020
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.
Contents
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 → Element Name Index• ATTS → Attribute Name Index• PATH → Path Index• NS → Namespaces• DOCS → Document Index2. Empty key as finalizer |
DiskData() |
2 a) Name Index Element/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 the node 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
, the Token Index in tokl
and tokr
. All have the same layout.
For a more detailed discussion and examples of these file formats please see Index File Structure.
Document Path Index: pth
Provides an index of all the document paths in the database. For databases with a large number of paths this file can be quite large so it is only generated the first time a function requesting a path lookup is run. For databases where path lookups are never used this file will not exist.
Note: On Windows/Mac systems this file is case insensitive (all paths are lower case). On UNIX-like systems this file is case sensitive. The behaviour of path look ups will vary between systems. Copying this file between system types may lead to unexpected behaviour.
ID/Pre Mapping: idp
This file is only created if incremental indexing (UPDINDEX) is enabled for a database. It is used to provide a quick look up of the pre value for a database node id.
Full-Text Fuzzy Index: ftxx
, ftxy
, ftxz
...may soon be reimplemented.