Baloo/Architecture

Baloo is a metadata and search framework by KDE.

Baloo focuses on providing a very small memory footprint along with with extremely fast searching. It also supports storing additional file based metadata via extended attributes.

• Mailing List: [email protected]
• IRC Channel: #kde-baloo on freenode

Architecture

 

Outdated architecture
The architecture evolved and Baloo is only a file indexer. The reference to PIM resources (emails, contacts) is historical. KDE PIM uses its own indexing system, which has a similar architecture.

Baloo focuses on decentralization of data and does not have any central database. At its core it is a set of 3 services:

• Data Stores
• Search Stores
• Relations

Baloo replaced Nepomuk, a more elaborate semantic search system and ontology. There may still be information about the older project in techbase::Projects/Nepomuk.

Data Store

A Data Store is a place to permanently store data. Its format and APIs are not dictated by Baloo and can be in any way that is most efficient for the data in question.

Search Store

A Search Store is a plugin which provides search capabilities for a specific kind of data. Each of these stores provide a plugin which can be used to search through the data using a consistent API. At this moment there are 3 search stores that have been implemented -

• File Search
• Email Search
• Contact Search

Each Store can also provide custom APIs for searching the data.

Relations

Baloo can be used to create relations between two uniquely identifiable identifiers. The key aspect is that when creating a relation between A and B, both of them are uniquely identifiable. For most of the PIM data, Akonadi already provides a unique identifier of the 'akonadi:?item=x'. For File data, Baloo provides unique identifiers of the form 'file:x'.

Each relation is stored in an independent store in the desired mechanism. A sample sqlite implementation is provided which maps both the identifiers in a 2 column sqlite database.

For example, there exists a TagRelation which maps a Tag and any unique identifier. Similarly, there can be an ActivityRelation, and relations for other aspects such as mapping a file received via Bluetooth to a Bluetooth device.

Implementation

Files

Baloo concentrates heavily on files and provides both a data store and search store for files. These are stored using a combination of sqlite and xapian.

Akonadi

In the case of PIM, since Akonadi already stores (caches) the data, only a Search Store is required. There are currently separate search stores for contacts and emails. They are both implemented using Xapian.

Tags

Tags are stored in a custom format. In order to relate tags with files / email, a TagRelation is provided.

Baloo and extended attributes

See the Baloo page on the Userbase wiki