JCR (Java Content Repository) is a standardized API for accessing content repositories. Specifications:

• JCR 1.0 — JSR-170 (2004)
• JCR 2.0 — JSR-283 (2009)

The implementation used in Sling is Apache Jackrabbit Oak.

• Functional data: application domain-specific content — articles, blog posts, comments, assets
• Structured data: address records, spreadsheet-like data
• Unstructured data: PDFs, Word Docs, XML, CSVs
• Other: workflow definitions, ACLs, code

JCR uses a hierarchical content model:

1. Repository — top-level container
2. Workspace — logical partition (default: "crx.default", typically only one used)
3. Node — hierarchical tree structure (like folders and files)
4. Property — holds data on a node
5. Session — opened/closed per request. Should be short-lived, never shared across threads.

Every node has a Primary Nodetype which defines mandatory/optional properties, data types, and subnodes:

• nt:unstructured — free data structure, no constraints (default)
• nt:folder — comparable to file system folder
• nt:file + nt:resource — files/binary data

Additionally, a node can have any number of Mixins (e.g. mix:title, mix:referenceable).

Properties hold the data of nodes. Supported data types:

• STRING, LONG, DOUBLE, DECIMAL, BOOLEAN, DATE, BINARY, REFERENCE
• All types support multi-valued variants (arrays): STRING[], LONG[], etc.

Rule of thumb: don't use it. "Content first, structure later, maybe."

• Exception: Use built-in JCR node types where appropriate (especially for files)
• For technical reasons, querying by nodetype is very efficient
• Often existing nodetypes are sufficient

JCR:

• nt:base — base type for all primary node types
• nt:unstructured — most flexible, no constraints
• nt:folder — represents folders
• nt:file — represents files (requires jcr:content child)
• nt:resource — holds binary data (requires jcr:data)

Sling:

• sling:OrderedFolder — default for folder structures
• sling:OsgiConfig — OSGi config stored in repository
• sling:Mapping — URL mappings

Main package: javax.jcr. The API defines different support levels:

• Level 1 — read access, export
• Level 2 — write access, import
• Advanced — observation, locking, versioning, access control

Jackrabbit Oak supports all levels/all features.

Note: The API is a bit outdated (designed 2002) — makes heavy use of checked exceptions and is cumbersome compared to modern Java.

Rule of thumb: don't use it. All normal read/write operations should use the Sling Resource API.

Use JCR API only for features not supported by Sling:

• Versioning and locking nodes
• Managing security and ACLs
• Observation use cases not supported by Sling eventing
• Copy or move whole subtrees
• Queries with special features like limiting