PhoenixmlDb

Core Concepts

Containers, documents, the XDM, and shredded node storage

#Core Concepts

Understanding the fundamental concepts of PhoenixmlDb will help you design efficient applications and write optimal queries.

#Architecture Overview

PhoenixmlDb is built on a layered architecture:

┌─────────────────────────────────────────────────────────┐
│                    Application Layer                     │
│         (Your C# Application / gRPC Clients)            │
├─────────────────────────────────────────────────────────┤
│                      Query Layer                         │
│   XQuery Engine │ XSLT Processor │ LINQ Provider        │
├─────────────────────────────────────────────────────────┤
│                    Indexing Layer                        │
│   Path │ Value │ Full-Text │ Structural Indexes         │
├─────────────────────────────────────────────────────────┤
│                  Data Model Layer                        │
│              XQuery Data Model (XDM)                     │
├─────────────────────────────────────────────────────────┤
│                    Storage Layer                         │
│                 LMDB Key-Value Store                     │
└─────────────────────────────────────────────────────────┘

#Key Components

#Database

The top-level container that manages all data. A database maps to a directory on disk containing LMDB data files.

csharp

using var db = new XmlDatabase("./mydata");

#Containers

Logical groupings of related documents within a database. Similar to tables in relational databases or collections in document databases.

csharp

var orders = db.CreateContainer("orders");
var customers = db.CreateContainer("customers");

#Documents

Individual XML or JSON documents stored within containers. Each document has a unique name within its container.

csharp

container.PutDocument("order-001.xml", xmlContent);

#Nodes

The atomic units of the XQuery Data Model. Documents are decomposed into nodes for efficient storage and querying.

#Shredded Node Storage

PhoenixmlDb stores XML and JSON documents using a technique called node shredding: each XDM node (element, attribute, text node, comment, etc.) is serialized into its own LMDB key-value entry rather than storing the document as a serialized string. This design enables:

Fine-grained access — queries can retrieve individual nodes without deserializing entire documents
Partial updates — XQuery Update expressions (e.g., replace value of node) modify only the affected node entries, not the whole document
Efficient indexing — index entries point directly to node-level keys, so the engine fetches only matching nodes

#The Identifier Hierarchy

Every object in PhoenixmlDb is identified by a strongly-typed integer value:

Type	.NET type	Scope
`ContainerId`	`uint`	Unique within a database
`DocumentId`	`ulong`	Unique within a container
`NodeId`	`ulong`	Unique within a document

A node is fully addressed by the triple (ContainerId, DocumentId, NodeId). LMDB keys are constructed from this triple, giving O(log n) lookup for any node in the database.

#Document Reassembly

When a document is retrieved in full, PhoenixmlDb performs a tree traversal over the structural index:

Resolve the DocumentId from the document name.
Fetch the document root node.
Walk the parent-child index recursively to emit child nodes in document order.
Serialize the resulting node sequence back to XML or JSON.

For query results that return only a subtree, the engine stops traversal at the requested root, so only the required nodes are read from LMDB.

#Namespace Interning

Namespace URIs are interned: each unique URI is stored once and assigned a NamespaceId (uint). Node entries reference the NamespaceId rather than the full URI string. This keeps node records compact and makes namespace-aware comparisons a fast integer equality check rather than a string comparison.

#Data Flow

XML Document → Parser → XDM Nodes → Storage
                           ↓
                       Indexing
                           ↓
Query → Parser → AST → Optimizer → Executor → Results