Extensions and Utilities

Loading…

Extensions and Utilities

Relevant source files

• extensions/Cargo.toml
• simd-r-drive-entry-handle/src/constants.rs
• simd-r-drive-entry-handle/src/lib.rs
• src/utils.rs
• src/utils/align_or_copy.rs
• src/utils/verify_file_existence.rs
• tests/align_or_copy_tests.rs

This document covers the utility functions, helper modules, and constants provided by the SIMD R Drive ecosystem. These components include the simd-r-drive-extensions crate for higher-level storage operations, core utility functions in the main simd-r-drive crate, and shared constants from simd-r-drive-entry-handle.

For details on the core storage engine API, see DataStore API. For performance optimization features like SIMD acceleration, see SIMD Acceleration. For alignment-related architecture decisions, see Payload Alignment and Cache Efficiency.

Extensions Crate Overview

The simd-r-drive-extensions crate provides storage extensions and higher-level utilities built on top of the core simd-r-drive storage engine. It adds functionality for common storage patterns and data manipulation tasks.

graph TB
    subgraph "simd-r-drive-extensions"
        ExtCrate["simd-r-drive-extensions"]
ExtDeps["Dependencies:\n- bincode\n- serde\n- simd-r-drive\n- walkdir"]
end
    
    subgraph "Core Dependencies"
        Core["simd-r-drive"]
Bincode["bincode\nBinary Serialization"]
Serde["serde\nSerialization Traits"]
Walkdir["walkdir\nDirectory Traversal"]
end
    
 
   ExtCrate --> ExtDeps
 
   ExtDeps --> Core
 
   ExtDeps --> Bincode
 
   ExtDeps --> Serde
 
   ExtDeps --> Walkdir
    
 
   Core -.->|provides| DataStore["DataStore"]
Bincode -.->|enables| SerializationSupport["Structured Data Storage"]
Walkdir -.->|enables| FileSystemOps["File System Operations"]

Crate Structure

Sources: extensions/Cargo.toml:1-22

Sources: extensions/Cargo.toml:13-17

Core Utilities Module

The main simd-r-drive crate exposes several utility functions through its utils module. These functions handle common tasks like alignment optimization, string formatting, and data validation.

graph TB
    subgraph "utils Module"
        UtilsRoot["src/utils.rs"]
AlignOrCopy["align_or_copy\nZero-Copy Optimization"]
AppendExt["append_extension\nString Path Handling"]
FormatBytes["format_bytes\nHuman-Readable Sizes"]
NamespaceHasher["NamespaceHasher\nHierarchical Keys"]
ParseBuffer["parse_buffer_size\nSize String Parsing"]
VerifyFile["verify_file_existence\nFile Validation"]
end
    
 
   UtilsRoot --> AlignOrCopy
 
   UtilsRoot --> AppendExt
 
   UtilsRoot --> FormatBytes
 
   UtilsRoot --> NamespaceHasher
 
   UtilsRoot --> ParseBuffer
 
   UtilsRoot --> VerifyFile
    
 
   AlignOrCopy -.->|used by| ReadOps["Read Operations"]
NamespaceHasher -.->|used by| KeyManagement["Key Management"]
FormatBytes -.->|used by| Logging["Logging & Reporting"]
ParseBuffer -.->|used by| Config["Configuration Parsing"]

Utility Functions Overview

Sources: src/utils.rs:1-17

align_or_copy Function

The align_or_copy utility function provides zero-copy deserialization with automatic fallback for misaligned data. It attempts to reinterpret a byte slice as a typed slice without copying, and falls back to manual decoding when alignment requirements are not met.

Function Signature

Sources: src/utils/align_or_copy.rs:44-50

Operation Flow

Sources: src/utils/align_or_copy.rs:44-73

Usage Patterns

Example Usage:

Sources: src/utils/align_or_copy.rs:38-43 tests/align_or_copy_tests.rs:7-12

Safety Considerations

The function uses unsafe for the align_to::<T>() call, which requires:

1. Starting address must be aligned to align_of::<T>()
2. Total size must be a multiple of size_of::<T>()

These requirements are validated by checking that prefix and suffix slices are empty before returning the borrowed slice. If validation fails, the function falls back to safe manual decoding.

Sources: src/utils/align_or_copy.rs:28-35 src/utils/align_or_copy.rs:53-60

Other Utility Functions

Sources: src/utils.rs:1-17

Entry Handle Constants

The simd-r-drive-entry-handle crate defines shared constants used throughout the storage system. These constants establish the binary layout of entries and alignment requirements.

graph TB
    subgraph "simd-r-drive-entry-handle"
        LibRoot["lib.rs"]
ConstMod["constants.rs"]
EntryHandle["entry_handle.rs"]
EntryMetadata["entry_metadata.rs"]
DebugAssert["debug_assert_aligned.rs"]
end
    
    subgraph "Exported Constants"
        MetadataSize["METADATA_SIZE = 20"]
KeyHashRange["KEY_HASH_RANGE = 0..8"]
PrevOffsetRange["PREV_OFFSET_RANGE = 8..16"]
ChecksumRange["CHECKSUM_RANGE = 16..20"]
ChecksumLen["CHECKSUM_LEN = 4"]
PayloadLog["PAYLOAD_ALIGN_LOG2 = 6"]
PayloadAlign["PAYLOAD_ALIGNMENT = 64"]
end
    
 
   LibRoot --> ConstMod
 
   LibRoot --> EntryHandle
 
   LibRoot --> EntryMetadata
 
   LibRoot --> DebugAssert
    
 
   ConstMod --> MetadataSize
 
   ConstMod --> KeyHashRange
 
   ConstMod --> PrevOffsetRange
 
   ConstMod --> ChecksumRange
 
   ConstMod --> ChecksumLen
 
   ConstMod --> PayloadLog
 
   ConstMod --> PayloadAlign
    
 
   PayloadAlign -.->|ensures| CacheLineOpt["Cache-Line Optimization"]
PayloadAlign -.->|enables| SIMDOps["SIMD Operations"]

Constants Module Structure

Sources: simd-r-drive-entry-handle/src/lib.rs:1-10 simd-r-drive-entry-handle/src/constants.rs:1-19

Metadata Layout Constants

The following constants define the fixed 20-byte metadata structure at the end of each entry:

Sources: simd-r-drive-entry-handle/src/constants.rs:3-11

Alignment Constants

These constants enforce 64-byte alignment for all payload data:

• PAYLOAD_ALIGN_LOG2 : Base-2 logarithm of alignment requirement (6 = 64 bytes)
• PAYLOAD_ALIGNMENT : Computed alignment value (64 bytes)

This alignment matches CPU cache line sizes and enables efficient SIMD operations. The maximum pre-padding per entry is PAYLOAD_ALIGNMENT - 1 (63 bytes).

Sources: simd-r-drive-entry-handle/src/constants.rs:13-18

Constant Relationships

Sources: simd-r-drive-entry-handle/src/constants.rs:1-19

sequenceDiagram
    participant Client
    participant EntryHandle
    participant align_or_copy
    participant Memory
    
    Client->>EntryHandle: get_payload_bytes()
    EntryHandle->>Memory: read &[u8] from mmap
    EntryHandle->>align_or_copy: align_or_copy<f32, 4>(bytes, f32::from_le_bytes)
    
    alt Aligned on 64-byte boundary
        align_or_copy->>Memory: validate alignment
        align_or_copy-->>Client: Cow::Borrowed(&[f32])
        Note over Client,Memory: Zero-copy: direct memory access\nelse Misaligned
        align_or_copy->>align_or_copy: chunks_exact(4)
        align_or_copy->>align_or_copy: map(f32::from_le_bytes)
        align_or_copy->>align_or_copy: collect into Vec<f32>
        align_or_copy-->>Client: Cow::Owned(Vec<f32>)
        Note over Client,align_or_copy: Fallback: allocated copy
    end

Common Patterns

Zero-Copy Data Access

Utilities like align_or_copy enable zero-copy access patterns when memory alignment allows:

Sources: src/utils/align_or_copy.rs:44-73 simd-r-drive-entry-handle/src/constants.rs:13-18

Namespace-Based Key Management

The NamespaceHasher utility enables hierarchical key organization:

Sources: src/utils.rs:11-12

Size Formatting for Logging

The format_bytes utility provides human-readable output:

Sources: src/utils.rs:7-8

Configuration Parsing

The parse_buffer_size utility handles size string inputs:

Sources: src/utils.rs:13-14

Integration with Core Systems

Relationship to Storage Engine

Sources: extensions/Cargo.toml:1-22 src/utils.rs:1-17 simd-r-drive-entry-handle/src/lib.rs:1-10

Performance Considerations

Sources: src/utils/align_or_copy.rs:1-74 simd-r-drive-entry-handle/src/constants.rs:13-18

Dismiss

Refresh this wiki

Enter email to refresh