Module trace 

Traces

A “trace” describes how a collection of key-value pairs changes over time. A “batch” is a mostly immutable trace. This module provides traits and structures for expressing traces in DBSP as collections of (key, val, time, diff) tuples.

The base trait for a trace is BatchReader, which allows a trace to be read in sorted order by key and value. BatchReader provides Cursor to step through a batch’s tuples without modifying them.

The Batch trait extends BatchReader with types and methods for creating new traces from ordered tuples (Batch::Builder) or unordered tuples (Batch::Batcher), or by merging traces of like types.

The Trace trait, which also extends BatchReader, adds methods to append new batches. New tuples must not have times earlier than any of the tuples already in the trace.

Time within traces

See the time module documentation for a description of logical times.

Traces are sorted by key and value. They are not sorted with respect to time: reading a trace might obtain out of order and duplicate times among the (time, diff) pairs associated with a key and value.

Re-exports

pub use crate::storage::file::Deserializable;

pub use crate::storage::file::Deserializer;

pub use crate::storage::file::Rkyv;

pub use crate::storage::file::Serializer;

pub use spine_async::BatchReaderWithSnapshot;

pub use spine_async::ListMerger;

pub use spine_async::MergerType;

pub use spine_async::Spine;

pub use spine_async::SpineSnapshot;

pub use spine_async::WithSnapshot;

pub use ord::FallbackIndexedWSet;

pub use ord::FallbackIndexedWSetBuilder;

pub use ord::FallbackIndexedWSetFactories;

pub use ord::FallbackKeyBatch;

pub use ord::FallbackKeyBatchFactories;

pub use ord::FallbackValBatch;

pub use ord::FallbackValBatchFactories;

pub use ord::FallbackWSet;

pub use ord::FallbackWSetBuilder;

pub use ord::FallbackWSetFactories;

pub use ord::FileIndexedWSet;

pub use ord::FileIndexedWSetFactories;

pub use ord::FileKeyBatch;

pub use ord::FileKeyBatchFactories;

pub use ord::FileValBatch;

pub use ord::FileValBatchFactories;

pub use ord::FileWSet;

pub use ord::FileWSetFactories;

pub use ord::OrdIndexedWSet;

pub use ord::OrdIndexedWSetBuilder;

pub use ord::OrdIndexedWSetFactories;

pub use ord::OrdKeyBatch;

pub use ord::OrdKeyBatchFactories;

pub use ord::OrdValBatch;

pub use ord::OrdValBatchFactories;

pub use ord::OrdWSet;

pub use ord::OrdWSetBuilder;

pub use ord::OrdWSetFactories;

pub use ord::VecIndexedWSet;

pub use ord::VecIndexedWSetFactories;

pub use ord::VecKeyBatch;

pub use ord::VecKeyBatchFactories;

pub use ord::VecValBatch;

pub use ord::VecValBatchFactories;

pub use ord::VecWSet;

pub use ord::VecWSetFactories;

pub use cursor::Cursor;

pub use cursor::MergeCursor;

pub use filter::Filter;

pub use filter::GroupFilter;

pub use layers::Trie;

Modules

cursor

Traits and types for navigating ordered sequences of (key, val, time, diff) tuples.

filter

Filtering predicates for traces.

layers

ord

spine_async

Implementation of a Trace which merges batches in the background.

Structs

TupleBuilder

Batch builder that accepts a full tuple at a time.

Enums

BatchLocation

Where a batch is stored.

Traits

Batch

A BatchReader plus features for constructing new batches.

BatchFactories

BatchReader

A set of (key, value, time, diff) tuples whose contents may be read in order by key and value.

BatchReaderFactories

Batcher

Functionality for collecting and batching updates.

Builder

Functionality for building batches from ordered update sequences.

DBData

Trait for data stored in batches.

DBWeight

Trait for data types used as weights.

Trace

A set of (key, val, time, diff) tuples that can be read and extended.

Functions

deserialize_indexed_wset

eq_batch

Compares two batches for equality. This works regardless of whether the batches are the same type, as long as their key, value, and weight types can be compared for equality.

merge_batches

Merges all of the batches in batches, applying key_filter and value_filter, and returns the merged result.

merge_batches_by_reference

Merges all of the batches in batches, applying key_filter and value_filter, and returns the merged result.

serialize_indexed_wset

unaligned_deserialize

Trait for data that can be serialized and deserialized with rkyv.

Type Aliases

WeightedItem