Crate re_arrow_store
source ·Expand description
The Rerun datastore, implemented on top of Apache Arrow
using the arrow2 crate.
This crate is an in-memory time series database for Rerun log data.
It is indexed by Entity path, component, timeline, and time.
It supports out-of-order insertions, and fast O(log(N)) queries.
- See
DataStorefor an overview of the core data structures. - See
DataStore::latest_atandDataStore::rangefor the documentation of the public read APIs. - See
DataStore::insertfor the documentation of the public write APIs.
Feature flags
-
deadlock_detection— Enablesparking_lot’s deadlock detection background thread. -
polars— Integration withpolars, to efficiently use the datastore with dataframes.
Re-exports
pub use arrow2::io::ipc::read::StreamReader;pub use arrow2::io::ipc::read::StreamState;pub use re_log_types::TimeInt;pub use re_log_types::TimeRange;pub use re_log_types::TimeType;pub use re_log_types::Timeline;
Modules
Structs
- A complete data store: covers all timelines, all entities, everything.
- An
IndexBucketholds a size-delimited (data size and/or number of rows) chunk of aIndexTable. - An
IndexTablemaps specific points in time to rows in component tables. - A query a given time, for a given timeline.
- A query over a time range, for a given timeline.
- An opaque type that directly refers to a row of data within the datastore, iff it is associated with a component name.