drasi_core/interface/

index_backend.rs

// Copyright 2025 The Drasi Authors.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
//     http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

//! Index Backend Plugin Trait
//!
//! This module defines the `IndexBackendPlugin` trait that external index backends
//! (like RocksDB, Garnet/Redis) must implement to integrate with Drasi.
//!
//! # Architecture
//!
//! The index plugin system follows pure dependency inversion:
//! - **Core** provides index traits (`ElementIndex`, `ResultIndex`, etc.) and a default
//!   in-memory implementation
//! - **Lib** uses this plugin trait but has no knowledge of specific implementations
//! - **External plugins** (in `components/indexes/`) implement this trait
//! - **Applications** optionally inject plugins; if none provided, the in-memory default is used

use async_trait::async_trait;
use std::fmt;
use std::sync::Arc;

use super::{
    CheckpointStore, ElementArchiveIndex, ElementIndex, FutureQueue, IndexError, LiveResultsWriter,
    OutboxWriter, ResultIndex, SessionControl,
};

/// Set of indexes for a query.
///
/// Groups the index types and session control needed for query evaluation into
/// a single unit.
/// This enables backends to create all indexes from a shared underlying resource
/// (e.g., a single RocksDB instance or Redis connection).
pub struct IndexSet {
    /// Element index for storing graph elements
    pub element_index: Arc<dyn ElementIndex>,
    /// Archive index for storing historical elements (for past() function)
    pub archive_index: Arc<dyn ElementArchiveIndex>,
    /// Result index for storing query results
    pub result_index: Arc<dyn ResultIndex>,
    /// Future queue for temporal queries
    pub future_queue: Arc<dyn FutureQueue>,
    /// Session control for atomic transaction lifecycle
    pub session_control: Arc<dyn SessionControl>,
}

impl fmt::Debug for IndexSet {
    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
        f.debug_struct("IndexSet")
            .field("element_index", &"<trait object>")
            .field("archive_index", &"<trait object>")
            .field("result_index", &"<trait object>")
            .field("future_queue", &"<trait object>")
            .field("session_control", &"<trait object>")
            .finish()
    }
}

/// Result of [`IndexBackendPlugin::create_indexes`].
///
/// Bundles the [`IndexSet`] together with an optional [`CheckpointStore`]
/// that shares the same underlying session state. Persistent backends return
/// `Some(store)`; volatile (in-memory) backends return `None`.
///
/// The store's `stage_checkpoint` calls land in the same database transaction
/// as index updates because both are derived from the same `SessionControl` /
/// session state instance — that is why the plugin returns them together
/// rather than via two separate calls.
pub struct CreatedIndexes {
    /// The set of indexes for the query.
    pub set: IndexSet,
    /// Atomic checkpoint store paired with the set's session state.
    /// `None` for volatile backends (no persistent storage to checkpoint into).
    pub checkpoint_store: Option<Arc<dyn CheckpointStore>>,
    /// Outbox writer for persisting query results for reaction replay.
    /// `None` for volatile backends or when outbox persistence is not needed.
    pub outbox_writer: Option<Arc<dyn OutboxWriter>>,
    /// Live results writer for persisting the current result snapshot.
    /// `None` for volatile backends or when live result persistence is not needed.
    pub live_results_writer: Option<Arc<dyn LiveResultsWriter>>,
}

impl fmt::Debug for CreatedIndexes {
    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
        f.debug_struct("CreatedIndexes")
            .field("set", &self.set)
            .field(
                "checkpoint_store",
                &self.checkpoint_store.as_ref().map(|_| "<trait object>"),
            )
            .field(
                "outbox_writer",
                &self.outbox_writer.as_ref().map(|_| "<trait object>"),
            )
            .field(
                "live_results_writer",
                &self.live_results_writer.as_ref().map(|_| "<trait object>"),
            )
            .finish()
    }
}

/// Plugin trait for external index storage backends.
///
/// Each storage backend (RocksDB, Garnet, etc.) implements this trait to provide
/// all index types needed for query evaluation from a single shared backend instance.
///
/// # Thread Safety
///
/// Implementations must be `Send + Sync` to allow use across async tasks.
///
/// # Example
///
/// ```ignore
/// use drasi_core::interface::{CreatedIndexes, IndexBackendPlugin, IndexError};
///
/// pub struct MyIndexProvider {
///     // configuration fields
/// }
///
/// #[async_trait]
/// impl IndexBackendPlugin for MyIndexProvider {
///     async fn create_indexes(&self, query_id: &str) -> Result<CreatedIndexes, IndexError> {
///         // Create and return all indexes (and an optional checkpoint store)
///         // from a shared backend instance
///     }
///     fn is_volatile(&self) -> bool { false }
/// }
/// ```
#[async_trait]
pub trait IndexBackendPlugin: Send + Sync {
    /// Create all indexes (and an optional checkpoint store) for a query
    /// from a single shared backend instance.
    ///
    /// This method creates the element index, archive index, result index,
    /// future queue, and session control backed by a shared storage resource
    /// (e.g., a single RocksDB database or Redis connection). This reduces
    /// resource overhead and enables cross-index atomic transactions.
    ///
    /// Persistent backends additionally return a [`CheckpointStore`] that
    /// shares the same session state as the returned `SessionControl`, so
    /// `stage_checkpoint` writes land in the same database transaction as
    /// index updates. Volatile backends return `checkpoint_store: None`.
    async fn create_indexes(&self, query_id: &str) -> Result<CreatedIndexes, IndexError>;

    /// Returns true if this backend is volatile (data lost on restart).
    ///
    /// Volatile backends (like in-memory) require re-bootstrapping after restart,
    /// while persistent backends (like RocksDB) retain data.
    fn is_volatile(&self) -> bool;
}