wf-connector-api 0.2.0

Minimal Arrow-native connector API for warp-fusion — BatchSource trait with RecordBatch support
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133

//! # wf-connector-api
//!
//! Minimal Arrow-native connector API for warp-fusion.
//!
//! ## Design
//!
//! `wp-connector-api` sources produce `SourceEvent { payload: RawData }`,
//! designed for downstream parse pipelines. CEP engines like warp-fusion
//! operate on Arrow `RecordBatch` directly.
//!
//! `wf-connector-api` fills this gap for Arrow-native source consumption.
//! (Sink output uses the existing `wp-connector-api` `SinkRuntime` —
//! adding `send_batch()` to it is sufficient, no new trait needed.)
//!
//! ## Relationship with `wp-connector-api`
//!
//! | | wp-connector-api | wf-connector-api |
//! |---|---|---|
//! | Source data | `SourceEvent { payload: RawData }` | `RecordBatch` (columnar) |
//! | Consumer | parse pipeline (WPL) | CEP engine (warp-fusion) |
//! | Error model | `SourceResult<T>` (orion-error) | `SourceResult<T>` (orion-error) |
//! | Lifecycle | `start()` / `receive()` / `close()` | `start()` / `receive_batch()` / `close()` |
//!
//! `wp-connectors` (the implementation crate) can implement BOTH traits
//! for the same connector (Kafka / File / TCP), sharing connection logic.

use arrow::record_batch::RecordBatch;
use async_trait::async_trait;
use orion_error::conversion::ToStructError;
use orion_error::{OrionError, StructError, UnifiedReason};
use std::error::Error as StdError;

// -- Error -------------------------------------------------------------------

/// Connector error reason.
///
/// All leaf variants carry detail via `err_detail()`. `SourceError` wraps
/// each variant with a detail string and optional source error.
#[derive(Debug, Clone, PartialEq, OrionError)]
pub enum SourceReason {
    /// End of stream — no more data will be produced.
    #[orion_error(message = "end of stream", identity = "sys.wf_connector.eof")]
    EOF,
    /// No data currently available (not EOF); caller should retry.
    #[orion_error(message = "no data available", identity = "sys.wf_connector.not_data")]
    NotData,
    /// I/O error from the underlying transport.
    #[orion_error(message = "I/O error", identity = "sys.wf_connector.io")]
    Io,
    /// Failed to establish connection / bind / subscribe.
    #[orion_error(message = "connection error", identity = "sys.wf_connector.connect")]
    Connect,
    /// Message / frame decoding failed.
    #[orion_error(message = "decode error", identity = "sys.wf_connector.decode")]
    Decode,
    /// Referenced connector not found in registry.
    #[orion_error(
        message = "connector not found",
        identity = "sys.wf_connector.not_found"
    )]
    NotFound,
    /// Catch-all for unexpected errors.
    #[orion_error(transparent)]
    General(UnifiedReason),
}

impl SourceReason {
    /// Create an error with detail message.
    pub fn err_detail<S: Into<String>>(self, detail: S) -> SourceError {
        self.to_err().with_detail(detail.into())
    }

    /// Create an error with a source (chained) error.
    pub fn err_source<E>(self, source: E) -> SourceError
    where
        E: StdError + Send + Sync + 'static,
    {
        self.to_err().with_source(source)
    }
}

pub type SourceError = StructError<SourceReason>;
pub type SourceResult<T> = Result<T, SourceError>;

// -- Source ------------------------------------------------------------------

/// A batch-oriented data source that produces Arrow [`RecordBatch`]es.
///
/// # Lifecycle
///
/// 1. `start()` — initialize (connect, subscribe, bind)
/// 2. `receive_batch()` — pull data in a loop
/// 3. `close()` — release resources (unsubscribe, close connections)
///
/// `close()` must be idempotent — safe to call multiple times, even before `start()`.
///
/// # Empty vs EOF
///
/// - Return `Ok(vec![])` when no data is currently available (caller should retry).
/// - Return `Err(SourceReason::EOF.into())` when the stream has ended.
#[async_trait]
pub trait BatchSource: Send {
    /// Initialize the source. Called once before the first `receive_batch()`.
    ///
    /// Default is a no-op.
    async fn start(&mut self) -> SourceResult<()> {
        Ok(())
    }

    /// Receive zero or more [`RecordBatch`]es.
    ///
    /// An empty `Vec` means "no data right now" — the caller should poll again.
    /// An error with `SourceReason::EOF` means the stream has ended.
    async fn receive_batch(&mut self) -> SourceResult<Vec<RecordBatch>>;

    /// Close the source and release all resources.
    ///
    /// Must be idempotent — safe to call multiple times or before `start()`.
    /// Default is a no-op.
    async fn close(&mut self) -> SourceResult<()> {
        Ok(())
    }

    /// Unique identifier for this source instance (logging / metrics).
    fn identifier(&self) -> &str;
}

// -- Sink (not needed as a separate trait) -----------------------------------
//
// Arrow-native sink output is handled by the existing `wp-connector-api`
// `SinkRuntime`. Adding a `send_batch()` method to `SinkRuntime` (which
// already has `send_record()`) is sufficient — no new trait required.
// File / Arrow IPC / TCP backends can natively accept RecordBatch.