Hacktoberfest 2026: le issue che i maintainer hanno segnato per ottobre, aperte e adatte ai principianti. Sfoglia le issue Hacktoberfest

datafusion: no way to set snapshot summary properties on the `insert_into` commit path

Aperta
  • Qualcosa non va?
#3,019 0 commenti 0 reazioni 0 assegnatari Vedi su GitHub

I maintainer di solito rispondono entro 3 giorni

Nessuno ha ancora preso questa issue.

Valutazione

Difficoltà
3/5
Tempo stimato
1-2 giorni
Idoneità per principianti
72/100
Tipo di issue
Funzionalità
Chiarezza
Specificata chiaramente
Stato di attività
Attiva
Stack tecnologico
rust
Ambito
backend-api-design, databases

Direzione di ricerca

Inizia da IcebergTableProvider::insert_into in crates/integrations/datafusion/src/table/mod.rs e dal flusso di commit in crates/integrations/datafusion/src/physical_plan/commit.rs, quindi confronta FastAppendAction::set_snapshot_properties in crates/iceberg/src/transaction/append.rs. Segui il valore del provider attraverso IcebergCommitExec e verifica che un commit INSERT INTO accetti le proprietà, preservando il comportamento di validazione esistente per le chiavi riservate.

Scritto dal modello di indicizzazione a partire dal testo della issue.

Descrizione

The transaction API supports user snapshot properties on an append — FastAppendAction::set_snapshot_properties (crates/iceberg/src/transaction/append.rs:70), carried into the snapshot summary and validated against reserved keys (#2744, #2725 show this surface is maintained). The DataFusion integration cannot reach it: IcebergTableProvider::insert_into (crates/integrations/datafusion/src/table/mod.rs:153) builds IcebergWriteExec + IcebergCommitExec (table/mod.rs:221-226), and the commit node creates the transaction internally —

// crates/integrations/datafusion/src/physical_plan/commit.rs:243-252
let tx = Transaction::new(&table);
let action = tx.fast_append().add_data_files(data_files);
let _updated_table = action
    .apply(tx)
    .map_err(to_datafusion_error)?
    .commit(catalog.as_ref())
    .await?;

— with no hook between building the action and committing it, so the properties parameter that exists one layer down is unreachable from an INSERT INTO.

(Line references are against main @ a500a2e7; the same shape ships in 0.10.x.)

Why this matters

Facts about a write are naturally snapshot summary properties: source scan counts, rows dropped by an ingest filter, cast-failure tallies, a recipe/job id — anything audit- or lineage-shaped that describes this append rather than the table. An embedder who wants them today has to reimplement the integration's whole write path beside the integration — the same RollingFileWriterBuilder + DataFileWriterBuilder calls as physical_plan/write.rs:239, the same fast_append as commit.rs:244 — solely to pass one HashMap the transaction API already accepts. That is what we ended up doing, and the duplicated path has to track every upstream improvement to the real one by hand.

Proposal

The smallest API that closes the gap: a builder-style setter on the provider, threaded through to the commit node —

let provider = IcebergTableProvider::try_new(...)
    .await?
    .with_snapshot_properties(props); // HashMap<String, String>

IcebergCommitExec gains the field and applies it:

let action = tx.fast_append()
    .add_data_files(data_files)
    .set_snapshot_properties(self.snapshot_properties.clone());

Granularity note: properties set this way are per-provider, which is per-statement for embedders that register a provider per statement, and that is the audit use case. A session/statement-option spelling (e.g. iceberg.snapshot-property.*) could layer on later for SQL-level use, but the provider hook alone removes the need to fork the write path.

Reserved-key validation stays where it is — the action already owns it, so an invalid property fails the commit exactly as it does through the transaction API directly.

Happy to send a PR if the shape sounds right.

Lingua principale
Rust
Stelle
1.4k
Fork
612
Merge medio
4g 6h
PR unite (30g)
83

Preparare l'ambiente

Apri in Codespaces

Avvia il container di sviluppo del progetto nel browser, con il tuo account GitHub.

Come iniziare

  1. Leggi tutta la issue e poi la guida ai contributi del progetto.
  2. Commenta sulla issue per dire che te ne occupi tu — evita che due persone facciano lo stesso lavoro.
  3. Fai un fork del repository e lavora su un branch.
  4. Apri una pull request che faccia riferimento al numero della issue.

Altre issue di apache/iceberg-rust

Tutte le issue di apache/iceberg-rust

Issue simili

Altre issue su Rust

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.