Compatibility matrix

Apache Iceberg compatibility matrix: engines and catalogs

Last verified September 30, 2026against each product's official documentation and release notes, Iceberg 1.11.0

"Supports Iceberg" covers a wide range. This matrix breaks it into the features that decide whether two systems can share a table, and cites the page each answer comes from. Select any cell for the detail and its source.

Yes: Supported, per the product's own documentation or release notes.
Partial: Supported with significant documented limits; the note says which.
Read only: Can read this feature but not write or create it.
No: The official documentation states it is not supported.
Unknown: No official statement found. Not a guess either way.

Query engines

What each engine can do with an Iceberg table. Row-level deletes are split into the three ways Iceberg records them: position delete files, equality delete files, and format v3 deletion vectors.

Query engines: Apache Iceberg feature support, verified September 30, 2026
Engine	Read	Write	Position deletes	Equality deletes	Deletion vectors	Format v3	Branches and tags	REST catalog	Iceberg views	Time travel
Apache Spark	Yes	Yes	Yes	Read only	Yes	Partial	Yes	Yes	Yes	Yes
Apache Flink	Yes	Yes	Unknown	Partial	Partial	Partial	Partial	Yes	Unknown	Yes
Trino	Yes	Yes	Yes	Read only	Partial	Partial	Partial	Yes	Yes	Yes
Dremio	Yes	Yes	Yes	Read only	Yes	Yes	No	Yes	Yes	Yes
DuckDB	Yes	Yes	Yes	Unknown	Yes	Partial	Unknown	Yes	Unknown	Yes
Snowflake	Yes	Yes	Partial	No	Yes	Partial	No	Yes	Unknown	Yes
BigQuery	Yes	Partial	Read only	Read only	Read only	Partial	Unknown	Partial	No	Yes
Amazon Athena	Yes	Yes	Yes	Read only	No	No	Unknown	No	No	Yes
StarRocks	Yes	Yes	Yes	Read only	Unknown	Partial	Yes	Yes	Yes	Yes
PyIceberg	Yes	Yes	Read only	No	Read only	Partial	Yes	Yes	Partial	Yes

Read: Reads Iceberg tables.
Write: Writes to Iceberg tables (INSERT or append at minimum).
Position deletes: Merge-on-read position delete files, read and/or write.
Equality deletes: Equality delete files, read and/or write.
Deletion vectors: Format v3 deletion vectors stored in Puffin files.
Format v3: Format version 3 tables and their features (row lineage, variant, default values, geo types, nanosecond timestamps, deletion vectors).
Branches and tags: Create, query, and write to Iceberg branches and tags.
REST catalog: Connects to an Iceberg REST catalog.
Iceberg views: Reads and/or creates views stored in the Iceberg view format.
Time travel: Queries a table as of a snapshot ID or timestamp.

Catalogs

What each catalog offers the engines that connect to it. The first row is the Iceberg REST specification itself, which defines what a REST catalog may offer; each implementation chooses which parts to serve.

Catalogs: Apache Iceberg feature support, verified September 30, 2026
Catalog	Serves REST API	Iceberg views	Format v3 tables	Credential vending	Multi-table commits
REST spec	Yes	Yes	Yes	Yes	Yes
Apache Polaris	Yes	Yes	Yes	Yes	Yes
AWS Glue	Yes	No	Partial	Yes	No
Hive Metastore	Yes	Yes	Unknown	Unknown	Yes
Nessie	Yes	Yes	No	Yes	Yes
Unity Catalog	Yes	No	Partial	Yes	Partial
JDBC catalog	No	Yes	Unknown	Unknown	Unknown

Serves REST API: Serves the Iceberg REST catalog API, so any REST client engine can connect.
Iceberg views: Stores and serves Iceberg views.
Format v3 tables: Supports format version 3 tables.
Credential vending: Hands engines short-lived, scoped storage credentials.
Multi-table commits: Commits changes to several tables atomically (REST commitTransaction, or catalog-level commits and branches).

Query engines: detail and sources

Apache Spark (Iceberg Spark runtime)

Evidence applies to: Iceberg 1.11.0 Spark runtime (Spark 3.5, 4.0, 4.1); docs/latest as of 2026-09-29. Main documentation

Feature	Status	What the documentation says
Read	Yes	Batch SELECT and DataFrame reads, incremental reads, and metadata tables. Source
Write	Yes	INSERT INTO, INSERT OVERWRITE, MERGE INTO, UPDATE, DELETE FROM and DataFrameWriterV2; CREATE TABLE, CTAS and RTAS via Spark DDL. Source
Position deletes	Yes	Row-level DELETE/UPDATE/MERGE write position deletes when write.*.mode is merge-on-read (v2+); needs Iceberg SQL extensions. Reads them. Source
Equality deletes	Read only	Reads equality deletes (1.11.0 fixed equality-delete reads with schema evolution); Spark writers produce position deletes or DVs, not equality deletes. Source
Deletion vectors	Yes	Writes DVs for v3 tables since 1.8.0; DVs surface in position_deletes metadata table; 1.9.0 rewrites v2 deletes to v3 DVs. Source
Format v3	Partial	Row lineage, DVs, variant (Spark 4.0+, shredded writes in 1.11) supported; nanosecond timestamps, geometry and geography not supported. Source
Branches and tags	Yes	CREATE/REPLACE/DROP BRANCH and TAG DDL, read via VERSION AS OF branch or tag, and writes to existing branches. Source
REST catalog	Yes	Configure a catalog with spark.sql.catalog.<name>.type = rest and a REST uri. Source
Iceberg views	Yes	CREATE, ALTER, DROP, SHOW and DESCRIBE Iceberg views on Spark 3.4 and above. Source
Time travel	Yes	SQL TIMESTAMP AS OF and VERSION AS OF (snapshot id, branch or tag) plus DataFrame read options. Source

Apache Flink (Iceberg Flink connector)

Evidence applies to: Iceberg 1.11.0 Flink runtime (Flink 1.20, 2.0, 2.1 jars); docs/latest as of 2026-09-29. Main documentation

Feature	Status	What the documentation says
Read	Yes	Batch and streaming reads via SQL and DataStream (FLIP-27 source); streaming reads are append-only, CDC read not supported. Source
Write	Yes	INSERT INTO, INSERT OVERWRITE, UPSERT and DataStream sinks incl. Dynamic Sink; CREATE TABLE via Flink DDL (no hidden partitioning). Source
Position deletes	Unknown	No official statement found on reading or writing position delete files; docs cover UPSERT and delete-file metrics only. Source
Equality deletes	Partial	UPSERT mode writes row-level changes keyed on equality fields; needs v2+ table with primary key or identifier fields. Source
Deletion vectors	Partial	1.11.0 added DV writing in IcebergSink and Dynamic Sink; DV read behavior not separately documented. Source
Format v3	Partial	Nanosecond timestamps, default values and row lineage readers supported; variant, geometry, geography listed as not supported in type mapping. Source
Branches and tags	Partial	Reads branches and tags (SQL options and DataStream) and writes to a branch via toBranch; no Flink DDL to create branches or tags. Source
REST catalog	Yes	CREATE CATALOG with 'catalog-type'='rest' and a REST uri. Source
Iceberg views	Unknown	No official statement found; Flink docs do not cover Iceberg views. Source
Time travel	Yes	Batch-mode read options snapshot-id and as-of-timestamp, plus branch and tag options. Source

Trino Iceberg connector

Evidence applies to: Trino 483 (docs/current, release notes 477 to 483). Main documentation

Feature	Status	What the documentation says
Read	Yes	Reads Iceberg spec v1 and v2 tables in Parquet, ORC, or Avro; v3 support is experimental. Source
Write	Yes	INSERT, UPDATE, DELETE, TRUNCATE, MERGE, plus CREATE TABLE and CREATE TABLE AS. Source
Position deletes	Yes	Row-level DELETE/UPDATE/MERGE on v2 tables write position delete files; the connector also reads them. Source
Equality deletes	Read only	Release notes cover reading tables with equality delete files; no statement that Trino writes them. Source
Deletion vectors	Partial	$files exposes DV fields; 480 notes add deletes on v3 tables, but connector docs still say v3 row-level deletes are unsupported. Source
Format v3	Partial	Experimental. Notes 480-482 add v3 create/write/delete, row lineage, defaults, variant, nanosecond ts, geometry; connector page still lists v3 limits. Source
Branches and tags	Partial	Queries branches and tags via FOR VERSION AS OF; Trino 477 added CREATE/DROP/ALTER BRANCH SQL. Tag creation and branch writes not documented. Source
REST catalog	Yes	iceberg.catalog.type=rest with OAuth2, SigV4, Google auth, and vended credentials (iceberg.rest-catalog.vended-credentials-enabled). Source
Iceberg views	Yes	With the REST catalog, view management uses the Iceberg View specification; view endpoints enabled by default. Source
Time travel	Yes	FOR VERSION AS OF snapshot id, branch, or tag, and FOR TIMESTAMP AS OF. Source

Dremio

Evidence applies to: Dremio Cloud (docs and changelog as of 2026-09-22); Dremio Software docs current 26.x checked for limits. Main documentation

Feature	Status	What the documentation says
Read	Yes	Iceberg is Dremio's native table format; reads v2 and v3 tables via Glue, REST, Snowflake Open Catalog, Unity Catalog. Source
Write	Yes	Full DML (INSERT, UPDATE, DELETE, MERGE), CREATE TABLE/CTAS, schema evolution, OPTIMIZE on v2 and v3 tables. Source
Position deletes	Yes	DML defaults to copy-on-write; with write.*.mode=merge-on-read, v2 tables get position delete files, read and written. Source
Equality deletes	Read only	Reads equality deletes (Aug 2026 fix); Software 26.x docs say global unpartitioned equality deletes in v2 are not read. Source
Deletion vectors	Yes	Merge-on-read DML on v3 tables writes Puffin deletion vectors and reads them; v2 position deletes convert on first MOR write. Source
Format v3	Yes	Create, upgrade, read, write v3: DVs, variant, row lineage, default values (primitive only), nanosecond ts. Changelog marks v3 Preview. Source
Branches and tags	No	Iceberg table-level branches and tags are not supported. Dremio's AT BRANCH, AT TAG, and AT COMMIT apply to catalog-level versioning in Nessie-backed (versioned) catalogs, not to Iceberg table refs. Confirmed by Alex Merced. Source
REST catalog	Yes	Connects to any Iceberg REST catalog (Polaris, Glue, S3 Tables, Nessie, OneLake, others) with optional vended credentials. Source
Iceberg views	Yes	Reads Iceberg views from catalogs, choosing the Dremio dialect; Open Catalog stores views as Iceberg views. Source
Time travel	Yes	SELECT ... AT SNAPSHOT '<id>' or AT TIMESTAMP, limited by snapshot retention. Source

DuckDB iceberg extension

Evidence applies to: DuckDB 1.5.x docs (docs/current; 1.5.6 released 2026-09-28). Main documentation

Feature	Status	What the documentation says
Read	Yes	iceberg_scan reads tables directly from storage (read-only), or read via an attached Iceberg REST catalog. Source
Write	Yes	Only through an attached REST catalog: CREATE TABLE/CTAS, INSERT, UPDATE, DELETE, MERGE INTO, ALTER TABLE. Source
Position deletes	Yes	Reads and writes positional deletes, merge-on-read only; no copy-on-write, and UPDATE/DELETE fail on sorted tables. Source
Equality deletes	Unknown	No official statement found Source
Deletion vectors	Yes	Reads and writes binary deletion vectors (Puffin) for v3 tables; DuckDB picks DVs automatically from format-version. Source
Format v3	Partial	Reads and writes v3: VARIANT, TIMESTAMP_NS, default values, DVs, row lineage. GEOGRAPHY and UNKNOWN types not yet supported. Source
Branches and tags	Unknown	No official statement found Source
REST catalog	Yes	ATTACH ... (TYPE iceberg) to REST catalogs; docs say only S3, S3 Tables, and GCS storage backends are supported. Source
Iceberg views	Unknown	No official statement found Source
Time travel	Yes	AT (VERSION => snapshot_id) or AT (TIMESTAMP => ...) on catalog tables; snapshot_from_id/timestamp in iceberg_scan. Source

Snowflake Apache Iceberg tables

Evidence applies to: Snowflake docs as of 2026-09 (Snowflake-managed, externally managed, and catalog-linked database tables). Main documentation

Feature	Status	What the documentation says
Read	Yes	Queries Snowflake-managed and externally managed tables (catalog integration or catalog-linked database). Iceberg spec v1, v2 and v3 supported. Source
Write	Yes	INSERT/UPDATE/DELETE/MERGE/COPY on managed and external REST tables. Remote CREATE TABLE only via catalog-linked DB; external writes are autocommit only. Source
Position deletes	Partial	Reads and writes v2 position deletes for UPDATE/DELETE/MERGE. Docs limit this to externally managed tables; managed v2 defaults to copy-on-write. Source
Equality deletes	No	Row-level equality deletes are listed as unsupported; Horizon IRC also rejects equality deletes from external engines. Source
Deletion vectors	Yes	Writes DVs for DELETE/UPDATE/MERGE on v3 tables (managed and external); merge-on-read is default for v3. Heuristics may fall back to copy-on-write. Source
Format v3	Partial	v3 create/read/write: variant, geo types, ns timestamps, unknown, defaults, DVs, row lineage. No multi-arg transforms, encryption keys, or in-place upgrade. Source
Branches and tags	No	Snowflake cannot create or query Iceberg branches or tags: no SQL for them is documented, and the Horizon catalog limits state creating Iceberg tags and branches isn't supported. Confirmed by Alex Merced. Source
REST catalog	Yes	Catalog integration for any Iceberg REST catalog (OAuth, bearer/PAT, SigV4, vended credentials); catalog-linked databases auto-discover and write tables. Source
Iceberg views	Unknown	No official statement found on reading or creating Iceberg-spec views; Snowflake views over Iceberg tables are Snowflake objects. Source
Time travel	Yes	Snowflake Time Travel within retention period. External tables: snapshots after table creation, if refreshed before the snapshot expires. Source

Google BigQuery (Iceberg managed tables, Lakehouse runtime catalog, external tables)

Evidence applies to: BigQuery and Lakehouse docs as of 2026-09 (BigLake names now Iceberg managed tables and Lakehouse runtime catalog). Main documentation

Feature	Status	What the documentation says
Read	Yes	Reads Iceberg managed tables, Lakehouse runtime catalog tables, and Iceberg external tables (v2, v3 in Preview). Parquet only. Source
Write	Partial	Full DML on BigQuery-managed Iceberg tables (GA). DDL/DML on Lakehouse runtime catalog tables is Preview, v2 only. External tables are read-only. Source
Position deletes	Read only	Reads merge-on-read position deletes (up to 10,000 delete files per data file). BigQuery writes only support copy-on-write. Source
Equality deletes	Read only	Reads equality deletes on external tables, max 100,000 equality deletes per data file. No documented equality delete writes. Source
Deletion vectors	Read only	Preview: BigQuery reads binary DVs on v3 tables; BigQuery writes to v3 tables aren't supported (use Spark, Flink or Trino). Source
Format v3	Partial	Preview, read-only from BigQuery: DVs only. New v3 types, default values, row lineage and encryption keys are not supported. Source
Branches and tags	Unknown	No official statement found on BigQuery querying or creating branches or tags; docs only note snapshot expiry is skipped when refs exist. Source
REST catalog	Partial	Uses Google Lakehouse runtime catalog (Iceberg REST). Remote REST catalogs (Glue, Unity, Horizon, Workday) only via cross-cloud federation (Preview). Source
Iceberg views	No	The Lakehouse runtime catalog doesn't support Apache Iceberg views. Source
Time travel	Yes	FOR SYSTEM_TIME AS OF on external table snapshots; BigQuery time travel limited to 7 days on managed and runtime catalog tables. Source

Amazon Athena

Evidence applies to: Athena engine version 3 (Iceberg library 1.4.2), docs as of 2026-09. Main documentation

Feature	Status	What the documentation says
Read	Yes	SELECT on Iceberg v2 tables in AWS Glue Data Catalog, including S3 Tables; Parquet, ORC and Avro data files. Source
Write	Yes	CREATE TABLE, CTAS, INSERT INTO, UPDATE, DELETE, MERGE INTO. Creates v2 tables only; DDL not supported on Lake Formation registered tables. Source
Position deletes	Yes	Reads position deletes; UPDATE, DELETE and MERGE always write position delete files (merge-on-read). Copy-on-write isn't supported. Source
Equality deletes	Read only	Queries merge-on-read both position and equality deletes; Athena writes only position deletes. Source
Deletion vectors	No	Athena only creates and operates on Iceberg v2 tables, so v3 deletion vectors are not supported. Source
Format v3	No	Athena only creates and operates on Iceberg v2 tables; Iceberg library version is 1.4.2. Source
Branches and tags	Unknown	No official statement found on branch or tag create, query or write; only the $refs metadata table is documented. Source
REST catalog	No	Only Iceberg tables in the AWS Glue Data Catalog are supported (S3 Tables via Glue catalog registration); no generic REST catalog connection. Source
Iceberg views	No	Views listed as unsupported for Iceberg: use Athena CREATE VIEW instead; Iceberg view specification not supported. Source
Time travel	Yes	FOR TIMESTAMP AS OF and FOR VERSION AS OF (snapshot ID) in engine version 3. Source

StarRocks

Evidence applies to: StarRocks 4.1 (docs latest; 4.1.5 release notes). Main documentation

Feature	Status	What the documentation says
Read	Yes	Queries v1 and v2 tables in Parquet and ORC via Iceberg catalogs (Hive metastore, Glue, REST, JDBC, others). Source
Write	Yes	CREATE TABLE, INSERT INTO/OVERWRITE (v3.1+), DELETE (v4.1+), TRUNCATE. UPDATE and MERGE not documented. Source
Position deletes	Yes	Reads v2 position deletes (v3.0/v3.1+); v4.1.0 added writing position delete files for DELETE. Source
Equality deletes	Read only	Queries on v2 tables support equality deletes from v3.1.10/v3.2.5/v3.3 (Parquet); no write support documented. Source
Deletion vectors	Unknown	No official statement found Source
Format v3	Partial	v4.1 adds v3 default values, row lineage (_row_id reads), VARIANT; unsupported v3 features fail fast. DVs, geometry, ns ts not documented. Source
Branches and tags	Yes	Beta: CREATE/DROP BRANCH and TAG, INSERT INTO a branch, and query branch or tag with VERSION AS OF. Source
REST catalog	Yes	iceberg.catalog.type=rest, with vended credentials from v4.0 and view endpoints enabled by default. Source
Iceberg views	Yes	CREATE VIEW makes Iceberg views (v3.5+); ALTER VIEW adds a StarRocks dialect to existing Iceberg views. Source
Time travel	Yes	Beta since v3.4: VERSION AS OF snapshot id, branch, or tag, and TIMESTAMP AS OF datetime or date. Source

PyIceberg

Evidence applies to: PyIceberg 0.12.0 (released 2026-09-01). Main documentation

Feature	Status	What the documentation says
Read	Yes	table.scan() with row filters and projections into Arrow, pandas, DuckDB, Polars, Ray, Daft and others. Source
Write	Yes	create_table, append, overwrite (with filter), delete and upsert via PyArrow; streaming RecordBatchReader writes only for unpartitioned tables. Source
Position deletes	Read only	Reads position deletes (DeleteFileIndex, 0.11.0); delete() warns merge-on-read is not yet supported and falls back to copy-on-write. Source
Equality deletes	No	Code raises 'PyIceberg does not yet support equality deletes' when equality delete files are encountered. Source
Deletion vectors	Read only	Deletion Vectors read support added in 0.10.0; writes use copy-on-write, so no DVs are produced. Source
Format v3	Partial	Reads v3; has row lineage fields, DV reads, nanosecond, default and geo types. Nanosecond writes in v3 not implemented; v3 metadata JSON write raises. Source
Branches and tags	Yes	manage_snapshots() creates branches and tags; writes to branches (WAP example) and fast-forward supported. Source
REST catalog	Yes	REST catalog configuration with OAuth, SigV4 and custom headers. Source
Iceberg views	Partial	create_view, load_view, list, exists and drop on REST catalogs; view support tied to the REST catalog. Source
Time travel	Yes	table.scan(snapshot_id=...) and snapshot_as_of_timestamp(); metadata tables accept snapshot_id. Source

Catalogs: detail and sources

Iceberg REST catalog spec

Evidence applies to: REST OpenAPI spec at Iceberg 1.11.0. Main documentation

Feature	Status	What the documentation says
Serves REST API	Yes	Defines the REST catalog API: config, namespaces, tables, views, scan planning, metrics, signing and credentials endpoints Source
Iceberg views	Yes	Defines list, create, load, replace, rename, drop and register view endpoints Source
Format v3 tables	Yes	TableMetadata allows format-version 1 through 3; update types include upgrade-format-version and row lineage fields Source
Credential vending	Yes	Defines X-Iceberg-Access-Delegation (vended-credentials, remote-signing), storage-credentials and a table credentials endpoint Source
Multi-table commits	Yes	Defines POST /v1/{prefix}/transactions/commit to commit updates to multiple tables atomically; servers may omit it Source

Apache Polaris (REST)

Evidence applies to: Apache Polaris 1.8.0 (released 2026-09-28; bundles Iceberg 1.11.x). Main documentation

Feature	Status	What the documentation says
Serves REST API	Yes	Built on the Iceberg REST protocol; any REST-compatible engine can read and write Polaris-managed tables Source
Iceberg views	Yes	Polaris view entities map to Iceberg views, managed through the REST CreateView/LoadView APIs Source
Format v3 tables	Yes	Polaris registers and serves Iceberg format v3 tables. Confirmed in practice by Alex Merced; the 1.8.0 release notes do not state it explicitly. Source
Credential vending	Yes	Returns short-lived, scoped storage credentials on table or view load for S3, Azure ADLS and GCS Source
Multi-table commits	Yes	Implements the REST commitTransaction endpoint; release notes cover commitTransaction fixes and retryable 409 conflicts Source

AWS Glue Data Catalog

Evidence applies to: AWS Glue docs as of 2026-09 (Iceberg REST endpoint; Glue 6.0 for v3). Main documentation

Feature	Status	What the documentation says
Serves REST API	Yes	Serves Iceberg REST at glue.<region>.amazonaws.com/iceberg with SigV4; single-level namespaces, no staged create (CTAS) Source
Iceberg views	No	Docs state View APIs in the Iceberg REST spec are not supported; Glue has its own multi-dialect Data Catalog views Source
Format v3 tables	Partial	Glue 6.0 Spark jobs create or upgrade v3 tables (variant, nanosecond timestamps, geo); REST endpoint page lists only v1 and v2 Source
Credential vending	Yes	Lake Formation vends temporary credentials to external engines once full table access for external engines is enabled Source
Multi-table commits	No	CommitTransaction is absent from the documented list of supported Iceberg REST operations Source

Hive Metastore

Evidence applies to: Iceberg 1.11.0 HiveCatalog; Hive 4.x Metastore Iceberg REST (Hive 4.2.1 latest). Main documentation

Feature	Status	What the documentation says
Serves REST API	Yes	HiveCatalog itself uses Thrift, but Hive 4.x Metastore also serves Iceberg REST endpoints (default port 9001 in Docker images) Source
Iceberg views	Yes	View support for the Hive catalog added in Iceberg 1.7.0; HMS REST endpoint list includes view operations Source
Format v3 tables	Unknown	No official statement found; Hive engine Iceberg docs only describe v1 and v2 tables. The catalog stores a metadata pointer; the writing engine sets the format version. Source
Credential vending	Unknown	No official statement found; the HMS REST endpoint list shows no credentials endpoint Source
Multi-table commits	Yes	HMS Iceberg REST advertises POST /v1/{prefix}/transactions/commit in its /v1/config endpoint list Source

Project Nessie

Evidence applies to: Nessie 0.108.8 (2026-09-09). Main documentation

Feature	Status	What the documentation says
Serves REST API	Yes	Serves Iceberg REST, which Nessie labels experimental; Nessie manages table and view metadata Source
Iceberg views	Yes	Versions Iceberg views alongside tables on branches Source
Format v3 tables	No	Release notes state Iceberg table spec v3 is not supported (0.103.0); no later statement found Source
Credential vending	Yes	Provides short-lived, down-scoped credentials or S3 request signing via Iceberg REST; GCS downscoping fixed in 0.108.8 Source
Multi-table commits	Yes	Cross-table transactions via single commits touching many tables, or via branches merged atomically Source

Unity Catalog

Evidence applies to: Databricks Unity Catalog docs as of 2026-09 (DBR 18 LTS+ for v3); OSS Unity Catalog 0.6.0 differs. Main documentation

Feature	Status	What the documentation says
Serves REST API	Yes	Databricks serves Iceberg REST at /api/2.1/unity-catalog/iceberg-rest (read and write managed Iceberg). OSS UC 0.6.0 only reads UniForm tables Source
Iceberg views	No	Databricks docs: external Iceberg clients can't access views Source
Format v3 tables	Partial	Managed tables support v3 DVs, variant, row lineage, geo types, write defaults; not initial defaults, unknown type, nanosecond timestamps Source
Credential vending	Yes	Vends short-lived credentials via Iceberg REST: read/write/create for managed Iceberg, read-only for foreign Iceberg Source
Multi-table commits	Partial	Multi-table SQL transactions on Databricks compute; Iceberg table support is Private Preview. No REST commitTransaction documented Source

Iceberg JDBC catalog

Evidence applies to: Iceberg 1.11.0 JdbcCatalog. Main documentation

Feature	Status	What the documentation says
Serves REST API	No	A client-side catalog library: each engine connects to the database itself. It does not serve the REST API. Source
Iceberg views	Yes	View support added in Iceberg 1.5.0; requires jdbc.schema-version=V1 in catalog properties Source
Format v3 tables	Unknown	No official statement found on format v3 for the JDBC catalog. The catalog stores a metadata pointer; the writing engine sets the format version. Source
Credential vending	Unknown	No official statement found; docs cover only JDBC connection settings Source
Multi-table commits	Unknown	No official statement found; docs describe atomic single-table commits via database transactions Source

How this was checked

Every cell was checked on September 30, 2026 by fetching the cited page and reading the relevant section. Sources are the vendor's or project's own documentation, release notes, or changelog. No third-party posts.
Unknown means the official documentation says nothing either way. It is not a polite "no". Many unknowns are features that work in practice but are not documented.
No is used only where the documentation states the limitation.
Where a product's docs contradict its release notes (Trino's v3 support, Flink's variant type, Snowflake's position deletes), the note says so and the status reflects the more cautious reading.
Preview and beta features are counted, and the note names them as such.
Iceberg moves quickly. Treat anything older than a few months as a starting point and check the linked source before you commit to an architecture.

The data behind this page: compatibility.json. For background on the choices, see comparing catalogs and comparing table formats.

Legend