v0.18

pgcopydb v0.18 (June 27, 2026)

pgcopydb v0.18 is a major feature release. PostgreSQL 17 and 18 are now
fully supported and tested. The pgoutput logical decoding plugin is now the
default, replacing test_decoding, and CDC filtering has been significantly
extended: publications are created to match the active filter set, wal2json
respects filter-tables, and new [include-only-extension] and
[exclude-extension] sections control which extensions are copied.

The internal format for CDC streaming data has been rewritten to use SQLite.
WAL changes are now stored in structured output.db and replay.db databases
instead of JSON files, enabling reliable processing across process restarts,
size-based file rotation, and a new stream cleanup command to reclaim disk
space.

Other major additions include support for cloning an entire Postgres
instance with --all-databases, Citus-to-Citus migration, regex patterns in
filter INI files, and arm64 Docker images.

Added

• Use SQLite as the file format for CDC streaming (#715)
• Add pgoutput logical decoding plugin support, now the default (#962)
• Add --all-databases flag to clone an entire Postgres instance (#967)
• Add size-based file rotation for SQLite outputDB/replayDB (#961)
• Add Citus-to-Citus migration support (#1011)
• Batch consecutive DELETEs into a single parameterised statement (#1010)
• Add regex/pattern syntax (~/pattern/) in filter INI files (#1004)
• Add filter-aware publication creation and wal2json filter-tables (#975)
• Add [exclude-extension] and [include-only-extension] filter sections (#989)
• Add --replay-no-op-updates option to CDC streaming (#983)
• Add support for the timescaledb extension (#849)
• Add PostgreSQL 17 and 18 build and test support (#958)
• Add arm64 Docker image build (#911)

Changed

• Batch SQLite catalog inserts and defer index creation for performance (#999)
• Cache table and attribute catalog lookups in transform hot path (#950)
• Add filter role model, command_log table, and conditional catalog invalidation (#1001)
• Add stream cleanup command to reclaim CDC file disk space (#1003)
• Fetch table attributes in a separate query using OID arrays (#993)
• Extract SQL queries from schema.c into standalone .sql files (#992)
• Set idle_in_transaction_session_timeout to zero on the target (#904)
• Use our own memory area for editing libpq error messages (#903)

Fix

• Fix client_encoding forced to UTF-8 breaking COPY from SQL_ASCII source databases (#874)
• Fix missing --split-tables-larger-than and --split-max-parts in copy table-data help (#921)
• Fix LOCK TABLE failure aborting COPY worker transaction; recover with SAVEPOINT (#875)
• Fix snapshot file not cleaned up when logical replication snapshot closes (#1014)
• Fix GUC settings not applied on replication slot creation connection (#1013)
• Fix list commands to be snapshot-agnostic (#1007)
• Fix COPY Transfer column to report actual wire bytes (#1002)
• Fix stale process/summary entries and reopen catalog after fork (#1000)
• Fix build compatibility with PostgreSQL 16-19 meson-built installations (#998)
• Replace md5() with sha256() in compare data checksum query (#997)
• Honour --requirements version pins when cloning extensions (#996)
• Detect binary-incompatible column types and fall back to text COPY (#994)
• Use named column access for IDENTIFY_SYSTEM to support AlloyDB (#995)
• Make extension copy resumable using summary table (#990)
• Fix noisy filter-mismatch warning for list/sentinel commands (#988)
• Fix --skip-collations failure when the same collation spans multiple objects (#986)
• Fix crash and ACL loss for function/table ACL entries (#985)
• Fix source URI check in catalog for list databases (#984)
• Fix list table-parts always reporting 0 parts (#982)
• Fix CDC apply for materialized views (#981)
• Preserve full double precision in CDC replay (#980)
• Fix single-quote escaping for VARCHAR/text in test_decoding CDC (#979)
• Validate and normalize filter names against source catalogs (#978)
• Prevent segfault in compare schema when object missing on target (#977)
• Skip GENERATED ALWAYS AS IDENTITY columns in CDC UPDATE SET clause (#976)
• Close SQLite catalog connections before fork (#974)
• Use (catoid, oid) object identity to fix filter OID collisions (#972)
• Fix double constraint creation for tables with EXCLUSION constraint indexes (#965)
• Fix KEEPALIVE crash when timestamp is empty (#959)
• Create large object on target if it doesn't exist (PG17+ compat) (#957)
• Fix no open transaction in clone --follow --snapshot (#956)
• Fix TRUNCATE ONLY and COPY FREEZE on partitioned target tables (#949)
• Fix test_decoding parser for REPLICA IDENTITY USING INDEX (#946)
• Fix parsing quoted null (#933)
• Fix invalid UPDATE stmt when none of the columns change (#883)
• Skip toast columns with unchanged values in test_decoding transform (#877)
• Fix deadlock during pipeline sync (#880)
• Fix bug in passing logSQL from spec to context (#878)
• Fix bad copy-paste in compare_fetch_schemas using source schema for target (#935)
• Fix database source already in use error when re-running commands (#953)
• Fix catalog mismatch after resume (#952)
• Fix extension OID collision in skip-extensions logic (#954)
• Fix archive TOC line parser truncating lines longer than 1024 chars (#867)
• Fix pg_dump \restrict / \unrestrict metacommand syntax (CVE-2025-8714) (#947)
• Fix build failure with PostgreSQL 18 (#938)
• Fix a segfault when SQL state is NULL (#902)
• Fix zero allocation during pgcopydb list progress (#891)
• Fix pipe done read condition (#890)
• Remove one double free (#908)