Release notes

1.3.5 (2025-06-18)

Added

  • Human-readable changelog in CHANGELOG.md.

  • Support for PostgreSQL 18.

Fixed

  • Regression introduced in 1.3.4 leading to assertion failures in reaper processes on cassert-enabled PostgreSQL builds.

  • Two potential cases of uninitialized variables usage reported by Clang.

  • Development and testing scripts were ported to Python 3.

  • Combiner processes used the server timezone when defining range bounds for automatically created partitions, if the partitioning attribute was of the TIMESTAMP WITH TIME ZONE type. That resulted in problems caused by overlapping or disjoint partition ranges after the default server timezone has changed for whatever reasons. Fixed by sticking to UTC when rounding timestamptz values to calculate partition range bounds.

  • Installing Tantor PipelineDB on PostgreSQL 17 could lead to the “function interval_combine() does not exist” error.

  • The sliding-window queries could work incorrectly or lead to a combiner process crash.

1.3.4 (2025-02-20)

Added

  • Support for PostgreSQL 17.

Fixed

  • If the pg_store_plans extension was installed together with Tantor PipelineDB, it could result in combiner failures when processing query execution plans created by Tantor PipelineDB for internal purposes. Fixed by configuring pg_store_plans to avoid examining execution plans created by Tantor PipelineDB background workers.

  • If Time-To-Live option was configured for partitioned CVs with some partitions converted to Hydra Columnar, the repear processes could end up in a broken state when deleting tuples from Columnar partitions. Fixed by disabling the TTL feature for partitioned CVs until partition-aware reaper functionality is implemented in a later release.

1.3.3 (2024-12-24)

Fixed

  • PostgreSQL 15.9 introduced an incompatible change resulting in the “missing lock for relation” warning message printed in the log during ALTER VIEW or ALTER SCHEMA commands in cassert-enabled Tantor PipelineDB builds when CVs were involved.

  • Typo in 1.3.1 → 1.3.2 extension upgrade script name.

1.3.2 (2024-12-05)

Added

  • New GUC pipelinedb.index_fillfactor controls fillfactor for automatically created indexes for CV materialized relations. This applies to both. partitioned and regular CVs.

1.3.1 (2024-11-20)

Fixed

  • The pipelinedb.fillfactor GUC did not work for partitioned CVs.

  • For partitioned CVs with aggregates, combiner processes could sometimes add new tuples with instead of updating existing ones.

  • Combiner processed could leak memory, if updating a partitioned CV failed with an error for whatever reasons.

  • Converting a partition of a partitioned CV to Hydra Columnar could fail with the “could not option relation with OID” error.

  • After converting a partition of a partitioned CV to Hydra Columnar combiner processes cound end up in a broken state, requiring a restart to continue processing data.

Removed

  • The “anonymous update checks” functionality that would periodically check for Tantor PipelineDB updates in the background and provide anonymized statistics at the same time.

1.3.0 (2024-11-12)

This is the first formal Tantor release.

Added

  • Support for PostgreSQL 15/16.

  • Support for partitioned continuous views.

  • Support for converting CV partitions to other table access methods such as Hydra Columnar.

Fixed

  • A number of memory management and functionality bugs revealed by AddressSanitizer and UndefinedBehaviorSanitizer.

1.0.0-13 (2019-02-05)

This was the last release from pipelinedb.com (GitHub).