High-level changes
- Improved support for primary-key, foreign-key, unique, and exclusion constraints.
- New histogram function added for getting the frequency of a column's values.
- Add support for using
DATE
as partition column. chunk_time_interval
now supportsINTERVAL
data types- Block several unsupported and/or dangerous operations on hypertables and chunks, including dropping or otherwise altering a chunk directly.
- Several bug fixes throughout the code.
Notable commits
- [afcb0b1] Fix NULL handling in first/last functions.
- [d53c705] Add script to dump meta data that can be useful for debugging.
- [aa904fa] Block adding constraints without a constraint name
- [a13039f] Fix dump and restore for tables with triggers and constraints
- [8cf8d3c] Improve the size utils functions.
- [2767548] Block adding constraints using an existing index
- [5cee104] Allow chunk_time_interval to be specified as an INTERVAL type
- [6232f98] Add histogram function.
- [2380033] Block ALTER TABLE and handle DROP TABLE on chunks
- [72d6681] Move security checks for ALTER TABLE ALTER COLUMN to C
- [19d3d89] Handle changing the type of dimension columns correctly.
- [17c4ba9] Handle ALTER TABLE rename column
- [66932cf] Forbid relocating extension after install.
- [d2561cc] Add ability to partition by a date type
- [48e0a61] Remove triggers from chunk and chunk_constraint
- [4dcbe61] Add support for hypertable constraints
Thanks
- @raycheung for reporting a segfault in
first
/last
- @meotimdihia, @noyez, and @andrew-blake for reporting issues with
UNQIUE
and other types of constraints
High-level changes
- Provide scripts for backing up and restoring single hypertables
Notable commits
- [683c078] Add backup/restore scripts for single hypertables
High-level changes
- Bug fix for a segmentation fault in the planner
- Shortcut when constraint-aware append excludes all chunks
- Fix edge case with negative timestamps when points fell right on the boundary
- Fix behavior of
time_bucket
forDATE
types by not converting toTIMESTAMPTZ
- Make the output of
chunk_relation_size
consistent
Notable commits
- [50c8c4c] Fix possible segfault in planner
- [e49e45c] Fix failure when constraint-aware append excludes all chunks
- [c3b6fb9] Fix bug with negative dimension values
- [3c69e4f] Fix semantics of time_bucket on DATE input
- [0137c92] Fix output order of chunk dimensions and ranges in chunk_relation_size.
- [645b530] Convert inserted tuples to the chunk's rowtype
Thanks
- @yadid for reporting a segfault (fixed in 50c8c4c)
- @ryan-shaw for reporting tuples not being correctly converted to a chunk's rowtype (fixed in 645b530)
High-level changes
- Exclude chunks when constraints can be constifyed even if they are
considered mutable like
NOW()
. - Support for negative values in the dimension range which allows for pre-1970 dates.
- Improve handling of default chunk times for integral date times by forcing it to be explicit rather than guessing the units of the time.
- Improve memory usage for long-running
COPY
operations (previously it would grow unbounded). VACUUM
andREINDEX
on hypertables now recurse down to chunks.
Notable commits
- [139fe34] Implement constraint-aware appends to exclude chunks at execution time
- [2a51cf0] Add support for negative values in dimension range
- [f2d5c3f] Error if add_dimension given both partition number and interval length
- [f3df02d] Improve handling of non-TIMESTAMP/TZ timestamps
- [6a5a7eb] Reduce memory usage on long-running COPY operations
- [953346c] Make VACUUM and REINDEX recurse to chunks
- [55bfdf7] Release all cache pins when a transaction ends
High-level changes
- "Upserts" are now supported via normal
ON CONFLICT DO UPDATE
/ON CONFLICT DO NOTHING
syntax. However,ON CONFLICT ON CONSTRAINT
is not yet supported. - Improved support for user-defined triggers on hypertables. Now handles both INSERT BEFORE and INSERT AFTER triggers, and triggers can be named arbitrarily (before, a _0_ prefix was required to ensure correct execution priority).
TRUNCATE
on a hypertable now deletes empty chunks.
Notable commits
- [23f9d3c] Add support for upserts (
ON CONFLICT DO UPDATE
) - [1f3dcd8] Make
INSERT
s use a custom plan instead of triggers - [f23bf58] Remove empty chunks on
TRUNCATE
hypertable.
High-level changes
- Users can now define their own triggers on hypertables (except for
INSERT AFTER
) - Hypertables can now be renamed or moved to a different schema
- Utility functions added so you can examine the size hypertables, chunks, and indices
Notable commits
- [83c75fd] Add support for triggers on hypertables for all triggers except
INSERT AFTER
- [e0eeeb9] Add hypertable, chunk, and indexes size utils functions.
- [4d2a65d] Add infrastructure to build update script files.
- [a5725d9] Add support to rename and change schema on hypertable.
- [142f58c] Cleanup planner and process utility hooks
IMPORTANT NOTE
Starting with this release, TimescaleDB will now
support upgrading between extension versions using the typical
ALTER EXTENSION
command, unless otherwise noted in future release notes. This
important step should make it easier to test TimescaleDB and be able
to get the latest benefits from new versions of TimescaleDB. If you
were previously using a version with the -beta
tag, you will need
to DROP
any databases currently using TimescaleDB and re-create them
in order to upgrade to this new version. To backup and migrate data,
use pg_dump
to save the table schemas and COPY
to write hypertable
data to CSV for re-importing after upgrading is complete. We describe
a similar process on our docs.
High-level changes
- More refactoring to stabilize and cleanup the code base for supporting upgrades (see above note)
- Correct handling of ownership and permission propagation for hypertables
- Multiple bug fixes
Notable commits
- [696cc4c] Provide API for adding hypertable dimensions
- [97681c2] Fixes permission handling
- [aca7f32] Fix extension drop handling
- [9b8a447] Limit the SubspaceStore size; Add documentation.
- [14ac892] Fix possible segfault
- [0f4169c] Fix check constraint on dimension table
- [71c5e78] Fix and refactor tablespace support
- [5452dc5] Fix partiton functions; bug fixes (including memory)
- [e75cd7e] Finer grained memory management
- [3c460f0] Fix partitioning, memory, and tests
- [fe51d8d] Add native scan for the chunk table
- [fc68baa] Separate out subspace_store and add it to the hypertable object as well
- [c8124b8] Use hypercube instead of dimension slice list
- [f5d7786] Change the semantics of range_end to be exclusive
- [700c9c8] Refactor insert path in C.
- [0584c47] Created chunk_get_or_create in sql with an SPI connector in C
- [7b8de0c] Refactor catalog for new schema and add native data types
- [d3bdcba] Start refactoring to support any number of partitioning dimensions
High-level changes
- A major cleanup and refactoring was done to remove legacy code and currently unused code paths. This change is backwards incompatible and will require a database to be re-initialized and data re-imported. This refactoring will allow us to provide upgrade paths starting with the next release.
COPY
andINSERT
commands now return the correct number of rows- Default indexes no longer duplicate existing indexes
- Cleanup of the Docker image and build process
- Chunks are now time-aligned across partitions
Notable commits
- [3192c8a] Remove Dockerfile and docker.mk
- [2a01ebc] Ensure that chunks are aligned.
- [73622bf] Fix default index creation duplication of indexes
- [c8872fe] Fix command-tag for COPY and INSERT
- [bfe58b6] Refactor towards supporting version upgrades
- [db01c84] Make time-bucket function parallel safe
- [18db11c] Fix timestamp test
- [97bbb59] Make constraint exclusion work with non-text partition keys
- [f2b42eb] Fix problems with partitioning logic for padded fields
- [997029a] if_not_exist flag to create_hypertable now works on hypertables with data as well
- [347a8bd] Reference the correct column when scanning partition epochs
- [88a9849] Fix bug with timescaledb.allow_install_without_preload GUC not working
High-level changes
- New
first(value, time)
andlast(value, time)
aggregates - Remove
setup_timescaledb()
function to streamline setup - Allow for use cases where restarting the server is not feasible by force loading the library
- Disable time series optimizations on non-hypertables
- Add some default indexes for hypertables if they do not exist
- Add "if not exists" flag for
create_hypertable
- Several bug fixes and cleanups
Notable commits
- [8ccc8cc] Add if_not_exists flag to create_hypertable()
- [2bc60c7] Fix time interval field name in hypertable cache entry
- [4638688] Improve GUC handling
- [cedcafc] Remove setup_timescaledb() and fix pg_dump/pg_restore.
- [34ad9a0] Add error when timescaledb library is not preloaded.
- [fc4ddd6] Fix bug with dropping chunks on tables with indexes
- [32215ff] Add default indexes for hypertables
- [b2900f9] Disable query optimization on regular tables (non-hypertables)
- [f227db4] Fixes command tag return for COPY on hypertables.
- [eb32081] Fix Invalid database ID error
- [662be94] Add the first(value, time),last(value, time) aggregates
- [384a8fb] Add regression tests for deleted unit tests
- [31ee92a] Remove unit tests and sql/setup
- [13d3acb] Fix bug with alter table add/drop column if exists
- [f960c24] Fix bug with querying a row as a composite type
High-level changes
- New
time_bucket
functions for doing roll-ups on varied intervals - Change default partition function (thanks @robin900)
- Variety of bug fixes
Notable commits
- [1c4868d] Add documentation for chunk_time_interval argument
- [55fd2f2] Fixes command tag return for
INSERT
s on hypertables. - [c3f930f] Add
time_bucket
functions - [b128ac2] Fix bug with
INSERT INTO...SELECT
- [e20edf8] Add better error checking for index creation.
- [72f754a] use PostgreSQL's own
hash_any
function as default partfunc (thanks @robin900) - [39f4c0f] Remove sample data instructions and point to docs site
- [9015314] Revised the
get_general_index_definition
function to handle cases where indexes have definitions other than justCREATE INDEX
(thanks @bricklen)