Skip to content

chore(deps): bump clickhouse-connect from 0.15.1 to 1.12.0 in /services/controlplane - #41

Open
dependabot[bot] wants to merge 1 commit into
mainfrom
dependabot/pip/services/controlplane/clickhouse-connect-1.12.0
Open

dependabot[bot] wants to merge 1 commit into
mainfrom
dependabot/pip/services/controlplane/clickhouse-connect-1.12.0

Conversation

@dependabot

@dependabot dependabot Bot commented on behalf of github Nov 16, 2026

Copy link
Copy Markdown
Contributor

Bumps clickhouse-connect from 0.15.1 to 1.12.0.

Release notes

Sourced from clickhouse-connect's releases.

v1.12.0

clickhouse-connect v1.12.0

Highlights

Features

  • ping() on the sync and async clients accepts an optional timeout in seconds (default 3). Closes #192.
  • New get_client(dedicated_pool=True) gives a synchronous HTTP client its own connection pool. Closing a client-owned pool now closes idle connections immediately. Closes #1121.

Performance

  • Faster Array column reads with the C extensions in all query methods. Related to #1008.
  • Faster Cython decoding of Date and Date32 columns.
  • Faster query_df reads of Map, nullable integer, nullable date, and timezone-aware nullable datetime columns with the Python codec.
  • AsyncClient.query_rows_stream reads each block in one executor call instead of one per row, so it is much faster.
  • Faster client creation with non-UTC IANA server timezones. Closes #1120.
  • Eligible SQLAlchemy Core and ORM SELECTs skip the extra LIMIT 0 metadata query for empty results. Closes #1108 and #1134.

Bug Fixes

  • Correct Date32 decoding before 1601 and in leap centuries such as 1600 and 2400.
  • Nullable(Tuple(...)) and Nullable(Point) now read and insert correctly with the Python codec, including NULLs. Closes #1102 and #1104.
  • pandas.Timestamp keeps nanoseconds in DateTime64(7..9) binds and inserts. Tuple, Nested, Map, and Variant parameters keep DateTime64 precision. Related to #429.
  • Datetime parameters whose names end in _64 no longer fail with KeyError in pyformat queries. Closes #1131.
  • Large queries are no longer form-encoded automatically, which avoids Field value too long errors. Closes #1105.
  • NumPy and pandas reads: LIMIT 0 queries return empty results, object-column rows keep the correct shape, and numeric LowCardinality columns return every row.
  • query_df with use_none=False works on Nullable(Date) and Nullable(Date32).
  • Async connection drops raise OperationalError or StreamFailureError instead of leaking aiohttp errors. Mid-stream drops on both clients report StreamFailureError and keep the original cause.
  • AsyncClient.query(context=...) applies per-call arguments, and query(context=...) runs a stored command statement.
  • Context tz_mode is preserved across all query methods.
  • Enum names with backslashes or control characters parse and insert correctly. See #1117.
  • Decimal256(S) resolves to precision 76.
  • Fixes to timezone file handling, Tuple serialization error reporting, Nullable(SimpleAggregateFunction) NULL inserts, and a C buffer crash on zero-length first reads.

Other

  • AsyncClient.raw_insert now types insert_block as an async iterable of bytes. Runtime behavior is unchanged.

Note for Python 3.14+

lz4 is no longer installed by default on Python 3.14 and later. In that case compress=True compresses inserts with zstd from the standard library. To keep using compress="lz4",install clickhouse-connect[lz4]. Nothing changes below 3.14.

Installation

pip install --upgrade clickhouse-connect

See the CHANGELOG for full details.

v1.11.0

clickhouse-connect v1.11.0

Highlights

... (truncated)

Changelog

Sourced from clickhouse-connect's changelog.

1.12.0, 2026-11-03

Bug Fixes

  • Fixed Date32 decoding before 1601 and in leap centuries such as 1600 and 2400. Python date results support years 1 through 9999, with out-of-range values raising ValueError. The same calendar correction applies to epoch-based DateTime64 decoding.
  • With the Python codec and extended pandas dtypes, query_df preserves Nullable(UInt64) values above 2**53 without rounding, keeps timezone-aware nullable datetime dtypes across all-NULL blocks, and reads valid Nullable(DateTime64) values beyond the nanosecond range on pandas 2.
  • Datetime parameters whose names end in _64 now retain their names in pyformat queries, fixing KeyError failures including SQLAlchemy queries with 64 datetime binds. The _64 precision hint still applies when the query uses the name without the suffix. Closes #1131.
  • The Python codec now decodes Nullable(Tuple(...)) correctly, preserving NULL tuples and non-NULL values, including nested tuples and tuple read formats. Closes #1102.
  • The Python codec now decodes Nullable(Point) columns, returning None for NULL points. Previously the decoder read the NULL map as point data and the query failed.
  • With the Rust codec, query_column_block_stream now returns unnamed Nullable(Tuple(...)) columns, including Nullable(Tuple()), as lists instead of tuples, matching the Python codec. Row results are unchanged.
  • The Python codec now inserts None into Nullable(Tuple(...)) and Nullable(Point) columns as NULL, including when nested in Array or Tuple. Previously these inserts failed with TypeError or AttributeError before reaching the server. Tuple elements keep their declared types' validation and input handling, including Decimal aliases and nullable SimpleAggregateFunction elements. Native insert headers now preserve the tuple's Nullable wrapper. Closes #1104.
  • The Python codec now inserts None into Nullable(SimpleAggregateFunction(...)) columns as NULL.
  • Tuple serialization now reports non-sequence rows with DataError and a row number instead of a raw TypeError or AttributeError, including None rows in dictionary input for a non-nullable named tuple.
  • Decimal256(S) now resolves to the supported precision of 76 instead of 79. Explicit Decimal256 type names now produce valid Native insert headers, and SQLAlchemy Decimal256 types report the correct precision.
  • Sync and async clients no longer automatically use form encoding when the final SQL exceeds 128 KiB, avoiding HTML Form Exception: Field value too long for large queries with large bound parameters. Explicit form encoding and queries with external data retain their existing limits. Closes #1105.
  • pandas.Timestamp values keep their nanoseconds when bound to a DateTime64(9) query parameter, including typed SQLAlchemy bind parameters and values wrapped in DT64Param, and when inserted as rows into DateTime64(7) through DateTime64(9) columns with the Python codec. Previously both paths truncated them to microseconds, so 2024-01-02 03:04:05.123456789 arrived as 2024-01-02 03:04:05.123456000. Python codec row inserts of a pandas.Timestamp whose sub-microsecond part rounds up, such as 03:04:05.999999999, also no longer store a value one second late in any DateTime64 column. Related to #429.
  • Server-side query parameters typed as a named Tuple, Nested, or Map now keep the sub-second precision of datetime values bound to their DateTime64 elements, and each element uses its declared timezone. Map parameters require dict_parameter_format="map". A Variant keeps precision and timezone only when its sole datetime alternative is a DateTime64. Enum labels and field names containing datetime don't count as datetime types. Variants with several datetime alternatives are unchanged.
  • Enum names that contain a backslash or a control character such as a tab now parse correctly from server type names and are escaped correctly in Native insert headers. Previously such names read back escaped, for example c\\d instead of c\d. Inserting them stored 0, which became the wrong label when 0 was a declared member and an unreadable value otherwise. See #1117.
  • The asynchronous client now raises OperationalError when the connection drops while raw_query, command, or a LIMIT 0 metadata query reads the response body, matching the synchronous client. Previously it leaked aiohttp.ClientPayloadError.
  • When the connection drops partway through a streamed query result, both sync and async clients now raise StreamFailureError: Stream failed during read (connection closed by server) and preserve the transport error as the cause. Previously, with show_clickhouse_errors=True, they could report buffered result bytes as unrecognized data found in stream: <hex> and suppress the cause. Recognizable ClickHouse errors written into the response before the drop are still reported.
  • AsyncClient.raw_insert now types insert_block as an async iterable of bytes instead of a sync generator. Async generators already worked but failed type checking. Runtime behavior is unchanged.
  • query_df and query_df_stream with use_none=False no longer raise TypeError on Nullable(Date) and Nullable(Date32), including inside arrays, tuples, maps, and LowCardinality columns. With extended dtypes (the default), NULLs return as NaT, matching use_none=True. NumPy queries with an extended-dtype context and use_none=False also support NULLs and retain datetime64[D] resolution.
  • AsyncClient.query(context=...) now applies the arguments passed with the call, such as parameters, settings, and query, on a copy of the context, matching the synchronous client. Previously the async client ignored them and ran the stored context unchanged, so concurrent calls that shared one context with different parameters all returned the stored context's result.
  • All sync and async query methods that accept a context now preserve its explicitly set tz_mode unless the call overrides it. Contexts without an explicit mode use the calling client's current default, as the sync client already did. Invalid context tz_mode assignments now raise ProgrammingError immediately.
  • query(context=...) now runs a command statement stored in the context, such as CREATE or SET, when no query argument is passed. Previously both the synchronous and asynchronous clients raised AttributeError: 'NoneType' object has no attribute 'rstrip'.
  • Unreadable timezone files and timezone directory names now follow the existing unrecognized-timezone handling. This prevents an import failure when TZ names a directory such as America and the tzdata package is installed.
  • The compiled Cython response buffer no longer segfaults when the first read from a stream is zero bytes, such as a zero-row Array, numeric, FixedString or Nullable(String) column, or an empty String value whose length byte arrived in its own chunk. Normal query results never take this path. The first chunk of a response is always more than one byte and loads the buffer. Only direct use of the buffer and datatype read APIs was affected.

Improvements

  • On Python 3.14 and later, lz4 is no longer installed by default. It publishes no wheels for Python 3.14+ free-threaded builds, CPython 3.15 or musllinux, and zstd is in the standard library from 3.14. Without lz4, compress=True compresses inserts with zstd. Query responses already use zstd because the server prefers it. Install clickhouse-connect[lz4] to use compress="lz4" on 3.14 and later. Nothing changes below 3.14.
  • Cython decoding of Date and Date32 columns is faster because Python date objects are constructed directly through the datetime C API. Results are the same as before.
  • ping() on the sync and async clients accepts an optional positive timeout in seconds, defaulting to 3 seconds when omitted or None. HTTP clients reject zero or negative values with ProgrammingError. The chDB client ignores the parameter. Closes #192.
  • Added get_client(dedicated_pool=True) for synchronous HTTP clients that need their own connection pool. Closing any client-owned pool now closes idle connections immediately and closes active connections when their responses return to the pool. Requests racing with close() can fail with OperationalError, including for existing TLS/proxy-owned pools. The async client accepts the option without effect because it already owns its connector. Closes #1121.
  • Client creation is faster with non-UTC IANA server timezones because timezone validation no longer scans the timezone database. Closes #1120.
  • query_np, query_df, query_np_stream, and query_df_stream no longer raise AttributeError for a query that ends in LIMIT 0. They now return an empty result, the same as a query that matches no rows. Related to #257. Empty results still don't include column names.
  • query_np now returns one row per result row when every column is returned as Python objects and some values are lists or tuples, for example a single Array or Tuple column. Previously numpy unpacked those values into extra dimensions, so SELECT [number, number + 1, number + 2] FROM numbers(2) came back with shape (3, 2, 1) instead of (2, 1). This also fixes a concatenation error for nested Array or Tuple results spread across multiple blocks.
  • query_np and query_df now return every row of a numeric LowCardinality column, such as LowCardinality(Int32). Previously the column was cut to the size of its dictionary, so 1000 rows with 300 distinct values came back as 300 rows, and a block with fewer rows than dictionary entries raised ValueError: iterator too short.
  • AsyncClient.query_rows_stream now reads each result block in the executor once and yields its rows on the event loop. Previously it made one executor call per row, so streaming rows asynchronously was far slower than the synchronous query_rows_stream.
  • The Python codec reads Map, nullable integer, nullable date, and timezone-aware nullable datetime columns faster in query_df by avoiding per-row pandas conversions.
  • With the C extensions, the Python codec reads Array columns faster in all query methods. Related to #1008.
  • Eligible compiled SQLAlchemy Core and ORM SELECT queries now skip the second LIMIT 0 metadata query for empty results and preserve result keys. A projection qualifies when it has a fixed column shape with unique compiled names and every result column uses a result processor that does not read the server type code. This covers SQLAlchemy's generic types whose processors depend only on dialect flags, such as String, Boolean, Numeric, Float, Enum, and types that inherit the default TypeEngine processor, the driver's reflected ChSqlaType types, and TypeDecorator wrappers over those, including nested wrappers, with_variant, and load_dialect_impl. The synthesized DB-API descriptions report None for type_code and null_ok. Plain DB-API and textual queries, raw SQL projections or modifiers, server-expanded columns, custom result_processor overrides, Enum subclasses, JSON, rewritten statements, and cursor execute wrappers retain the LIMIT 0 metadata fallback. Closes #1108 and #1134.

1.11.0, 2026-10-16

Bug Fixes

  • The synchronous HTTP client now waits before retrying HTTP 429, 503, and 504 responses and closes each failed response before the next attempt, matching the asynchronous client.

Improvements

... (truncated)

Commits
  • f04e7c1 release prep for v1.12.0 (#1145)
  • cbd11b4 Fix extended Date32 decoding and speed up Cython date conversion (#1142)
  • 96ef200 Fix crash reading an empty Array column with the C buffer (#1140)
  • 8c5c12f Speed up Array column reads with a Cython row builder (#1138)
  • 5520216 Make lz4 optional on Python 3.14+ (#1139)
  • 6f24bb8 Add optional timeout parameter to ping() (#1137)
  • 93408a0 Keep DateTime64 precision in pandas.Timestamp binds and inserts and in Tuple,...
  • f92d24d Skip empty SQLAlchemy metadata probes for decorators and generic types (#1135)
  • a5a68cf Decode backslash escapes in Enum names from server type names (#1117)
  • 7df5b4c Speed up Map, nullable DateTime/Date and nullable Int columns in query_df (#1...
  • Additional commits viewable in compare view

Dependabot compatibility score

Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting @dependabot rebase.


Dependabot commands and options

You can trigger Dependabot actions by commenting on this PR:

  • @dependabot rebase will rebase this PR
  • @dependabot recreate will recreate this PR, overwriting any edits that have been made to it
  • @dependabot show <dependency name> ignore conditions will show all of the ignore conditions of the specified dependency
  • @dependabot ignore this major version will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)
  • @dependabot ignore this minor version will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)
  • @dependabot ignore this dependency will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)

Bumps [clickhouse-connect](https://github.com/ClickHouse/clickhouse-connect) from 0.15.1 to 1.12.0.
- [Release notes](https://github.com/ClickHouse/clickhouse-connect/releases)
- [Changelog](https://github.com/ClickHouse/clickhouse-connect/blob/main/CHANGELOG.md)
- [Commits](ClickHouse/clickhouse-connect@v0.15.1...v1.12.0)

---
updated-dependencies:
- dependency-name: clickhouse-connect
  dependency-version: 1.12.0
  dependency-type: direct:production
  update-type: version-update:semver-major
...

Signed-off-by: dependabot[bot] <support@github.com>
@dependabot
dependabot Bot requested a review from DevChiniwala as a code owner November 16, 2026 02:37
@dependabot dependabot Bot added dependencies Pull requests that update a dependency file python Pull requests that update python code labels Nov 16, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

dependencies Pull requests that update a dependency file python Pull requests that update python code

Projects

None yet

Development

Successfully merging this pull request may close these issues.

0 participants