Db2 backend ibis clean by sma-2004 · Pull Request #11 · Amitkumar293/ibis

sma-2004 · 2026-10-13T10:42:55Z

Description of changes

Issues closed

Signed-off-by: hdm-db2-eco-system <Hdm-dev-persona-db2-eco-system@ibm.com>

…formatting cleanup

This test was added to main after the db2 branch was originally cut (commit 5f1dff0). DB2 has no STR_TO_DATE function and requires a FROM clause for scalar SELECT, so both param variants fail at runtime. Mark them notimpl with IbmDb2Error so CI skips rather than fails.

1.0.0 had sqlglot>=30.8.0,<30.10.0 which caps sqlglot for the entire project. 1.1.0 drops the upper bound. Set the minimum to 1.1.0 so the broken version can never be resolved.

…lect >=1.1.0

DB2 has no STR_TO_DATE function so both as_date and as_timestamp with format string fail at runtime with SQL0440N.

…rmat DB2 supports as_timestamp with format (passes), but not as_date (STR_TO_DATE doesn't exist). Scope the notimpl marker to date only.

The exclude for DataFusion on windows-latest with Python 3.14 was added alongside the Db2 matrix entries but is unrelated to Db2. Restore the upstream behaviour so DataFusion keeps running on that combination.

No directory named temp exists in the repo, so this entry has no effect. Removing it avoids a global pytest config change in a backend-specific PR.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

## Description of changes `ibis.to_sql` looks a dialect name up as an attribute of `ibis.backends.sql.compilers`, but that package never imported the `singlestoredb` compiler module. So `ibis.to_sql(expr, dialect="singlestoredb")` raised `ValueError: Unknown dialect singlestoredb` unless something had imported `ibis.backends.singlestoredb` first. Our backend test jobs always import the backend, which is why CI never caught it. This imports the module and adds `SingleStoreDBCompiler` to `__all__`, matching every other SQL backend. The new test compiles a trivial table for each SQL backend name, so a future backend that misses this list fails in the core test job. Found while moving compile-only tests around in ibis-project#12153. 🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

…12152) This PR contains the following updates: | Package | Update | Change | |---|---|---| | apache/impala | patch | `4.5.0-impalad_coord_exec` → `4.5.3-impalad_coord_exec` | | apache/impala | patch | `4.5.0-catalogd` → `4.5.3-catalogd` | | apache/impala | patch | `4.5.0-statestored` → `4.5.3-statestored` | | apache/impala | patch | `4.5.0-impala_quickstart_hms` → `4.5.3-impala_quickstart_hms` | --- ### Configuration 📅 **Schedule**: (UTC) - Branch creation - "after 10pm every weekday,every weekend,before 5am every weekday" - Automerge - At any time (no schedule defined) 🚦 **Automerge**: Enabled. ♻ **Rebasing**: Whenever PR is behind base branch, or you tick the rebase/retry checkbox. 🔕 **Ignore**: Close this PR and you won't be reminded about these updates again. --- - [ ] If you want to rebase/retry this PR, check this box --- This PR was generated by [Mend Renovate](https://mend.io/renovate/). View the [repository job log](https://developer.mend.io/github/ibis-project/ibis).  Co-authored-by: renovate[bot] <29139614+renovate[bot]@users.noreply.github.com>

…s-project#12041) ## Description of changes `translate()` and `add_query_to_expr()` rebuild the `WITH` clause from its CTEs and drop the original `With` node, which lost the `RECURSIVE` flag. The flag is now passed to `with_(recursive=...)` when the clause is rebuilt. In `add_query_to_expr()` it's kept if either the user's query or the ibis expression had it, so a recursive CTE coming from the table passed to `Table.sql()` also works. CTE aliases are now built as `TableAlias` instead of a bare identifier, because sqlglot's Trino/Presto generator reads `alias.columns` when the clause is recursive. The same change is applied to SingleStoreDB's copy of `add_query_to_expr()`. On MSSQL this makes CTE projections explicit (`[t0].[key] AS [key]`), so one snapshot changed, with the same meaning. Known limitations: BigQuery doesn't allow a nested `WITH RECURSIVE`; Oracle rejects the `RECURSIVE` keyword, which sqlglot still emits; and on Trino/Athena a recursive query over a plain table still renders the CTE alias as `"t"("*")`, which already failed on main. ## Issues closed * Resolves ibis-project#11949 * Resolves ibis-project#11922 --------- Co-authored-by: Deepyaman Datta <deepyaman.datta@utexas.edu>

- import the external dialect unconditionally; remove the stub registered in ibis/backends/sql/dialects.py - put Db2Generator first in the MRO and merge TRANSFORMS/TYPE_MAPPING on top of the external dialect instead of spreading the base Generator dicts - remove dead and incorrect overrides (tablesample_sql, dateadd_sql, visit_Sample, div_sql, log_sql) - lower Log2/Log10 and implement visit_Log with LN/LOG10 (Db2 has no two-argument LOG) - build CONCAT with || since Db2 CONCAT takes two arguments - remove db2 xfail marks for log and integer true-division tests

The external dialect's extract_sql renders DAYOFWEEK/DAYOFYEAR as function calls; our override emitted EXTRACT(dayofyear FROM ...), which Db2 rejects (SQL0104N). Also update the test_cte_refs_in_topo_order snapshot for the intended WHERE 1 -> WHERE TRUE change.

- add UNSUPPORTED_OPS (42 ops with no Db2 equivalent) so they fail at compile time with OperationNotDefinedError, the same as other backends, instead of emitting SQL that references nonexistent functions - delete the hand-written raise-only visit_ methods this replaces, and the SIMPLE_OPS exclusions that only protected them (Unnest's raise was being silently overwritten by __init_subclass__) - implement RandomUUID with LOWER(GENERATE_UUID()), a built-in v4 UUID function in Db2 12.1, instead of GENERATE_UNIQUE() which is not a UUID - update test marks to raises=OperationNotDefinedError; drop the stale test_uuid_function xfail; regenerate the impure-ops snapshot

… revert codespell list - compose.yaml: pin icr.io/db2_community/db2 to 12.1.5.0 (was :latest); GENERATE_UUID() is verified on 12.1.5.0 - pyproject.toml/uv.lock: db2 extra pandas<3 -> <4 like every other extra; remove packaging (not imported by the backend) - .codespellrc: restore upstream ignore-words-list (the extra words do not appear in the PR) - delete stale test_union_aliasing/db2 snapshot (contained EXPLODE and ANY_VALUE; the test is notimpl for db2)

…bis-project#12163) This PR contains the following updates: | Package | Type | Update | Change | |---|---|---|---| | [actions/create-github-app-token](https://redirect.github.com/actions/create-github-app-token) | action | minor | `v3.2.0` → `v3.3.0` | --- ### Release Notes <details> <summary>actions/create-github-app-token</summary> ### [`v3.3.0`](https://redirect.github.com/actions/create-github-app-token/releases/tag/v3.3.0) [Compare Source](https://redirect.github.com/actions/create-github-app-token/compare/v3.2.0...v3.3.0) ##### Features - update permission inputs ([#&ibis-project#8203;399](https://redirect.github.com/actions/create-github-app-token/issues/399)) ([a47236f](https://redirect.github.com/actions/create-github-app-token/commit/a47236f79f79fcbc3532af49fe3a5ded89529b6e)) ##### Bug Fixes - **deps:** bump nested undici from 6.24.0 to 6.29.0 ([#&ibis-project#8203;403](https://redirect.github.com/actions/create-github-app-token/issues/403)) ([12461a3](https://redirect.github.com/actions/create-github-app-token/commit/12461a3a794e67c032f1eff5bb95ece9427eb4fb)) - **deps:** bump the production-dependencies group across 1 directory with 3 updates ([#&ibis-project#8203;395](https://redirect.github.com/actions/create-github-app-token/issues/395)) ([a69c19d](https://redirect.github.com/actions/create-github-app-token/commit/a69c19ddfdf8e1ab093900e0064524ce9309c807)) </details> --- ### Configuration 📅 **Schedule**: (UTC) - Branch creation - "after 10pm every weekday,every weekend,before 5am every weekday" - Automerge - At any time (no schedule defined) 🚦 **Automerge**: Enabled. ♻ **Rebasing**: Whenever PR is behind base branch, or you tick the rebase/retry checkbox. 🔕 **Ignore**: Close this PR and you won't be reminded about this update again. --- - [ ] If you want to rebase/retry this PR, check this box --- This PR was generated by [Mend Renovate](https://mend.io/renovate/). View the [repository job log](https://developer.mend.io/github/ibis-project/ibis).  Co-authored-by: renovate[bot] <29139614+renovate[bot]@users.noreply.github.com>

This PR contains the following updates: | Package | Type | Update | Change | |---|---|---|---| | [flink](https://hub.docker.com/_/flink) ([source](https://redirect.github.com/apache/flink-docker)) | final | patch | `1.20.3` → `1.20.5` | --- ### Configuration 📅 **Schedule**: (UTC) - Branch creation - "after 10pm every weekday,every weekend,before 5am every weekday" - Automerge - At any time (no schedule defined) 🚦 **Automerge**: Enabled. ♻ **Rebasing**: Whenever PR is behind base branch, or you tick the rebase/retry checkbox. 🔕 **Ignore**: Close this PR and you won't be reminded about this update again. --- - [ ] If you want to rebase/retry this PR, check this box --- This PR was generated by [Mend Renovate](https://mend.io/renovate/). View the [repository job log](https://developer.mend.io/github/ibis-project/ibis).  Co-authored-by: renovate[bot] <29139614+renovate[bot]@users.noreply.github.com>

db2-sqlglot-dialect is an optional extra, but compilers/__init__.py imported every compiler on `import ibis`. Without the extra, every backend job failed with ModuleNotFoundError: No module named 'db2_sqlglot' while loading conftest. Resolve `db2` and `Db2Compiler` through a module-level __getattr__, so the optional package is only imported on access. A missing package now raises ImportError with an install hint. Other missing modules and other names are unchanged. Db2 is the only compiler loaded this way, because lookups such as to_sql(dialect=...) and test_sql.py resolve compilers by name on this module. Do not turn this back into a top-level import. Known follow-up: non-Db2 jobs fail on the db2 dialect params in test_dot_sql.py (sqlglot: "Unknown dialect 'db2'") because the lazy import now lets them reach those tests. The skip is in a separate commit; db2-parametrized tests without a job marker are unchanged.

test_table_dot_sql_transpile and test_con_dot_sql_transpile are parametrized over every backend name as a sqlglot dialect, so every backend job gets a `[<backend>-db2]` case. Those tests call sqlglot directly with dialect="db2", which is only registered when the optional db2-sqlglot-dialect package (the `db2` extra) is installed. Without it they fail with ValueError: Unknown dialect 'db2'. Skip the db2 dialect case only, and only when db2_sqlglot cannot be imported. Other backends and the Db2 job (which installs the extra) are unaffected.

Prepare the statement and read the result-set metadata with ibm_db.field_type, field_precision, field_scale and field_nullable instead of returning VARCHAR for every column. Wrap prepare failures in ibm_db_dbi.ProgrammingError, as cursor.execute does. Remove the hand-written ibis/backends/db2/datatypes.py and its tests, which are no longer used.

test_cte, test_unnamed_columns and test_scalar_dot_sql pass now that _get_schema_using_query returns real types.

The driver defers statement preparation by default, so syntax errors and undefined names in non-SELECT statements were silently accepted and an empty schema was returned. Disable deferred prepare so these raise ibm_db_dbi.ProgrammingError again.

…rations

Catalog views are visible on the same connection after a committed DDL statement, and the connection now goes through ibm_db_dbi.connect with explicit autocommit, so the reconnect is not needed. Remove _reconnect() and its callers.

SYSCAT.TABLES reports views as V and global temporary tables as G, so the TYPE = 'T' filter hid both.

Create memtables as GLOBAL TEMPORARY tables instead of permanent ones, now that create_table no longer reconnects, and drop the per-call list_tables() existence check since the base class already tracks registered memtables. Raise IbisTypeError for null-typed columns, as the Oracle and Postgres backends do, and update the matching test marks.

Table names were upper-cased on the way in (table, create_table, insert, drop_table, drop_view and memtable registration) and lower-cased on the way out of list_tables, so names did not round-trip and raw SQL against a quoted lowercase table failed. Quote identifiers everywhere, as Oracle does, and quote the table names in ci/schema/db2.sql.

Raw SQL against quoted lowercase tables, upsert (MERGE INTO), views in list_tables, and to_sql with the default backend now work on Db2, so the notimpl/notyet marks for them are stale.

The database argument was accepted but ignored; the query always used the current schema.

Generate the statement through SQLGlot and schema.to_sqlglot_column_defs instead of concatenating strings, as the Oracle backend does, so column types in DDL go through the same dialect as CAST. Emits identical SQL.

This PR contains the following updates: | Package | Update | Change | |---|---|---| | [postgres](https://hub.docker.com/_/postgres) ([source](https://redirect.github.com/docker-library/postgres)) | minor | `18.4-alpine` → `18.6-alpine` | --- ### Configuration 📅 **Schedule**: (UTC) - Branch creation - "after 10pm every weekday,every weekend,before 5am every weekday" - Automerge - At any time (no schedule defined) 🚦 **Automerge**: Enabled. ♻ **Rebasing**: Whenever PR is behind base branch, or you tick the rebase/retry checkbox. 🔕 **Ignore**: Close this PR and you won't be reminded about this update again. --- - [ ] If you want to rebase/retry this PR, check this box --- This PR was generated by [Mend Renovate](https://mend.io/renovate/). View the [repository job log](https://developer.mend.io/github/ibis-project/ibis).  Co-authored-by: renovate[bot] <29139614+renovate[bot]@users.noreply.github.com>

…bis-project#12161) ## Description of changes The MSSQL dialect mapped a bare `sge.Stddev` to `STDEVP`, but T-SQL's `STDEV` is the sample standard deviation, so `STDEV` in user SQL was silently turned into a population one. Oracle had the same problem: `STDDEV` became `STDDEV_POP`, even though Oracle's `STDDEV` is sample. This shows up through `Table.sql()` and `con.sql(..., dialect=...)`, not `Backend.sql()` without a dialect. It also affected cross-dialect queries, e.g. a DuckDB or Postgres `STDDEV` sent to MSSQL or Oracle came back as the population value. This removes the `sge.Stddev` overrides for MSSQL and Oracle, so sqlglot's defaults are used, which already render `STDEV` and `STDDEV`. Since the Oracle override was applied to sqlglot's global Oracle dialect, removing it also fixes plain `sqlglot.transpile(..., write="oracle")` for anyone who imported ibis. SQL generated from ibis expressions like `t.x.std()` doesn't change, since that always goes through `stddev_samp`/`stddev_pop`. Tests: `test_dot_sql_transpile_stddev` in `test_dot_sql.py` runs DuckDB SQL over a few rows on every backend and compares against the sample standard deviation (with markers for SQLite, Druid, RisingWave and BigQuery, where the query can't run as written). MySQL isn't changed here. sqlglot parses MySQL's `STDDEV`/`VARIANCE` (population) into the same nodes as DuckDB's sample versions, so fixing that needs a change in sqlglot (tobymao/sqlglot#8626). MySQL and SingleStoreDB are marked `notyet` in the new test for now. ## Issues closed * Resolves ibis-project#12057 --------- Co-authored-by: Deepyaman Datta <deepyaman.datta@utexas.edu>

The Db2 image runs every script in /var/custom after the database is created. Mount docker/db2/tablespaces.sh there so the tablespaces needed for GLOBAL TEMPORARY tables exist before tests run, instead of creating them from the test conftest. The healthcheck now waits for them.

Tables are created with quoted lowercase names, so raw SQL must quote them too. Fixes test_client.py::test_sql on a fresh database.

db2-sqlglot-dialect documents support for Python 3.10 - 3.12 only.

…2179) This PR contains the following updates: | Package | Type | Update | Change | |---|---|---|---| | [actions/setup-python](https://redirect.github.com/actions/setup-python) | action | major | `v6` → `v7` | --- ### Release Notes <details> <summary>actions/setup-python</summary> ### [`v7.0.0`](https://redirect.github.com/actions/setup-python/releases/tag/v7.0.0) [Compare Source](https://redirect.github.com/actions/setup-python/compare/v6.3.0...v7.0.0) #### What's Changed ##### Enhancements - Migrate to ESM and upgrade dependencies by [@&ibis-project#8203;priyagupta108](https://redirect.github.com/priyagupta108) in [#&ibis-project#8203;1330](https://redirect.github.com/actions/setup-python/pull/1330) - Pin SHA commits and update docs with latest versions by [@&ibis-project#8203;HarithaVattikuti](https://redirect.github.com/HarithaVattikuti) in [#&ibis-project#8203;1338](https://redirect.github.com/actions/setup-python/pull/1338) - Remove the pip-install input by [@&ibis-project#8203;gowridurgad](https://redirect.github.com/gowridurgad) in [#&ibis-project#8203;1336](https://redirect.github.com/actions/setup-python/pull/1336) ##### Bug Fix - Fix to Classify stderr warning messages as warnings instead of errors in annotations by [@&ibis-project#8203;lmvysakh](https://redirect.github.com/lmvysakh) in [#&ibis-project#8203;1335](https://redirect.github.com/actions/setup-python/pull/1335) - Validate and retry manifest fetch to prevent silent failures by [@&ibis-project#8203;priyagupta108](https://redirect.github.com/priyagupta108) in [#&ibis-project#8203;1332](https://redirect.github.com/actions/setup-python/pull/1332) ##### Dependency Upgrade - Bump certifi from 2020.6.20 to 2024.7.4 in /**tests**/data by [@&ibis-project#8203;dependabot](https://redirect.github.com/dependabot) in [#&ibis-project#8203;1328](https://redirect.github.com/actions/setup-python/pull/1328) - Remove EOL Python versions and Bumps numpy text fixture by [@&ibis-project#8203;priya-kinthali](https://redirect.github.com/priya-kinthali) in [#&ibis-project#8203;1333](https://redirect.github.com/actions/setup-python/pull/1333) - Upgrade [@&ibis-project#8203;actions/cache](https://redirect.github.com/actions/cache) to 6.2.0 by [@&ibis-project#8203;philip-gai](https://redirect.github.com/philip-gai) in [#&ibis-project#8203;1337](https://redirect.github.com/actions/setup-python/pull/1337) #### New Contributors - [@&ibis-project#8203;lmvysakh](https://redirect.github.com/lmvysakh) made their first contribution in [#&ibis-project#8203;1335](https://redirect.github.com/actions/setup-python/pull/1335) - [@&ibis-project#8203;philip-gai](https://redirect.github.com/philip-gai) made their first contribution in [#&ibis-project#8203;1337](https://redirect.github.com/actions/setup-python/pull/1337) **Full Changelog**: <actions/setup-python@v6...v7.0.0> </details> --- ### Configuration 📅 **Schedule**: (UTC) - Branch creation - "after 10pm every weekday,every weekend,before 5am every weekday" - Automerge - At any time (no schedule defined) 🚦 **Automerge**: Enabled. ♻ **Rebasing**: Whenever PR is behind base branch, or you tick the rebase/retry checkbox. 🔕 **Ignore**: Close this PR and you won't be reminded about this update again. --- - [ ] If you want to rebase/retry this PR, check this box --- This PR was generated by [Mend Renovate](https://mend.io/renovate/). View the [repository job log](https://developer.mend.io/github/ibis-project/ibis).  Co-authored-by: renovate[bot] <29139614+renovate[bot]@users.noreply.github.com>

…cursive CTEs - STDDEV renders as STDDEV_SAMP (Db2's STDDEV is the population one) - cast boolean to string renders 'true'/'false' - drop the RECURSIVE keyword and name the columns of recursive CTEs, which Db2 requires

This PR contains the following updates: | Package | Update | Change | |---|---|---| | apache/druid | major | `36.0.0` → `38.0.0` | --- ### Configuration 📅 **Schedule**: (UTC) - Branch creation - "after 10pm every weekday,every weekend,before 5am every weekday" - Automerge - At any time (no schedule defined) 🚦 **Automerge**: Enabled. ♻ **Rebasing**: Whenever PR is behind base branch, or you tick the rebase/retry checkbox. 🔕 **Ignore**: Close this PR and you won't be reminded about this update again. --- - [ ] If you want to rebase/retry this PR, check this box --- This PR was generated by [Mend Renovate](https://mend.io/renovate/). View the [repository job log](https://developer.mend.io/github/ibis-project/ibis).  Co-authored-by: renovate[bot] <29139614+renovate[bot]@users.noreply.github.com>

…ging extra Db2 `a - b` on dates, timestamps and times returns a packed-decimal duration, not a count, so date/timestamp subtraction silently returned wrong values (a 1-day gap came back as 11 days 13:46:40). Map DateDiff to DAYS_BETWEEN, TimestampDiff to SECONDS_BETWEEN and TimeDiff to MIDNIGHT_SECONDS differences. Date literals compiled to DATE_FROM_PARTS, which Db2 does not have; emit DATE('YYYY-MM-DD'). Interval operands compiled to INTERVAL 'n' UNIT, which Db2 rejects; emit labeled durations `(n) DAYS` (week -> 7 days, quarter -> 3 months, millisecond -> 1000 microseconds, nanosecond unsupported). DateDelta and TimestampDelta ignored their unit and always returned DAYS_BETWEEN. Truncate both sides to the unit and use the matching *_BETWEEN function. TimeDelta is unsupported since the *_BETWEEN functions reject TIME. Remove the db2 notimpl marks these changes make stale, and add `packaging` back to the db2 extra: ibis/backends/sql/compilers/datafusion.py imports it at module level, so `import ibis` fails after installing the extra without it (upstream main has the same behaviour in a clean environment).

Db2's VARIANCE is the population variance, so Variance silently returned the wrong value, and VARIANCE_POP does not exist. Render the sample and population forms with VAR_SAMP and VAR_POP, and remove the two db2 marks this makes stale.

Db2 does not accept a bare TRUNCATE TABLE; IMMEDIATE is required, so the base implementation failed with SQL0104N. Add a Db2 truncate_table and use it from insert(overwrite=True).

_get_schema_using_query lower-cased result column names, but Db2 stores an unquoted alias in upper case, so con.sql(...) followed by any further expression referenced a column that does not exist (SQL0206N). Return the names verbatim, as get_schema does, and document the string length, NaN, division by zero and literal alias limits.

hdm-db2-eco-system added 10 commits October 13, 2026 12:05

feat(db2): add IBM Db2 backend

2a3383e

Signed-off-by: hdm-db2-eco-system <Hdm-dev-persona-db2-eco-system@ibm.com>

fix(db2): CI test markers, backend fixes, tablespace setup, lint and …

f8bc8cd

…formatting cleanup

fix(db2): bump db2-sqlglot-dialect lower bound to >=1.1.0

23f972a

1.0.0 had sqlglot>=30.8.0,<30.10.0 which caps sqlglot for the entire project. 1.1.0 drops the upper bound. Set the minimum to 1.1.0 so the broken version can never be resolved.

fix(deps): sync uv.lock specifier with pyproject.toml db2-sqlglot-dia…

9be6140

…lect >=1.1.0

fix(db2): mark test_string_as_temporal_with_format as notimpl

9b37567

DB2 has no STR_TO_DATE function so both as_date and as_timestamp with format string fail at runtime with SQL0440N.

style(db2): fix ruff-format line length in notimpl marker

a58e4ea

fix(db2): notimpl only date param for test_string_as_temporal_with_fo…

0d8ffa2

…rmat DB2 supports as_timestamp with format (passes), but not as_date (STR_TO_DATE doesn't exist). Scope the notimpl marker to date only.

ci: drop unrelated datafusion windows exclude

980d863

The exclude for DataFusion on windows-latest with Python 3.14 was added alongside the Db2 matrix entries but is unrelated to Db2. Restore the upstream behaviour so DataFusion keeps running on that combination.

ci: remove **/temp from norecursedirs

5f87342

No directory named temp exists in the repo, so this entry has no effect. Removing it avoids a global pytest config change in a backend-specific PR.

github-actions Bot added sql tests docs nix ci dependencies labels Oct 13, 2026

hdm-db2-eco-system and others added 14 commits October 13, 2026 18:48

ci: trigger re-run

03933b1

docs: say "window function" instead of "analytic function" in prose

63d885f

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

docs(ir): describe what ops.WindowFunction represents

eb78c9a

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

hdm-db2-eco-system and others added 21 commits October 18, 2026 01:11

test(db2): remove notimpl marks that pass with real schema inference

6cc89dc

test_cte, test_unnamed_columns and test_scalar_dot_sql pass now that _get_schema_using_query returns real types.

docs(db2): document tablespace requirement, unsupported types and ope…

5d31cfd

…rations

fix(db2): stop reconnecting after create_table

4844332

Catalog views are visible on the same connection after a committed DDL statement, and the connection now goes through ibm_db_dbi.connect with explicit autocommit, so the reconnect is not needed. Remove _reconnect() and its callers.

fix(db2): list views and global temporary tables in list_tables

b47abec

SYSCAT.TABLES reports views as V and global temporary tables as G, so the TYPE = 'T' filter hid both.

test(db2): remove marks that pass with quoted identifiers

f4a53cd

Raw SQL against quoted lowercase tables, upsert (MERGE INTO), views in list_tables, and to_sql with the default backend now work on Db2, so the notimpl/notyet marks for them are stale.

fix(db2): honour the database argument in list_tables

403493b

The database argument was accepted but ignored; the query always used the current schema.

refactor(db2): build CREATE TABLE with sge.Create

cc9b3ec

Generate the statement through SQLGlot and schema.to_sqlglot_column_defs instead of concatenating strings, as the Oracle backend does, so column types in DDL go through the same dialect as CAST. Emits identical SQL.

test(db2): quote table names in raw SQL tests

5a5e818

Tables are created with quoted lowercase names, so raw SQL must quote them too. Fixes test_client.py::test_sql on a fresh database.

docs(db2): document why Python 3.14 is excluded from CI

32c5346

db2-sqlglot-dialect documents support for Python 3.10 - 3.12 only.

Merge upstream/main into db2-backend-ibis-clean

65dc79e

github-actions Bot added the impala label Oct 18, 2026

renovate Bot and others added 8 commits October 18, 2026 09:02

Merge remote-tracking branch 'upstream/main' into db2-backend-ibis-clean

1c32266

fix(db2): use VAR_SAMP and VAR_POP for variance

a195678

Db2's VARIANCE is the population variance, so Variance silently returned the wrong value, and VARIANCE_POP does not exist. Render the sample and population forms with VAR_SAMP and VAR_POP, and remove the two db2 marks this makes stale.

fix(db2): add truncate_table

8dd644c

Db2 does not accept a bare TRUNCATE TABLE; IMMEDIATE is required, so the base implementation failed with SQL0104N. Add a Db2 truncate_table and use it from insert(overwrite=True).

test(db2): sort frames before comparing, Db2 row order is unspecified

d9ee0d2

Navigation Menu

Repository navigation

Db2 backend ibis clean - #11

Db2 backend ibis clean#11
sma-2004 wants to merge 78 commits into
mainfrom
db2-backend-ibis-clean

sma-2004 commented Oct 13, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

7 participants

Conversation

sma-2004 commented Oct 13, 2026

Description of changes

Issues closed

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

7 participants