Skip to content

adbcBridge v0.1.2

Choose a tag to compare

@github-actions github-actions released this 15 Sep 05:26
· 57 commits to main since this release
17fe5a1

adbcBridge 0.1.2

One behaviour change on PostgreSQL. Columns declared with reduced precision now read at that precision: TIMESTAMP(0) / TIMESTAMPTZ(0) as timestamp[s], TIMESTAMPTZ(1..3) as timestamp[ms, tz=UTC], TIME(0) as time32[s], TIME(1..3) as time32[ms]. Values are unchanged; only the Arrow unit is, and it now matches the column. Code that assumed every PostgreSQL timestamp arrives as microseconds sees the declared unit instead. Every other database keeps its exact behaviour. (#107)

Also on PostgreSQL (#107): bulk ingest creates the column the Arrow type says, TIMESTAMP(n) for a zone-less timestamp and TIMESTAMP(n) WITH TIME ZONE for a zoned one, TIME(n) for every time unit; adbc.connection.db_schema answers the current schema (also on DuckDB); GetObjects at schema depth names the catalog. The ADBC Driver Foundry validation suite now passes 293 of 327 on PostgreSQL, the remaining 34 being declared PostgreSQL limits such as nanosecond precision; details in tests/validation/RESULTS.md.

Windows fix (#109): PostgreSQL bulk ingest bound its array literals as narrow text, so non-ASCII text stored through psqlodbc on Windows was double-encoded (héllo landed as héllo). The literals are now bound wide, as the row path always was. psqlodbc also receives LFConversion=0 on Windows, so fetched text keeps the newlines the server stored instead of CR LF.

Verified on Linux, macOS and Windows before the tag (PR descriptions carry the per-platform runs). A PostgreSQL job now runs in CI against a real server on every push.


What's Changed

  • docs: ArcadeDB #7178 in the upstream tracker (maintainer-opened, from the native-probe follow-up) by @singhpratech in #79
  • docs: ArcadeDB #7178 — fix PR #7179 open, verified on its CI build by @singhpratech in #80
  • docs: ArcadeDB #7178 row — fix build re-measured, use_copy=false reads work, #7180 by @singhpratech in #81
  • docs: Arrow PostgreSQL driver row — use_copy=false re-measure on four wire-compatible servers by @singhpratech in #82
  • docs: ArcadeDB #7178 row — fix PR #7179 merged (release pending) by @singhpratech in #84
  • docs: ArcadeDB row — native-driver status after the #7178 fix by @singhpratech in #85
  • docs: ArcadeDB #7178 fix ships in 26.10.1 (in 26.10.1-SNAPSHOT now) by @singhpratech in #86
  • validation: PostgreSQL backend + Linux re-run of the Foundry suite (SQLite 212/327, PostgreSQL 265/327) by @singhpratech in #87
  • docs: ArcadeDB GetTableSchema fixed in #7187; #7188 (binary COPY) tracker row by @singhpratech in #88
  • docs: Firebird ODBC #300 confirmed on Windows, widened to NUMERIC/DECIMAL upstream by @singhpratech in #89
  • docs: Firebird ODBC #300/#301 fixed upstream in PRs #302/#303, verified on Linux by @singhpratech in #90
  • docs: Firebird ODBC #299 fixed upstream in PR #308, verified on Linux (all three findings fixed) by @singhpratech in #91
  • docs: psqlodbc #207 verified on the merge commit; follow-up #211 verified (UPSTREAM row) by @singhpratech in #92
  • docs: psqlodbc #207 and #208 fixed in 18.00.0003, verified on the release build by @singhpratech in #101
  • docs: ArcadeDB #7188 implemented in #7398 (26.10.1), verified on the snapshot; Doris #67306 conflicts resolved by @singhpratech in #102
  • docs: ArcadeDB #7188 framing finding reported on the issue by @singhpratech in #103
  • docs: Materialize #38606 merged, verified on its main image (psqlodbc #208 pair closed) by @singhpratech in #104
  • docs: tracker — ArcadeDB #7188 reopened and in progress; Materialize #38606 results posted by @singhpratech in #105
  • docs: ArcadeDB #7188 fixed on main (#7607), native driver passes with defaults on 26.10.1-SNAPSHOT by @singhpratech in #106
  • docs: tracker — ArcadeDB #7188 closed by the maintainer, re-verified on the snapshot by @singhpratech in #108
  • Honour declared temporal precision and zones on PostgreSQL; db_schema option; schema catalog fill (validation P1-P4, P7, P8) by @singhpratech in #107
  • ingest: bind PostgreSQL array literals wide; LFConversion=0 for psqlodbc on Windows by @singhpratech in #109
  • release: 0.1.2 by @singhpratech in #110

Full Changelog: v0.1.1...v0.1.2