fix: DbtProcess/DbtColumnProcess extend correct process parent (SHA-887)

[Aryamanz29]

Same root cause as the DbtMeasure fix in 9.7.0 (#919) — Atlas typedefs list Dbt first in superTypes, the class generator picks super_types[0], and atlan-publish-app's issubclass(cls, Process) / issubclass(cls, ColumnProcess) checks (used to identify lineage types that must publish after the entities they reference) silently return False. The dbt-process classes then land in Batch 1 alongside their upstream entities and ATLAS-404 on every inputs / outputs lookup. ~35 publish failures observed in a live dbt run, plus the same pattern hits SAP BW connectors.

Atlas typedefs

Class	superTypes (Atlas)	Picked by generator	Should pick
DbtProcess	[\"Dbt\", \"Process\"]	Dbt	Process
DbtColumnProcess	[\"Dbt\", \"ColumnProcess\"]	Dbt	ColumnProcess

Sibling type DbtMeasure was fixed via the same mechanism in 9.7.0 — this PR uses the same _SUPERCLASS_OVERRIDES map.

Changes

• pyatlan/generator/class_generator.py — extend _SUPERCLASS_OVERRIDES with \"DbtProcess\": \"Process\" and \"DbtColumnProcess\": \"ColumnProcess\". The override only applies when the named supertype is actually declared on the entity, so a future typedef change that drops it cleanly errors rather than silently reverting.
• pyatlan/model/assets/dbt_process.py — class DbtProcess(Process) and class Attributes(Process.Attributes).
• pyatlan/model/assets/dbt_column_process.py — same shape against ColumnProcess.

Blast radius

Same as the DbtMeasure fix:

• isinstance(x, Dbt) on a DbtProcess / DbtColumnProcess was True, is now False. That check was incorrect — Atlas does not model DbtProcess as a Dbt subtype in a way that's meaningful for ordering. A grep across pyatlan and its tests finds no such isinstance usage.
• Serialised JSON (type_name, attribute keys) is unchanged — every dbt_* and Process / ColumnProcess field is declared inline on the respective Attributes classes, so no fields are lost; only the MRO is corrected.

Tests

• tests/unit/model/dbt_process_test.py — 7 cases (issubclass guards both directions, __bases__[0], type_name immutability, dbt_* and Process attribute round-trip on the same instance).
• tests/unit/model/dbt_column_process_test.py — same 7 cases for the ColumnProcess parent.

Verified

• ./qa-checks — ruff format / ruff check / mypy all pass
• pytest tests/unit/model/dbt_process_test.py tests/unit/model/dbt_column_process_test.py -q — 14/14 pass
• Full unit suite clean (no new failures)
• Downstream verification in atlan-publish-app once this ships: DbtProcess / DbtColumnProcess move to a strictly later batch than the entities they reference

Notes

Release notes will land in the next version bump (HISTORY.md is left alone here so 9.7.0's already-shipped section isn't retroactively edited). The typing.get_type_hints shim that was bundled into the closed PR #922 has been dropped — the better fix is on the publish-app side (switching to pydantic v1's ModelField.type_ / .shape, which are already fully resolved at SDK import time via update_forward_refs). A separate publish-app PR will land that change.

🤖 Generated with Claude Code