make "schema.log"."target_id" column optional

[n1ru4l]

Background

Prerequisite for #8031; Continuation of efforts from #8082

Description

#8031 will introduce that a schema_log can be associated with ANY schema version within the same project.

With the current database schema, deleting a target triggers a cascade delete of its associated schema versions and schema_logs. Since schema logs can be shared across multiple schema versions, this may unintentionally remove references that are still required by other versions (which will be introduced in #8031), leaving the remaining schema version relationships inconsistent.

To prevent this, this PR introduces the following changes:

• Update the insert logic to stop persisting target_id on schema_log, as it is no longer used by any relevant business logic.
• Make schema_log.target_id nullable and remove its foreign key constraint to prevent cascade deletions when a target is deleted

Drawback / Consideration

This change introduces a minor concern.

A target can have multiple schema versions, and each schema version can reference multiple schema log entries through the schema_version_to_log relation. In addition, a project directly owns many schema log entries.

When a target is deleted, all associated schema versions are removed via cascade deletion. However, the related schema_log records are not automatically deleted, since they may also be referenced independently at the project level.

As a result, this can lead to orphaned schema_log entries that are no longer associated with any schema version.

To mitigate this, we could introduce a periodic cleanup process (for example, a cron job) in the future to identify and remove unreferenced schema log records.

I added this to a follow up issue #8052

[gemini-code-assist]

Code Review

This pull request removes the targetId / target_id field from schema log entities, including proposals, push logs, delete logs, and base schema definitions. However, two critical issues were identified: first, targetId is still being passed to insertPushSchemaLog at its call site in createSchemaVersion, leading to a TypeScript compilation error; second, in deleteSubgraphFromTarget, target_id was removed from the INSERT columns but remains in the RETURNING clause and the subsequent Zod parser, which will cause a runtime validation failure.

[github-actions]

🚀 Snapshot Release (alpha)

The latest changes of this PR are available as alpha on npm (based on the declared changesets):

Package	Version	Info
@graphql-hive/cli	0.59.3-alpha-20260529123937-4ccbe160054a20fd649e11ac7235caf549d4af67	npm ↗︎ unpkg ↗︎
hive	11.2.0-alpha-20260529123937-4ccbe160054a20fd649e11ac7235caf549d4af67	npm ↗︎ unpkg ↗︎

[github-actions]

🐋 This PR was built and pushed to the following Docker images:

Targets: build

Platforms: linux/arm64

Image Tag: 4ccbe160054a20fd649e11ac7235caf549d4af67