Helo#31
Merged
Merged
Conversation
Only migrate loaded the project config; the other database-touching commands
read this.app.getConfig('adapter') directly. Under harnesses that hand the
command a fresh CliApp without applying the config (e.g. the Arkstack CLI),
the adapter was undefined — so migrate:fresh silently took the Prisma-schema
path instead of resetting the configured (Kysely) database, i.e. it did not
tear anything down. rollback/seed/models:sync/migrate:history had the same
latent gap.
Each now calls loadArkormConfig() first (idempotent; matches migrate).
migrate:fresh itself is confirmed correct: adapter.resetDatabase() drops all
base tables and enum types in the current schema, then every migration is
re-run from a clean state.
Test: migrate:fresh reaches the adapter reset path when the adapter is only
reachable through arkormx.config (no configureArkormRuntime).
A migration's up()/down() is invoked both to gather its schema operations
and, when it uses DB.raw() directly, as a real side effect. Several flows
re-invoke up() purely to (re)build column-mapping / feature metadata:
MigrateCommand's per-migration plan, and syncPersistedColumnMappingsFromState
/ validatePersistedMetadataFeaturesForMigrations. On rollback, sync re-planned
every *still-applied* migration's up(), replaying its raw SQL (e.g. an ADD
CONSTRAINT) against a schema it was never meant to touch again — throwing
'already exists' and leaving state half-updated so later commands also failed.
migrate survived only because the raw ran twice but idempotently.
- Add a migration-planning context: while active, DB.raw is a no-op. It is
anchored on globalThis via Symbol.for so the CLI bundle (dist/cli.mjs) and the
library bundle (dist/index.mjs) share one AsyncLocalStorage — otherwise the
flag would not cross from getMigrationPlan() to the migration's DB.raw().
- getMigrationPlan(migration, direction, { inert }) runs the body inertly.
- Planning-only callers pass inert: MigrateCommand's plan step, column-mapping
rebuild, and feature validation. Actual apply/rollback stay non-inert, so the
raw runs exactly once.
Verified end to end against a scratch database: migrate runs the raw once, and
rollback of a later batch no longer replays a still-applied migration's raw SQL.
Tests: DB.raw runs on normal planning but is suppressed under inert planning
while operations are still collected.
…ffects fix(migrations): don't replay DB.raw side effects during plan collection
Fix/cli commands load config
…urfacing fix(cli): rollback --step default + surface query error causes
3m1n3nc3
merged commit ccee8c3
into
fix/rollback-step-and-error-surfacing
8 checks passed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.