Version Packages

[github-actions]

This PR was opened by the Changesets release GitHub action. When you're ready to do a release, you can merge this and the packages will be published to npm automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to master, this PR will be updated.

Releases

sequelize-typescript-migration@1.0.0

Major Changes

• 484fc69 Thanks @kimjbstar! - Modernized for Sequelize 6, with experimental Sequelize 7 support.

  This release fixes several bugs that were silently changing what got generated. Run your
  first migration with preview: true and read the output — you will see differences that
  are not schema changes you made.

  Data that used to disappear

  • Default values now reach the migration. @Default(...) never made it into a generated
    file: the value was computed and then discarded when the attribute object was reassigned.
    Falsy defaults (@Default(false), @Default(0)) were dropped even earlier, by a truthy
    guard.
  • JSON, DOUBLE, FLOAT and REAL columns now appear. The type lookup missed them —
    DataTypes.JSON's class is named JSONTYPE — so they degraded to VIRTUAL and the caller
    skipped the column entirely, with nothing logged.
  • Unrecognised types now raise an error instead of quietly dropping the column. Returning
    a placeholder is what made the two bugs above possible.

  Migrations that could not run

  • BLOB no longer emits a bare identifier. Sequelize.BLOB((long)) threw
    ReferenceError the moment sequelize-cli loaded the file; DataType.BLOB without a length
    threw TypeError while generating.
  • ARRAY and RANGE no longer recurse infinitely. Both branches extracted the inner
    type and then recursed on themselves.
  • CHAR(n) keeps its length.

  Postgres

  • Identifiers are quoted per dialect. Unquoted FROM SequelizeMeta was folded to
    lowercase and failed with relation "sequelizemeta" does not exist, which made the tool
    unusable on Postgres (upstream #3, #4, #10).
  • The stored snapshot is parsed when a dialect returns it as a string. MySQL hands back a
    parsed object; SQLite has no JSON type and returns the raw text, which read as "nothing has
    ever been migrated" and regenerated the whole schema every run.

  Ordering

  • Foreign keys are ordered by a real topological sort. The previous pairwise-swap pass
    settled a chain of three but not four — measured, half of the input orderings of a
    four-table chain came out wrong, and 95 of 120 for five. This is the README's old
    "undo(down) action may not work" admission, now closed.
  • references.model is read correctly in all three places (two were misspelled).

  Breaking changes

  • sequelize and sequelize-typescript are now peer dependencies, and no database
    driver is bundled — install the one for your dialect yourself.
  • makeMigration no longer calls process.exit. It returns a discriminated union:
    { status: 'no-changes' | 'preview' | 'written' }.
  • Failures throw instead of being reported as success. A failure to record the snapshot
    used to be swallowed, leaving the next run diffing against a stale state.
  • preview is genuinely read-only — it no longer creates the bookkeeping tables.
  • Index identity changed. Indexes are identified by what they do rather than by the shape
    of the object that described them, so reordering decorator options no longer looks like a
    schema change. Existing indexes are re-created once, on the first run after upgrading.
  • Requires Node 18.18+.

  New

  • A real CLI. npx sequelize-typescript-migration --config ./db.js --out-dir ./migrations.
    The bin field previously pointed at a file with no entry point.
  • Experimental Sequelize 7 support. The package now requires no Sequelize at runtime and
    adapts to whichever version you pass it. v7 is still alpha upstream, hence experimental.
  • Published type declarations, and reads the snapshot table under both this package's name
    and the one the -lts / @techntools forks use, so switching from a fork keeps your history.

Patch Changes

• 2f82f62 Thanks @kimjbstar! - Fixed underscored: true and explicit field names being ignored.

  Migrations were generated against attribute names rather than column names, so a model with
  underscored: true produced a table of camelCase columns — one the model itself could not
  then read. The same applied to any column with an explicit field, and to index field
  lists: an index declared on actorName was created on actorName rather than actor_name.

  Sequelize already resolves this onto attribute.field, in both v6 and v7, so this needs no
  naming-convention option — unlike the useSnakeCase flag some forks added.

  If you use either feature, your previously generated migrations named the wrong columns.