durcno@1.0.0-alpha.12

Major Changes

• f680e01 - Thanks to @almahdi404 !

  change!: auto-convert schema identifiers from camelCase to snake_case

  Durcno now automatically converts all schema identifiers from camelCase to snake_case when generating SQL and migration files. You can write your entire schema in TypeScript-idiomatic camelCase — Durcno handles the PostgreSQL side.

  This affects every identifier passed to a schema builder:

  Builder	Identifiers converted
  table(schema, name, ...)	schema, table name, column names
  enumtype(schema, name, ...)	schema, enum name
  sequence(schema, name, ...)	schema, sequence name

  Enum values are not converted — they remain exactly as written.

  export const BlogPosts = table("public", "blogPosts", {
    id: pk(),
    authorId: bigint({ notNull }),
    publishedAt: timestamp({ notNull }),
  });

  Generates:

  CREATE TABLE "public"."blog_posts" (
    "id" bigserial PRIMARY KEY,
    "author_id" bigint NOT NULL,
    "published_at" timestamp NOT NULL
  );

  Breaking changes:

  • Any existing schema using camelCase identifiers that relied on the raw (unconverted) name being used in SQL or migrations will now produce snake_case identifiers instead. Re-generate your migrations after upgrading.
  • Constraint name helpers (check(), primaryKey(), unique()) no longer auto-prefix or auto-suffix names. The name you provide is used exactly as-is. Previously, primaryKey and unique would prepend the table name if it wasn't already present, and check would append _check if missing. You must now supply the full constraint name yourself. The recommended conventions are:
    • Primary key: pk_<table>
    • Unique: unique_<table>_<col>[_and_<col>]*
    • Check: check_<table>_<col>[_and_<col>]*[_<suffix>]?

• 7bf19a8 - Thanks to @almahdi404 !

  change(columns)!: introduce and helpers for array dimensions

  The dimension option on column configs now requires a Dimension object instead of a raw tuple literal. Two helper functions are exported from durcno to construct these objects:

  • array() — creates a variable-length (unbounded) dimension, equivalent to type[]
  • tuple(size) — creates a fixed-length dimension of size elements, equivalent to type[N]

  Both helpers return a Dimension instance that exposes .array() and .tuple(size) methods for chaining multi-dimensional shapes.

  // Before (no longer valid)
  tags: varchar({ length: 50, dimension: [null] as const });
  coordinates: integer({ dimension: [3] as const });
  matrix: integer({ dimension: [null, null] as const });
  
  // After
  tags: varchar({ length: 50, dimension: array() });
  coordinates: integer({ dimension: tuple(3) });
  matrix: integer({ dimension: array().array() });
  vectors: integer({ dimension: tuple(2).array() });

  Breaking change: the dimension property type changed from Readonly<[number | null, ...(number | null)[]]> to Dimension<readonly (number | null)[]>. Raw tuple literals are no longer accepted.

Minor Changes

• eda5522 - Thanks to @almahdi404 !

  feat(cli/squash): validate db state and sync tracking records

  The squash command now connects to the database to validate migration state and keep the durcno.migrations tracking table consistent after a squash.

  Mixed-state validation: if the squash range contains a mix of applied and unapplied migrations, the command exits with an error listing which are in each bucket.

  Tracking sync: when all migrations in the range are applied, the individual tracking records are deleted and replaced with a single entry pointing to the squashed migration's start name.

  --skip-db flag: pass --skip-db to bypass all database interaction entirely — useful in environments where the database is unavailable.

  # Normal usage — connects to DB for validation and tracking sync
  durcno squash 2024-01-01T00-00-00.000Z 2024-06-01T00-00-00.000Z
  
  # Skip DB — no validation, no tracking update
  durcno squash 2024-01-01T00-00-00.000Z 2024-06-01T00-00-00.000Z --skip-db

Patch Changes

• 0425ef9 - Thanks to @almahdi404 !

  perf: use columnsBySql index for faster column lookups

  Replace repeated snakeToCamel string conversions with a pre-built columnsBySql index keyed by SQL column names. This optimization eliminates runtime string transformations during result row conversion in query builders (Select, Insert, InsertReturning, FirstQuery, Update).

  The Table configuration now includes a columnsBySql object that maps SQL column names directly to column definitions, enabling O(1) lookups instead of string conversion followed by object lookup. This reduces CPU overhead on every row returned from the database.