supabase_schema_core 1.10.1

Internal package for Supabase schema fetching and type mapping. Used by supafreeze and suparepo. Not intended for direct use.

Changelog

1.10.1 - 2026-06-28

Fixed

• OpenAPI table introspection now reports defaults/nullability correctly. Follow-up to 1.10.0, which first made the OpenAPI table path reachable.
  • A NOT NULL column with a default is no longer marked nullable. PostgREST omits such columns from the required list (they are optional on insert), so !required.contains(col) alone was wrong; the presence of a default now also implies non-nullable.
  • String defaults are normalized to the single-quoted PG-literal form (JPY → 'JPY') so consumers that parse column_default recognize them. Expression defaults (gen_random_uuid(), now()) stay unquoted and are treated as non-literal.

1.10.0 - 2026-06-28

Fixed

• OpenAPI spec fetch no longer fails with HTTP 400. The request appended ?apikey=<key> to /rest/v1/, which PostgREST parsed as a column filter (PGRST100). The key is already sent as a header, so the redundant query parameter is dropped. This previously aborted all RPC return-type introspection.
• Catalog introspection works with an aggregating execute_sql. RPC/enum introspection queries wrap their projection in json_agg(sub); an execute_sql that itself aggregates rows with jsonb_agg(t) (rather than the README's EXECUTE ... INTO single-row form) double-wrapped the result as [{"json_agg": [...]}], throwing type 'Null' is not a subtype of type 'String'. Responses are now normalized for both conventions.

Added

• SchemaFetcher.rowsFromAggregatedResponse — normalizes a raw execute_sql response into the inner json_agg row list under either execute_sql convention.
• SchemaFetcher.applyResultModels — merges consumer result_models column overrides onto introspected RPC columns (overrides set type/nullability; unmentioned introspected columns are retained).

1.9.0 - 2026-06-26

Added

• RpcTableColumn.nullable — new field (default false) marking whether a RETURNS TABLE column may be null in the result row. Introspection cannot infer nullability from pg_proc, so it stays false for introspected columns and is intended to be overridden via consumers' YAML result_models definitions. toString() now includes the flag.

1.8.0 - 2026-05-19

Added

• SqlMigrationParser — pure-Dart parser for local SQL migration files. Extracts RPC return-type info from CREATE [OR REPLACE] FUNCTION ... RETURNS ... statements and CREATE TYPE ... AS (...) composite types, with no database round-trip. Supports:
  • Scalar returns (RETURNS BOOLEAN, TEXT, UUID, INTEGER, BIGINT, etc.)
  • RETURNS TABLE(col1 type1, ...)
  • RETURNS SETOF <type> (including composite types resolved to typed columns)
  • Dollar-quoted bodies (correctly ignores CREATE FUNCTION text inside function bodies / RAISE NOTICE strings)
  • Line (--) and block (/* */) comments
  • Trigger functions are skipped (not RPC-callable)
• SchemaFetcher.mergeSqlMigrationInfo — merges parser results into the function list, filling in only those still resolved to void

Changed

• The execute_sql failure warning now lists rpc.migrations_path as the recommended workaround alongside the existing options

1.7.3 - 2026-05-19

Added

• SchemaFetcher.warnings getter that exposes non-fatal diagnostics collected during the last fetchRpcFunctions() call (e.g. execute_sql introspection failures, suspiciously high ratio of RPC functions resolved to void)
• parseRpcFunctions now resolves $ref schemas against definitions / components.schemas and falls back to OpenAPI 3.0 responses.200.content.<mediaType>.schema when Swagger 2.0 responses.200.schema is absent

Fixed

• When execute_sql RPC is not available, fetchRpcFunctions() no longer silently degrades to Future<void> for every RPC. A clear warning is recorded explaining the failure and pointing to rpc.return_types as a workaround (fixes #2)

1.7.2 - 2026-04-08

Fixed

• Fix _resolveType order: coalesce/NOT/EXISTS checks now run before :: cast detection, preventing ::type inside nested subqueries from being misidentified as the expression type
• Harden parseJsonBuildObject and _tokenizeBuildObjectArgs to skip SQL string literals during parenthesis depth tracking

1.7.1 - 2026-04-08

Fixed

• Improve _resolveType to infer types from complex SQL expressions:
  • coalesce(..., false/true) → bool
  • coalesce(..., 0) → int4
  • NOT expr → bool
  • EXISTS(...) → bool
  • Boolean operators (AND, OR, IS NULL) → bool

1.7.0 - 2026-04-08

Added

• RpcTableColumn.nestedColumns and isArray fields for nested JSON structure
• SchemaFetcher.parseNestedJsonColumns() — parses json_agg(json_build_object(...)) patterns to extract nested column schemas with alias matching
• SchemaFetcher._fetchRpcNestedJsonColumns() — auto-detects nested structures for RETURNS TABLE json columns

1.6.0 - 2026-04-08

Added

• SchemaFetcher.parseRpcErrorCodes() — parses PL/pgSQL function source to extract snake_case error code literals from return query select false, 'code'::text and error := 'code' patterns
• RpcFunctionInfo.errorCodes field for detected error codes
• SchemaFetcher._fetchRpcErrorCodes() — queries pg_proc.prosrc for functions with error text columns in RETURNS TABLE

1.5.2 - 2026-04-08

Added

• TypeMapper.isJsonType() — returns true for json/jsonb PostgreSQL types, allowing callers to scope dynamic mapping to specific contexts (e.g. RPC result models) instead of globally

Fixed

• Revert json/jsonb mapping back to Map<String, dynamic> in TypeMapper — the global dynamic change in v1.5.1 was too broad; callers that need dynamic should use isJsonType() to decide

1.5.1 - 2026-04-08

Fixed

• Map json / jsonb PostgreSQL types to dynamic instead of Map<String, dynamic> — JSON values can be arrays (json_agg), objects, or scalars, so dynamic is the safe Dart type

1.5.0 - 2026-04-08

Added

• Auto-detect JSON column schemas from json_build_object() / jsonb_build_object() in PL/pgSQL function bodies via pg_proc.prosrc
• SchemaFetcher.parseJsonBuildObject() — parses function source to extract key names and infer types from variable declarations
• SchemaFetcher._fetchRpcJsonColumns() — queries pg_proc for RETURNS json/jsonb functions and auto-detects column schemas

1.4.2 - 2026-04-03

Added

• TypeMapper.isDateTimeType() — checks whether a PostgreSQL type (date, timestamp, timestamptz) maps to Dart DateTime
• TypeMapper.isDateOnlyType() — checks whether a PostgreSQL type is date-only (date) with no time component

1.4.1 - 2026-03-11

Fixed

• Fix mergeEnumTypes failing to match OpenAPI-derived enum names (tableName_columnName format) to pg_enum names, resulting in dynamic type
  • Add openApiEnums parameter to retain OpenAPI enum info before TypeMapper is cleared

1.4.0 - 2026-03-11

Added

• EnumInfo class for representing PostgreSQL enum types (name + values)
• SchemaFetcher.fetchEnums() — queries pg_enum + pg_type + pg_namespace catalog to fetch PostgreSQL enum definitions; falls back to OpenAPI-detected enums when execute_sql is not available
• SchemaFetcher.mergeEnumTypes() static method — replaces OpenAPI-derived enum type names (tableName_columnName) with actual PostgreSQL type names in table column definitions
• TypeMapper.useEnumTypes static flag — when true, mapType() returns PascalCase Dart enum type names instead of String for registered enum types
• TypeMapper.enumTypeName() — converts PostgreSQL enum type name to Dart enum type name (e.g. campaign_type → CampaignType)

1.3.0 - 2026-03-03

Added

• RpcTableColumn class for representing columns in RETURNS TABLE(...) definitions
• RpcFunctionInfo.tableColumns field — non-null when the function uses RETURNS TABLE(col1 type1, ...)
• SchemaFetcher._fetchRpcTableColumns() — queries pg_proc catalog (proargmodes, proargnames, proallargtypes) to extract TABLE column names and types
• SchemaFetcher.mergeTableColumns() static method to merge TABLE column info into RpcFunctionInfo list
• fetchRpcFunctions() now calls mergeTableColumns after mergeReturnTypes to populate tableColumns for RETURNS TABLE functions

1.2.2 - 2026-03-03

Fixed

• mergeReturnTypes() now correctly handles RETURNS TABLE(...) functions — previously treated as void because pg_proc reports them as record type, now detected as setof jsonb when proretset = true

1.2.1 - 2026-02-23

Fixed

• _fetchRpcReturnTypes() now wraps the pg_proc query with json_agg so that execute_sql (which uses EXECUTE ... INTO) returns all rows as a single JSON array instead of only the first row

1.2.0 - 2026-02-23

Added

• RpcFunctionInfo.copyWith() method for updating return type fields
• SchemaFetcher.mergeReturnTypes() static method to correct void return types using pg_proc catalog
• SchemaFetcher._fetchRpcReturnTypes() to query pg_proc for accurate return type information

Changed

• fetchRpcFunctions() now queries pg_proc catalog to correct scalar return types (e.g. bool, int4) that PostgREST OpenAPI spec reports as empty schema (void)
  • Only active when execute_sql RPC function is available
  • Falls back to OpenAPI-only behavior when execute_sql is not configured

1.1.0 - 2026-02-11

Added

• RpcParamInfo class for representing RPC function parameters
• RpcFunctionInfo class for representing RPC function metadata
• SchemaFetcher.fetchRpcFunctions() to fetch RPC function definitions from OpenAPI spec
• SchemaFetcher.parseRpcFunctions() to parse RPC functions from OpenAPI JSON

Changed

• Extracted _fetchOpenApiSpec() as a shared method for both table and RPC fetching

1.0.0 - 2025-12-11

Added

• Initial release
• Schema fetching from Supabase OpenAPI spec
• PostgreSQL to Dart type mapping
• Base configuration loader with environment variable support
• Foreign key detection from column naming conventions