fix(datasource-active-record): map inet/cidr/macaddr(8) columns to String

[matthv]

What

Postgres network address column types (inet, cidr, macaddr, macaddr8) weren't in the TYPES map in column.rb, so they fell through to the unknown-type fallback. The fallback (String) was already correct, but it also logged a misleading INFO message on every schema generation, once per unmapped column: unknown type 'inet' for field named '...', 'String' type use by default.

Added the four types to TYPES, mirroring the existing Postgres-specific handling for citext/hstore/jsonb.

Why

Fixes #372.

How tested

• New spec cases in column_spec.rb (one per type) asserting both the String mapping and that the logger is no longer invoked for these types.
• Full forest_admin_datasource_active_record suite green locally (191 examples, 0 failures)
• rubocop clean on the changed files

🤖 Generated with Claude Code

Note

Map Postgres inet, cidr, macaddr, and macaddr8 columns to String

Adds inet, cidr, macaddr, and macaddr8 to the TYPES mapping in column.rb so get_column_type returns 'String' for these Postgres types instead of falling into the unknown-type logging path. Adds parameterized tests in column_spec.rb verifying the mapping and that no unknown-type log is emitted.

Changes since #377 opened

• Removed macaddr8 type mapping from the TYPES constant in ForestAdminDatasourceActiveRecord::Parser::Column and updated the corresponding spec [0995097]

^{Macroscope summarized c65ba85.}

[qltysh]

Coverage Impact

This PR will not change total coverage.

🚦 See full report on Qlty Cloud »

🛟 Help

• Diff Coverage: Coverage for added or modified lines of code (excludes deleted files). Learn more.

• Total Coverage: Coverage for the whole repository, calculated as the sum of all File Coverage. Learn more.

• File Coverage: Covered Lines divided by Covered Lines plus Missed Lines. (Excludes non-executable lines including blank lines and comments.)

  • Indirect Changes: Changes to File Coverage for files that were not modified in this PR. Learn more.

[christophebrun-forest]

macaddr8: 'String' is unreachable — #372 is only fixed for 3 of the 4 types

ActiveRecord never produces the :macaddr8 symbol. The type isn't registered anywhere, at either the class or the instance level:

$ grep -rn "macaddr8" activerecord-6.1.7.9/lib/ activerecord-8.1.3.1/lib/
(no match)

Verified empirically by resolving the real Postgres type map (PostgreSQLAdapter.initialize_type_map plus the three instance-level register_class_with_precision calls at postgresql_adapter.rb:758-765), on activerecord 8.1.3.1:

pg name	resolved class	.type
inet	OID::Inet	:inet ✅
cidr	OID::Cidr	:cidr ✅
macaddr	OID::Macaddr	:macaddr ✅
macaddr8	ActiveModel::Type::Value	nil ❌

The path is deterministic: get_oid_type → load_additional_types only maps an OID when its typname is already a registered key (TypeMapInitializer#run: mapped = nodes.extract! { |row| @store.key? row["typname"] }). macaddr8 isn't, so type_map.fetch falls into the block → Type.default_value → Value#type returns nil.

Concrete effect on this branch. For a real macaddr8 column, TYPES.key?(nil) is false, so we still hit the else in column.rb:31-37 and log:

unknown type '' for field named 'device_mac', 'String' type use by default

Confirmed with a throwaway spec run against this branch (it passes):

column = instance_double(ActiveRecord::ConnectionAdapters::SQLite3::Column,
                         name: 'device_mac', type: nil)   # what the PG adapter actually reports
expect(dummy_class.get_column_type(User, column)).to eq 'String'
expect(logger).to have_received(:log).with(
  'Info', "unknown type '' for field named 'device_mac', 'String' type use by default"
)
# => 1 example, 0 failures

So the misleading INFO log survives for macaddr8, and it's now harder to diagnose than before the PR, since the type name renders empty.

Worth noting the reason this slipped through: the new spec injects type: :macaddr8, a value no adapter ever emits, so it validates the map key that was chosen rather than what the adapter reports. All four cases pass with equal confidence even though one of them is wrong. (inet, cidr and macaddr are all correct — I checked them against the real type map, and the citext/hstore/jsonb analogy in the description holds.)

Two ways to close it

(a) Simplest — drop the macaddr8 entry and remove it from the spec's %i[...] array, then note on #372 that ActiveRecord doesn't map this type, so there's nothing the datasource can key off.

(b) More general — fall back to sql_type when the type doesn't resolve. This covers macaddr8 plus every other unregistered Postgres type (tsrange, domain types, extension types):

def get_column_type(model, column)
  return 'Enum' if model.respond_to?(:defined_enums) && model.defined_enums.key?(column.name)

  is_array = column.respond_to?(:array) && column.array == true
  type = TYPES[column.type] || TYPES[sql_type_key(column)]

  unless type
    type = TYPES[:string]
    ForestAdminAgent::Facades::Container.logger.log(
      'Info',
      "unknown type '#{column.type || column.sql_type}' for field named '#{column.name}', " \
      "'#{TYPES[:string]}' type use by default"
    )
  end

  is_array ? [type] : type
end

private

def sql_type_key(column)
  return nil unless column.respond_to?(:sql_type) && column.sql_type

  column.sql_type.delete_suffix('[]').to_sym
end

The delete_suffix('[]') matters: for macaddr8[], sql_type is "macaddr8[]". (inet[] is already fine either way — OID::Array does delegate :type, to: :subtype, so column.type == :inet and this PR correctly yields ['String'].)

Everything else in the PR looks good: no behaviour change beyond the log (both branches already returned 'String'), and the new specs are genuine guards — removing the four TYPES entries makes all four fail on expect(logger).not_to have_received(:log).

[christophebrun-forest]

IMO simple solution : delete macaddr8

[christophebrun-forest]

LGTM

[forest-bot]

🎉 This PR is included in version 1.39.4 🎉

The release is available on GitHub release

Your semantic-release bot 📦🚀