Make Scenic Integration Optional

[calmacleod]

Resolve Scenic Integration Causes Issues with Multiple Databases by adding optional configuration block with option to disable Scenic integration

[CLAassistant]

All committers have signed the CLA.

[murrayju]

Thanks for the submission! Just a bit of feedback, if you don't mind addressing it.

[calmacleod]

@murrayju Please note that I also bumped some tool versions to ensure that the project is up to date when running the tests locally.

For what it's worth, to get all the tests running locally, I had to make some changes to some surrounding code. I haven't committed it yet but it may be worth it to get it included.

Specifically the test helper teardown_tables doesn't properly handle dropping materialized views and the ContinuousAggregates model doesn't seem to scope jobs/chunks properly in it's has_many relationships.

diff --git a/lib/timescaledb/continuous_aggregates.rb b/lib/timescaledb/continuous_aggregates.rb
index 5c4e61f..85d4293 100644
--- a/lib/timescaledb/continuous_aggregates.rb
+++ b/lib/timescaledb/continuous_aggregates.rb
@@ -3,11 +3,15 @@ module Timescaledb
     self.table_name = "timescaledb_information.continuous_aggregates"
     self.primary_key = 'materialization_hypertable_name'
 
-    has_many :jobs, foreign_key: "hypertable_name",
-      class_name: "Timescaledb::Job"
+    has_many :jobs,
+    foreign_key: "hypertable_name",
+    primary_key: "view_name",
+    class_name: "Timescaledb::Job"
 
-    has_many :chunks, foreign_key: "hypertable_name",
-      class_name: "Timescaledb::Chunk"
+    has_many :chunks,
+    foreign_key: "hypertable_name",
+    primary_key: "materialization_hypertable_name",
+    class_name: "Timescaledb::Chunk"
 
     scope :resume, -> do
       {
diff --git a/spec/support/active_record/schema.rb b/spec/support/active_record/schema.rb
index 6924962..b0a2a91 100644
--- a/spec/support/active_record/schema.rb
+++ b/spec/support/active_record/schema.rb
@@ -36,6 +36,10 @@ def setup_tables
 end
 
 def teardown_tables
+  Timescaledb::ContinuousAggregates.all.each do |continuous_aggregate|
+    ActiveRecord::Base.connection.execute("DROP MATERIALIZED VIEW IF EXISTS #{continuous_aggregate.view_name} CASCADE")
+  end
+
   ActiveRecord::Base.connection.tables.each do |table|
     ActiveRecord::Base.connection.drop_table(table, force: :cascade)
   end

[murrayju]

@calmacleod lets get the tests to pass in this PR. Please add your fixes

[calmacleod]

@murrayju The PR has been updated with the following:

• Updated bundler + gem version to take care of deprecation warnings
• Updated the Schema Dumper spec to account for newer Active Record 8.1 method definitions
• Added new testing.md document with detailed explanations for how the test suite can be run.

[murrayju]

Thanks @calmacleod. It looks like the tests are still failing in the checks here. Can you take another look?

[calmacleod]

@murrayju I've made some changes that should hopefully fix those CI failures - The workflow needs maintainer approval before it can run however.

[calmacleod]

@murrayju There's some weird, potentially inconsistent, between local and CI testing.

This is the test failure I'm getting:

Failures:

  1) Timescaledb::MigrationHelpers.create_caggs when using refresh policies is expected not to be empty
     Failure/Error: expect(model.caggs.first.jobs).not_to be_empty
       expected `#<ActiveRecord::Associations::CollectionProxy []>.empty?` to be falsey, got true
     # ./spec/timescaledb/migration_helper_spec.rb:131:in `block (4 levels) in <top (required)>'

Initially I dealt with the same failure locally, however, I was able to get the test passing by modifying continuous_aggregates.rb to use view_name as a foreign key to the jobs table rather than materialization_hypertable_name.

This is because the cAgg refresh policy job gets created with the table name of the cAgg rather than the table name of the hyper table for which the cAgg is being created for.

The databases which we're testing against in CR are all 12m old images where as I'm using Timescale on PG18 to test locally, perhaps the structure between cAggs and Jobs have changed in newer versions?

The cAgg create by the test:

hypertable_schema	hypertable_name	view_schema	view_name	view_owner	materialized_only	compression_enabled	materialization_hypertable_schema	materialization_hypertable_name	view_definition	finalized
public	ticks	public	ohlc_1m	postgres	true	false	_timescaledb_internal	_materialized_hypertable_746	SELECT time_bucket('00:01:00'::interval, created_at) AS "time",symbol,first(price, created_at) AS open,max(price) AS high,min(price) AS low,last(price, created_at) AS close,sum(volume) AS volume FROM ticks GROUP BY (time_bucket('00:01:00'::interval, created_at)), symbol;	true

And the refresh policy job created for the cAgg:

job_id	application_name	schedule_interval	max_runtime	max_retries	retry_period	proc_schema	proc_name	owner	scheduled	fixed_schedule	config	next_start	initial_start	hypertable_schema	hypertable_name	check_schema	check_name
1264	Refresh Continuous Aggregate Policy [1264]	00:01:00	00:00:00	-1	00:01:00	_timescaledb_functions	policy_refresh_continuous_aggregate	postgres	true	false	{"end_offset":"PT1M","start_offset":"P1M","mat_hypertable_id":407}	2026-01-09 16:27:39.415483+00		public	ohlc_1m	_timescaledb_functions	policy_refresh_continuous_aggregate_check