Bump the minimum version of PostgreSQL to 9.3

[yahonda]

Summary

This pull request bumps the minimum version of PostgreSQL to 9.3, dropping PostgreSQL 9.1 and 9.2 support.

https://www.postgresql.org/support/versioning/

• 9.1 EOLed on September 2016.
• 9.2 EOLed on September 2017.

9.3 is also not supported since Nov 8, 2018. https://www.postgresql.org/about/news/1905/
I think it may be a little bit early to drop PostgreSQL 9.3 yet.

Here are other changes included in this pull request:

• Deprecated supports_ranges? since no other databases support range data type

• Add supports_materialized_views? to abstract adapter
  Materialized views itself is supported by other databases, other connection adapters may support them

• Drop CI against PostgreSQL 9.2

[kamipo]

Is this query deterministic order even if the ORDER BY is removed?

[yahonda]

Thanks for the comment.

In my environment, it returns the primary key order as expected without order by but It does not look like deterministic since it does not Sort at the end of the query.

Was thinking about adding order by a.attnum but it just shows the column order of the table, not the index key order.

  create_table :countries_treaties, force: true, primary_key: [:country_id, :treaty_id] do |t|
    t.string :country_id, null: false
    t.string :treaty_id, null: false
  end

• Without ORDER BY i.idx

activerecord_unittest=#               SELECT a.attname
              FROM (
                     SELECT indrelid, indkey, generate_subscripts(indkey, 1) idx
                       FROM pg_index
                      WHERE indrelid = 'countries_treaties'::regclass
                        AND indisprimary
                   ) i
              JOIN pg_attribute a
                ON a.attrelid = i.indrelid
               AND a.attnum = i.indkey[i.idx];
  attname
------------
 country_id
 treaty_id
(2 rows)


activerecord_unittest=# explain SELECT a.attname
activerecord_unittest-#               FROM (
activerecord_unittest(#                      SELECT indrelid, indkey, generate_subscripts(indkey, 1) idx
activerecord_unittest(#                        FROM pg_index
activerecord_unittest(#                       WHERE indrelid = 'countries_treaties'::regclass
activerecord_unittest(#                         AND indisprimary
activerecord_unittest(#                    ) i
activerecord_unittest-#               JOIN pg_attribute a
activerecord_unittest-#                 ON a.attrelid = i.indrelid
activerecord_unittest-#                AND a.attnum = i.indkey[i.idx];
                                                         QUERY PLAN
-----------------------------------------------------------------------------------------------------------------------------
 Hash Join  (cost=329.65..357.92 rows=43 width=64)
   Hash Cond: ((pg_index.indrelid = a.attrelid) AND (pg_index.indkey[(generate_subscripts(pg_index.indkey, 1))] = a.attnum))
   ->  ProjectSet  (cost=0.27..13.30 rows=1000 width=34)
         ->  Index Scan using pg_index_indrelid_index on pg_index  (cost=0.27..8.29 rows=1 width=30)
               Index Cond: (indrelid = '207561'::oid)
               Filter: indisprimary
   ->  Hash  (cost=218.75..218.75 rows=7375 width=70)
         ->  Seq Scan on pg_attribute a  (cost=0.00..218.75 rows=7375 width=70)
(8 rows)

activerecord_unittest=#

• WITH ORDER BY i.idx

activerecord_unittest=# explain SELECT a.attname
              FROM (
                     SELECT indrelid, indkey, generate_subscripts(indkey, 1) idx
                       FROM pg_index
                      WHERE indrelid = 'countries_treaties'::regclass
                        AND indisprimary
                   ) i
              JOIN pg_attribute a
                ON a.attrelid = i.indrelid
               AND a.attnum = i.indkey[i.idx] ORDER BY i.idx;
                                                            QUERY PLAN
-----------------------------------------------------------------------------------------------------------------------------------
 Sort  (cost=359.09..359.20 rows=43 width=68)
   Sort Key: (generate_subscripts(pg_index.indkey, 1))
   ->  Hash Join  (cost=329.65..357.92 rows=43 width=68)
         Hash Cond: ((pg_index.indrelid = a.attrelid) AND (pg_index.indkey[(generate_subscripts(pg_index.indkey, 1))] = a.attnum))
         ->  ProjectSet  (cost=0.27..13.30 rows=1000 width=34)
               ->  Index Scan using pg_index_indrelid_index on pg_index  (cost=0.27..8.29 rows=1 width=30)
                     Index Cond: (indrelid = '207561'::oid)
                     Filter: indisprimary
         ->  Hash  (cost=218.75..218.75 rows=7375 width=70)
               ->  Seq Scan on pg_attribute a  (cost=0.00..218.75 rows=7375 width=70)
(10 rows)

activerecord_unittest=#

[yahonda]

Well, in the meantime I'm restoring this change to the current master one since it looks difficult (for me) to guarantee the order of index key, not table column order.

[y-yagi]

Can also remove branch condition in test_reconnection_after_actual_disconnection_with_verify?

rails/activerecord/test/cases/adapters/postgresql/connection_test.rb

Lines 163 to 176 in d6fcc41

	if @connection.send(:postgresql_version) >= 90200
	secondary_connection = ActiveRecord::Base.connection_pool.checkout
	secondary_connection.query("select pg_terminate_backend(#{original_connection_pid.first.first})")
	ActiveRecord::Base.connection_pool.checkin(secondary_connection)
	elsif ARTest.config["with_manual_interventions"]
	puts "Kill the connection now (e.g. by restarting the PostgreSQL " \
	'server with the "-m fast" option) and then press enter.'
	$stdin.gets
	else
	# We're not capable of terminating the backend ourselves, and
	# we're not allowed to seek assistance; bail out without
	# actually testing anything.
	return
	end

[yahonda]

Removed PostgreSQL version condition and with_manual_interventions from test_reconnection_after_actual_disconnection_with_verify.