Sha256: cf1a9e1f2e02d946d5bad7648d2048a67b38c7a50998967f953a63e72bc203ee
Contents?: true
Size: 1.93 KB
Versions: 11
Compression:
Stored size: 1.93 KB
Contents
require 'scenic/adapters/postgres'
require 'scenic/adapters/postgres/views'
module Timescaledb
module Scenic
class Views < ::Scenic::Adapters::Postgres::Views
# All of the views that this connection has defined, excluding any
# Timescale continuous aggregates. Those should be defined using
# +create_continuous_aggregate+ rather than +create_view+.
#
# @return [Array<Scenic::View>]
def all
ts_views = views_from_timescale.map { |v| to_scenic_view(v) }
pg_views = views_from_postgres.map { |v| to_scenic_view(v) }
ts_view_names = ts_views.map(&:name)
# Skip records with matching names (includes the schema name
# for records not in the public schema)
pg_views.reject { |v| v.name.in?(ts_view_names) }
end
private
def views_from_timescale
connection.execute(<<-SQL.squish)
SELECT
view_name as viewname,
view_definition AS definition,
'm' AS kind,
view_schema AS namespace
FROM timescaledb_information.continuous_aggregates
SQL
end
end
class Adapter < ::Scenic::Adapters::Postgres
# Timescale does some funky stuff under the hood with continuous
# aggregates. A continuous aggregate is made up of:
#
# 1. A hypertable to store the materialized data
# 2. An entry in the jobs table to refresh the data
# 3. A view definition that union's the hypertable and any recent data
# not included in the hypertable
#
# That doesn't dump well, even to structure.sql (we lose the job
# definition, since it's not part of the DDL).
#
# Our schema dumper implementation will handle dumping the continuous
# aggregate definitions, but we need to override Scenic's schema dumping
# to exclude those continuous aggregates.
def views
Views.new(connection).all
end
end
end
end
Version data entries
11 entries across 11 versions & 1 rubygems