Awesome
Pillar
Elixir client for ClickHouse, a fast open-source Online Analytical Processing (OLAP) database management system.
Features
- Direct Usage with connection structure
- Pool of workers
- Async insert
- Buffer for periodical bulk inserts
- Migrations
- DateTime Timezones
- Switching between HTTP adapters
Usage
Direct Usage with connection structure
conn = Pillar.Connection.new("http://user:password@localhost:8123/database")
# Params are passed in brackets {} in SQL query, and map strtucture does fill
# query by values.
sql = "SELECT count(*) FROM users WHERE lastname = {lastname}"
params = %{lastname: "Smith"}
{:ok, result} = Pillar.query(conn, sql, params)
result
#=> [%{"count(*)" => 347}]
Pool of workers
Recommended usage, because of limited connections and supervised workers.
defmodule ClickhouseMaster do
use Pillar,
connection_strings: [
"http://user:password@host-master-1:8123/database",
"http://user:password@host-master-2:8123/database"
],
name: __MODULE__,
pool_size: 15
end
ClickhouseMaster.start_link()
{:ok, result} = ClickhouseMaster.select(sql, %{param: value})
Async insert
connection = Pillar.Connection.new("http://user:password@host-master-1:8123/database")
Pillar.async_insert(connection, "INSERT INTO events (user_id, event) SELECT {user_id}, {event}", %{
user_id: user.id,
event: "password_changed"
}) # => :ok
Buffer for periodical bulk inserts
For this feature required Pool of workers.
defmodule BulkToLogs do
use Pillar.BulkInsertBuffer,
pool: ClickhouseMaster,
table_name: "logs",
# interval_between_inserts_in_seconds, by default -> 5
interval_between_inserts_in_seconds: 5,
# on_errors is optional
on_errors: &__MODULE__.dump_to_file/2
@doc """
dump to file function store failed inserts into file
"""
def dump_to_file(_result, records) do
File.write("bad_inserts/#{DateTime.utc_now()}", inspect(records))
end
@doc """
retry insert is dangerous (but it is possible and listed as proof of concept)
this function may be used in `on_errors` option
"""
def retry_insert(_result, records) do
__MODULE__.insert(records)
end
end
:ok = BulkToLogs.insert(%{value: "online", count: 133, datetime: DateTime.utc_now()})
:ok = BulkToLogs.insert(%{value: "online", count: 134, datetime: DateTime.utc_now()})
:ok = BulkToLogs.insert(%{value: "online", count: 132, datetime: DateTime.utc_now()})
....
# All this records will be inserted with 5 second interval.
on_errors parameter allows you to catch any error of bulk insert (for example: one of batch is bad or clickhouse was not available )
Migrations
Migrations can be generated with mix task mix pillar.gen.migration migration_name
.
Multi-statement migration Example for this UseCase
defmodule Pillar.Migrations.CreateMultipleTables do
def up do
# for MultiStatement migration result of this function should be List of Strings
(0..4) |> Enum.map(fn i ->
"CREATE TABLE IF NOT EXISTS shard_#{i} (field FixedString(10)) ENGINE = Memory"
end)
end
end
mix pillar.gen.migration events_table
But for launching them we have to write own task, like this:
defmodule Mix.Tasks.MigrateClickhouse do
use Mix.Task
def run(_args) do
connection_string = Application.get_env(:my_project, :clickhouse_url)
conn = Pillar.Connection.new(connection_string)
Pillar.Migrations.migrate(conn)
end
end
And launch this via command.
mix migrate_clickhouse
Timezones
In order to be able to use Timezones add timezones database to your project and configure your app:
config :elixir, :time_zone_database, Tzdata.TimeZoneDatabase
Details here https://hexdocs.pm/elixir/1.12/DateTime.html#module-time-zone-database
HTTP Adapters
If you have problems with default Pillar HTTP Adapter (Mint over Tesla), you can use alternative one, based on :httpc or define your own and pass it through config.
config :pillar, Pillar.HttpClient, http_adapter: Pillar.HttpClient.TeslaMintAdapter
Adapter should define one function post/3
and return 2 possible results (%Pillar.HttpClient.Response{}
, %Pillar.HttpClient.TransportError{}
)
Contribution
Feel free to make a pull request. All contributions are appreciated!