Skip to content

Elixir library client for work with ClickHouse

Notifications You must be signed in to change notification settings

ifoo/pillar

 
 

Repository files navigation

Pillar

github.com hex.pm hex.pm hex.pm hex.pm github.com

Elixir client for ClickHouse, a fast open-source Online Analytical Processing (OLAP) database management system.

Features

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.

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

Contribution

Feel free to make a pull request. All contributions are appreciated!

About

Elixir library client for work with ClickHouse

Resources

Stars

Watchers

Forks

Packages

No packages published

Languages

  • Elixir 100.0%