Experiments with DuckDB in Elixir.

Note

The current API tries to map 1-to-1 to DuckDB C API. This is very verbose and low-level but it enables more idiomatic libraries to be built on top of this one without any performance or feature sacrifice.

For an example of a higher level library, please see DuxDB.Ecto.

Here's how it can be used on its own:

brew install duckdb llvm

duckdb --version
# v1.2.1 8e52ec4395

clang --version
# Homebrew clang version 19.1.7

export CC=$(which clang)
export DUXDB_LDFLAGS=-L/opt/homebrew/opt/duckdb/lib

Mix.install([{:duxdb, github: "ruslandoga/duxdb"}], force: true)

db = DuxDB.open(":memory:", %{"max_memory" => "1GB"})
conn = DuxDB.connect(db)

stmt =
  DuxDB.prepare(conn, """
  SELECT * FROM (VALUES
    (101, 'Hello, ' || $database || '!', today(), -1.0),
    (102, 'Insert a lot of rows per batch', today() - 1, 1.41421),
    (102, 'Sort your data based on your commonly-used queries', today() + 1, 2.718),
    (101, 'Granules are the smallest chunks of data read', $day, $pi)
  ) AS my_first_table(user_id, message, timestamp, metric)
  """)

DuxDB.bind_varchar(stmt, DuxDB.bind_parameter_index(stmt, "database"), "ClickHouse")
DuxDB.bind_double(stmt, DuxDB.bind_parameter_index(stmt, "pi"), 3.14159)
DuxDB.bind_date(stmt, DuxDB.bind_parameter_index(stmt, "day"), ~D[2025-03-14])

result = DuxDB.execute_prepared(stmt)

read_chunks =
  fn result ->
    Stream.repeatedly(fn -> DuxDB.fetch_chunk(result) end)
    |> Stream.take_while(&is_reference/1)
    |> Enum.map(fn chunk ->
      Map.new(0..(DuxDB.column_count(result) - 1), fn i ->
        {DuxDB.column_name(result, i), DuxDB.data_chunk_get_vector(chunk, i)}
      end)
    end)
  end

read_chunks.(result)
# [
#   %{
#     "message" => [
#       "Hello, ClickHouse!",
#       "Insert a lot of rows per batch",
#       "Sort your data based on your commonly-used queries",
#       "Granules are the smallest chunks of data read"
#     ],
#     "metric" => [-1.0, 1.41421, 2.718, 3.14159],
#     "timestamp" => [~D[2024-10-29], ~D[2024-10-28], ~D[2024-10-30], ~D[2025-03-14]],
#     "user_id" => [101, 102, 102, 101]
#   }
# ]

For more examples, please see doctests and tests.