Text
Page: 1
Red Arrow
Ruby and Apache Arrow
Sutou Kouhei
ClearCode Inc.
RubyKaigi Takeout 2021
2021-09-11
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 2
Sutou Kouhei
A president Rubyist
The president of ClearCode Inc.
クリアコードの社長
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 3
Sutou Kouhei
An Apache Arrow contributor
✓ A member of PMC of Apache Arrow
PMC: Project Management Committee
Apache Arrowのプロジェクト管理委員会メンバー
✓ #2 commits （コミット数2位）
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 4
Sutou Kouhei
The pioneer in Ruby and Arrow
✓ The author of Red Arrow
Red Arrowの作者
✓ Red Arrow:
✓ The official Apache Arrow library for Ruby
公式のRuby用のApache Arrowライブラリー
✓ GObject Introspection based bindings
GObject Introspectionベースのバインディング
✓ Apache Arrow GLib is developed for Red Arrow
Red ArrowのためにApache Arrow GLibも開発
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 5
GObject Introspection?
A way to implement bindings
バインディングの実装方法の1つ
Ruby bindings 2016
How to create bindings 2016
Kouhei Sutou
ClearCode Inc.
RubyKaigi 2016
2016-09-09
Ruby bindings 2016 - How to create bindings 2016
Powered by Rabbit 2.2.0
https://rubykaigi.org/2016/presentations/ktou.html
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 6
Why do I work on Red Arrow?
なぜRed Arrowの開発をしているか
✓ To use Ruby for data processing!
データ処理でRubyを使いたい！
✓ At least a part of data processing
データ処理の全部と言わず一部だけでも
✓ Results of my 5 years of work:
私のここ5年の仕事の成果
✓ We can use Ruby for some data processing!
いくつかのデータ処理でRubyを使える！
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 7
Goal of this talk
このトークのゴール
✓ You want to use Ruby
for some data processing
いくつかのデータ処理でRubyを使いたくなる
✓ You join Red Data Tools project
Red Data Toolsプロジェクトに参加する
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 8
Red Data Tools project?
Red Data Tools is a project that
provides data processing tools for
Ruby
Red Data ToolsはRuby用のデータ処理ツールを提供するプロジェクト
https://red-data-tools.github.io/
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 9
Data processing?
... how?
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 10
0. Why do you want?
0. データ処理の目的を明らかにする
✓ What problem do you want to resolve?
どんな問題を解決したい？
✓ What data is needed for it?
そのためにはどんなデータが必要？
✓ ...
No Red Arrow support in this area
このあたりにはRed Arrowを使えない
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 11
1. Collect data
1. データ収集
✓ Where are data?
データはどこにある？
✓ Where are collected data stored?
集めたデータはどこに保存する？
✓ ...
Some Red Arrow supports in this area
このあたりでは少しRed Arrowを使えない
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 12
Common dataset
よく使われるデータセット
require "datasets"
Datasets::Iris.new
Datasets::PostalCodeJapan.new
Datasets::Wikipedia.new
Red Datasets
https://github.com/red-data-tools/red-datasets
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 13
Output: Local file
出力先：ローカルファイル
require "datasets-arrow"
dataset = Datasets::PostalCodeJapan.new
dataset.to_arrow.save("codes.csv")
dataset.to_arrow.save("codes.arrow")
Red Datasets Arrow
https://github.com/red-data-tools/red-datasets-arrow
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 14
#save
✓ General serialize API for table data
テーブルデータ用の汎用シリアライズAPI
✓ Serialize as the specified format
指定したフォーマットにシリアライズ
✓ If you use Red Arrow object for in-memory
table data, you can serialize to many
formats! Cool!
メモリー上のテーブルデータをRed Arrowオブジェクトにするといろんな
フォーマットにシリアライズできる！かっこいい！
✓ Extensible!
拡張可能！
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 15
#save: Implementation
module Arrow
class Table
def save(output)
saver = TableSaver.new(self, output)
saver.save
end
end
end
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 16
#save: Implementation
class Arrow::TableSaver
def save
format = detect_format(@output)
__send__("save_as_#{format}")
end
def save_as_csv
end
end
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 17
#save: Extend by Red Parquet
module Parquet::ArrowTableSavable
def save_as_parquet
end
Arrow::TableSaver.include(self)
end
Red Parquet is a subproject of Red Arrow
Red ParquetはRed Arrowのサブプロジェクト
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 18
#save: Extended
require "datasets-arrow"
require "parquet"
dataset = Datasets::PostalCodeJapan.new
dataset.to_arrow.save("codes.parquet")
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 19
Output: Online storage: Fluentd
出力先：オンラインストレージ：Fluentd
✓ fluent-plugin-s3-arrow:
✓ Collect data by Fluentd
Fluentdでデータ収集
✓ Format data as Apache Parquet by Red Arrow
Red ArrowでApache Parquet形式にデータを変換
✓ Store data to Amazon S3 by fluent-plugin-s3
fluent-plugin-s3でAmazon S3にデータを保存
✓ By @kanga33 at Speee/Red Data Tools
Speee/Red Data Toolsの香川さんが開発
https://github.com/red-data-tools/fluent-plugin-s3-arrow/
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 20
Output: Online storage: Red Arrow
出力先：オンラインストレージ：Red Arrow
require "datasets-arrow"
require "arrow-dataset"
dataset = Datasets::PostalCodeJapan.new
url = URL("s3://mybucket/codes.parquet")
dataset.to_arrow.save(url)
Implementing...
実装中。。。
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 21
#save: Implementing...
class Arrow::TableSaver
def save
if @output.is_a?(URI)
__send__("save_to_uri")
else
__send__("save_to_file")
end
end
end
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 22
Collect data w/ Red Arrow: Wrap up
Red Arrowでデータ収集：まとめ
✓ Usable as serializer for common formats
よくあるフォーマットにシリアライズするツールとして使える
✓ Usable as writer to common locations
in the near future...
近いうちによくある出力先に書き出すツールとして使える
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 23
2. Read data
2. データ読み込み
✓ What format is used?
どんなフォーマットで保存されている？
✓ Where are collected data?
収集したデータはどこ？
✓ How large is collected data?
データはどれかで大きい？
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 24
Format
フォーマット
require
table =
table =
table =
table =
Red Arrow - Ruby and Apache Arrow
"arrow"
Arrow::Table.load("data.csv")
Arrow::Table.load("data.json")
Arrow::Table.load("data.arrow")
Arrow::Table.load("data.orc")
Powered by Rabbit 3.0.1
Page: 25
.load
✓ General deserialize API for table data
テーブルデータ用の汎用デシリアライズAPI
✓ Deserialize common formats
よく使われているフォーマットからデシリアライズ
✓ Extensible!
拡張可能！
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 26
.load: Implementation
module Arrow
def Table.load(input)
loader = TableLoader.new(self, input)
loader.load
end
end
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 27
.load: Implementation
class Arrow::TableLoader
def load
format = detect_format(@output)
__send__("load_as_#{format}")
end
def load_as_csv
end
end
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 28
.load: Extend by Red Parquet
module Parquet::ArrowTableLoadable
def load_as_parquet
end
Arrow::TableLoader.include(self)
end
Red Parquet is a subproject of Red Arrow
Red ParquetはRed Arrowのサブプロジェクト
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 29
.load: Extended
require "parquet"
table = Arrow::Table.load("data.parquet")
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 30
.load: More extensible
class Arrow::TableLoader
def load
if @output.is_a?(URI)
__send__("load_from_uri")
else
__send__("load_from_file")
end
end
end
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 31
.load: Extend by Red Arrow Dataset
module ArrowDataset::ArrowTableLoadable
def load_from_uri
end
Arrow::TableLoader.include(self)
end
Red Arrow Dataset is a subproject of Red Arrow
Red Arrow DatasetはRed Arrowのサブプロジェクト
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 32
Location: Online storage
場所：オンラインストレージ
require "arrow-dataset"
url = URI("s3://bucket/path...")
table = Arrow::Table.load(url)
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 33
Location: RDBMS
場所：RDBMS
require "arrow-activerecord"
User.all.to_arrow
Red Arrow Active Record
https://github.com/red-data-tools/red-arrow-activerecord
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 34
Location: Network
場所：ネットワーク
require "arrow-flight"
client = ArrowFlight::Client.new(url)
info = client.list_flights[0]
reader = client.do_get(info.endpoints[0].ticket)
table = reader.read_all
Introducing Apache Arrow Flight: A Framework for Fast Data Transport
https://arrow.apache.org/blog/2019/10/13/introducing-arrow-flight/
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 35
Large data
大規模データ
✓ Apache Arrow format
✓ Designed for large data
大規模データ用に設計されている
✓ For large data
大規模データ用に必要なもの
✓ Fast load
高速にロードできること
✓ ...
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 36
Fast load: Benchmark
高速ロード：ベンチマーク
require "datasets-arrow"
dataset = Datasets::PostalCodeJapan.new
table = dataset.to_arrow # 124271 records
n = 5
n.times do |i|
table.save("codes.#{i}.csv")
table.save("codes.#{i}.arrow")
CSV.read("codes.#{i}.csv")
Arrow::Table.load("codes.#{i}.csv")
Arrow::Table.load("codes.#{i}.arrow")
table = table.concatenate([table])
end
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 37
Fast load: Benchmark: All
高速ロード：ベンチマーク：すべて
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 38
Fast load: Benchmark: Red Arrow
高速ロード：ベンチマーク：Red Arrow
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 39
How to implement fast load
高速ロードの実装方法
Apache Arrowフォーマットは
なぜ速いのか
須藤功平
株式会社クリアコード
db tech showcase ONLINE 2020
2020-12-08
Apache Arrowフォーマットはなぜ速いのか
Powered by Rabbit 3.0.1
https://slide.rabbit-shocker.org/authors/kou/db-tech-showcase-online-2020/
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 40
Read data with Red Arrow: Wrap up
Red Arrowでデータ読み込み：まとめ
✓ Easy to read common formats
よくあるフォーマットのデータを簡単に読める
✓ Easy to read from common locations
よくある場所にあるデータを簡単に読める
✓ Large data ready
大規模データも扱える
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 41
3. Explore data
3. データ探索
✓ Preprocess data （データを前処理）
✓ Filter out needless data （不要なデータを除去）
✓ ...
✓ Summarize data and visualize them
（データを要約して可視化）
✓ ...
Red Arrow can be used for some operations
いくつかの操作でRed Arrowを使える
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 42
Filter: Red Arrow
絞り込み：Red Arrow
table = Datasets::PostalCodeJapan.new.to_arrow
table.n_rows # 124271
filtered_table = table.slice do |slicer|
slicer.prefecture == "東京都" # Tokyo
end
filtered_table.n_rows # 3887
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 43
Filter: Performance
絞り込み：性能
dataset = Datasets::PostalCodeJapan.new
arrow_dataset = dataset.to_arrow
dataset.find_all do |row|
row.prefecture == "東京都" # Tokyo
end # 1.256s
arrow_dataset.slice do |slicer|
slicer.prefecture == "東京都" # Tokyo
end # 0.001s
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 44
Filter: Performance
絞り込み：性能
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 45
Apache Arrow data: Interchangeable
Apache Arrow data：交換可能
✓ With low cost thanks to fast load
高速ロードできるので低コスト
✓ Apache Arrow data ready systems are
increasing
Apache Arrowデータを扱えるシステムは増加中
✓ e.g. DuckDB: in-process SQL OLAP DBMS
(SQLite like DBMS for OLAP)
OLAP: OnLine Analytical Processing
例：DuckDB：同一プロセス内で動くデータ分析用SQL DB管理システム
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 46
Filter: DuckDB
絞り込み：DuckDB
require "arrow-duckdb"
codes = Datasets::PostalCodeJapan.new.to_arrow
db = DuckDB::Database.open
c = db.connect
c.register("codes", codes) do # Use codes without copy
c.query("SELECT * FROM codes WHERE prefecture = ?",
"東京都", # Tokyo
output: :arrow) # Output as Apache Arrow data
.to_table.n_rows # 3887
end
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 47
Summarize: Group + aggregation
要約：グループ化して集計
iris = Datasets::Iris.new.to_arrow
iris.group(:label).count(:sepal_length)
#
count(sepal_length)
label
# 0
50
Iris-setosa
# 1
50
Iris-versicolor
# 2
50
Iris-virginica
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 48
Visualize: Charty
可視化：Charty
require "charty"
Charty.backends.use("pyplot")
Charty.scatter_plot(data: iris,
x: :sepal_length,
y: :sepal_width,
color: :label)
.save("iris.png")
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 49
Visualize: Charty: Result
可視化：Charty：結果
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 50
4. Use insight
4. 知見を活用
✓ Write report
（レポートにまとめたり）
✓ Build a model
（モデルを作ったり）
✓ ...
No Red Arrow support in this area for now
Can be used for passing data to other tools like DuckDB and Charty
今のところこのあたりにはRed Arrowを使えない
DuckDBやChartyにデータを渡すように他のツールにデータを渡すためには使える
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 51
Data processing and Red Arrow
Red Arrowでデータ処理
✓ Red Arrow helps us in some areas
いくつかの領域ではRed Arrowを使える
✓ Collect, read and explore data
データを収集して読み込んで探索するとか
✓ Some tools can integrate with Red Arrow
いくつかのツールはRed Arrowと連携できる
✓ Fluentd, DuckDB, Charty, ...
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 52
Red Arrow and Ruby 3.0
✓ MemoryView support
✓ Ractor support
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 53
MemoryView
MemoryView provides the features to
share multidimensional homogeneous
arrays of fixed-size element on
memory among extension libraries.
MemoryViewは多次元数値配列（数値はすべて同じ型）を共有する機能を提供します。
https://docs.ruby-lang.org/en/master/doc/memory_view_md.html
https://tech.speee.jp/entry/2020/12/24/093131 (Japanese)
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 54
Numeric arrays in Red Arrow
Red Arrow内の数値配列
✓ Arrow::NumericArray family
✓ 1-dimensional numeric array
1次元数値配列
✓ Arrow::Tensor
✓ Multidimensional homogeneous numeric arrays
多次元数値配列
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 55
MemoryView: Red Arrow
✓ Arrow::NumericArray family
✓ Export as MemoryView: Support
MemoryViewとしてエクスポート：対応済み
✓ Import from MemoryView: Not yet
MemoryViewをインポート：未対応
✓ Arrow::Tensor
✓ Export/Import: Not yet
エクスポート・インポート：未対応
Join Red Data Tools to work on this!
対応を進めたい人はRed Data Toolsに来てね！
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 56
MemoryView: C++
✓ Some problems are found by this work
Red Arrowの対応作業でいくつかの問題が見つかった
✓ Can't use private as member name
メンバー名にprivateを使えない
✓ Can't assign to const variable with cast
キャストしてもconst変数に代入できない
✓ Ruby 3.1 will fix them
Ruby 3.1では直っているはず
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 57
Ractor
Ractor is designed to provide a
parallel execution feature of Ruby
without thread-safety concerns.
Ractorはスレッドセーフかどうかを気にせずに並列実行するための機能です。
https://docs.ruby-lang.org/en/master/doc/ractor_md.html
https://techlife.cookpad.com/entry/2020/12/26/131858 (Japanese)
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 58
Red Arrow and concurrency
Red Arrowと並列性
✓ Red Arrow data are immutable
Red Arrowデータは変更不可
✓ Ractor can share frozen objects
Ractorはfrozenなオブジェクトを共有可能
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 59
Ractor: Red Arrow
require "datasets-arrow"
table = Datasets::PostalCodeJapan.new.to_arrow
Ractor.make_shareable(table)
Ractor.new(table) do |t|
t.slice do |slicer|
slicer.prefecture == "東京都" # Tokyo
end
end
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 60
Ractor: Red Arrow: Benchmark
n_ractors = 4
n_jobs_per_ractor = 1000
n_jobs = n_ractors * n_jobs_per_ractor
n_jobs.times do
table.slice {|s| s.prefecture == "東京都"}
end
n_ractors.times.collect do
Ractor.new(table, n_jobs_per_ractor) do |t, n|
n.times {t.slice {|s| s.prefecture == "東京都"}}
end
end.each(&:take)
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 61
Ractor: Red Arrow: Benchmark
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 62
Wrap up
まとめ
✓ Ruby can be used
in some data processing work
いくつかのデータ処理作業にRubyを使える
✓ Red Arrow helps you!
Red Arrowが有用なケースがあるはず！
✓ Ruby 3.0 has useful features for data
processing work
Ruby 3.0にはデータ処理作業に有用な機能があるよ
✓ Red Arrow starts supporting them
Red Arrowはそれらのサポートを進めている
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 63
Goal of this talk
このトークのゴール
✓ You want to use Ruby
for some data processing
いくつかのデータ処理でRubyを使いたくなる
✓ You join Red Data Tools project
あなたがRed Data Toolsプロジェクトに参加する
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 64
Feature work
今後の仕事
✓ Implement DataFusion bindings by
adding C API to DataFusion
DataFusionにC APIを追加してバインディングを実装
✓ DataFusion: Apache Arrow native query
execution framework written in Rust
https://github.com/apache/arrow-datafusion/
DataFusion：Rust実装のApache Arrowベースのクエリー実行フレームワーク
✓ Add Active Record like API to Red Arrow
Red ArrowにActive Record風のAPIを追加
✓ Improve MemoryView/Ractor support
MemoryView/Ractorサポートを進める
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 65
Red Data Tools
Join us!
https://red-data-tools.github.io/
https://gitter.im/red-data-tools/en
https://red-data-tools.github.io/ja/
https://gitter.im/red-data-tools/ja
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 66
OSS Gate on-boarding
OSS Gateオンボーディング
✓ Supports accepting newcomers by OSS
projects such as Ruby & Red Arrow
RubyやRed ArrowといったOSSプロジェクトが新人を受け入れることを支援
✓ Contact me! 興味がある人は私に教えて！
✓ OSS project members who want to accept newcomers
新人を受け入れたいOSSプロジェクトのメンバー
✓ Companies which want to support OSS Gate on-boarding
OSS Gateオンボーディングを支援したい会社
https://oss-gate.github.io/on-boarding/
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
Page: 67
ClearCode Inc.
✓ Recruitment: Developer to work on Red
Arrow related business
採用情報：Red Arrow関連のビジネスをする開発者
✓ https://www.clear-code.com/recruitment/
✓ Business: Apache Arrow/Red Arrow
related technical support/consulting:
仕事：Apache Arrow/Red Arrow関連の技術サポート・コンサルティング
✓ https://www.clear-code.com/contact/
Red Arrow - Ruby and Apache Arrow
Powered by Rabbit 3.0.1
