Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Charty - Visualize Real-world Data with Ruby

7ef4ab70ee295c821f8b77fab3aa87cf?s=47 秒速284km
September 06, 2019

Charty - Visualize Real-world Data with Ruby

Charty - Visualize Real-world Data with Ruby

7ef4ab70ee295c821f8b77fab3aa87cf?s=128

秒速284km

September 06, 2019
Tweet

Transcript

  1. RUBYCONFTH Charty - Visualize Real-world Data with Ruby @284km (Kazuma

    Furuhashi) Speee Inc.
  2. RUBYCONFTH I came to Thailand for the first time. I've

    wanted to see the Chao Phraya River. I was able to achieve it yesterday. Thank you for holding the RubyConfTH
  3. about me (@284km) a programmer came from Japan, Tokyo I

    made Ruby2.6 standard CSV library 1.5x-3x faster a creator of a visualization library named Charty a member of Asakusa.rb
  4. If you ever come to Japan, come to Asakusa.rb

  5. Btw, what langage do you use? Ruby? Java? Python? JavaScript?

    English? Thai? Japanese? or many other languages?
  6. I often write Ruby, but speaking language is not Ruby.

  7. I recently learned some interesting things common to English, Thai

    and Japanese.
  8. I think these are similar pronunciations Shopping (ʃάpɪŋ) อป ง

    (cʰɔ́ ɔp pîŋ) ショッピング(shoppingu)
  9. There are similar online (ɔ́ nlàɪn) ออนไล (ɔɔn lai) オンライン

    (onrain)
  10. There are also similar donut (dóʊn`ʌt) โด ท (doo nát)

    ドーナツ (Dōnatsu)
  11. are these similar? ( :tea: ) 茶 (cha) ชา (cʰaa)

    tea (tíː)
  12. (cʰaa) is pronounced (tíː) (cʰaa) is (tíː) ชา (cʰaa) tea

    (tíː)
  13. ชา tea ...!! do you know?

  14. ชา tea "Charty" red-data-tools/charty So today I would like to

    talk about Charty.
  15. About Charty Charty is a open-source Ruby library for visualizing

    your data in a simple way. For example, Charty outputs these graphs. We can easily plot using Charty
  16. Charty using Ruby, Java, Python and JavaScript. These days, there

    is no de facto standard visualization tool for Ruby On the other hand, each above languages has its own good plotting library Thus, Charty is visualizing your data by standing on the shoulders of giants
  17. Characteristics of Charty Charty has 2 abstract layer Data Abstraction

    Layer (abstract data structure) Plotting Abstraction Layer (abstract backend plotting libraries) (I will explain about it later)
  18. This is basic usage of Charty (I will show a

    demo) /examples/iris_dataset.ipynb
  19. Next, focus on the code we need to write.

  20. This is output images How code we need to write?

  21. e.g. Pyplot backend require 'charty' charty = Charty::Plotter.new(:pyplot) scatter =

    charty.scatter do iris.group_by(:label).groups.each do |label, index| records = iris.row[*index] series records[:petal_length].to_a, records[:petal_width].to_a, label: label[0] end xlabel "Petal Length" ylabel "Petal Width" end scatter.render('pyplot.png')
  22. Gruff backend example If we want to use another backend,

    difference is only one line. require 'charty' charty = Charty::Plotter.new(:gruff) scatter = charty.scatter do iris.group_by(:label).groups.each do |label, index| records = iris.row[*index] series records[:petal_length].to_a, records[:petal_width].to_a, label: label[0] end xlabel "Petal Length" ylabel "Petal Width" end scatter.render('gruff.png')
  23. about Plotting Abstraction Layer

  24. about Plotting Abstraction Layer From the previous example, the difference

    is one line to change backend Here is one of the features of Charty We can easily switch backend libraries with almost the same code
  25. More about Plotting Abstraction Layer Currently supported backends is below

    pyplot gruff Rubyplot google-chart bokeh plotly plotly.js chart.js JFreeChart (Charty now works with JRuby )
  26. How to develop backend I feel that Pyplot has the

    largest number of graph types that can be output. When we want to add a graph to support, we often implement Pyplot first as a reference implementation. After that, we will implement other libraries.
  27. (other cases) For example, google-charts, bokeh, plotly These were implemented

    by a pull request that "I'd like to use Charty if this library is supported by the backend" If there is a real User and Real-world use case exists, it depends on the priority with other work, but consider support for a new backend
  28. about Data Abstraction Layer

  29. about Data Abstraction Layer Charty supports these data structures (I

    will show a demo) daru numo/narray nmatrix ActiveRecord Thus, Charty can respond to various data structures. That's because Charty::Table is abstracted.
  30. Feature summary of Charty Charty has two abstraction layers. Data

    Abstraction Layer Plotting Abstraction Layer. Thus we can use the data structures we need We can use output libraries we want to use. We can use them in any combination we need with almost no code rewrite.
  31. Introduction of various use cases of Charty Recently, we introduced

    Charty in our production environment of Web Application, which is our job. This Web Application is a common Rails Application. At that time, we were asking for Charty to output json, not image file. Here is an example using plotly.js (I will show a demo)
  32. Code # controller plotlyjs = Charty::Plotter.new(:plotlyjs) plotlyjs.table = DataModel.where(foo: bar)

    json_data = plotlyjs.to_json layout_data = plotlyjs.layout # view <div id="sample"></div> # javascript import * as Plotly from 'plotly.js-dist'; Plotly.newPlot("sample", json_data, layout_data);
  33. This combination is also possible. Because Charty has Data Abstraction

    Layer to support various data structures. https://github.com/284km/benchmark_driver- output-charty
  34. benchmark-driver outputs as below by default. $ gem install $

    benchmark-driver examples/parse.yaml Calculating ------------------------------------- csv 3.1.1 csv 3.0.1 unquoted 61.332 38.149 i/s - 100.000 times in 1.630461s 2.621311s quoted 30.558 17.023 i/s - 100.000 times in 3.272469s 5.874313s mixed 40.932 23.047 i/s - 100.000 times in 2.443082s 4.339030s include_col_sep 11.167 10.657 i/s - 100.000 times in 8.955275s 9.383878s include_row_sep 11.180 4.339 i/s - 100.000 times in 8.944608s 23.044523s encode_utf-8 39.129 31.525 i/s - 100.000 times in 2.555671s 3.172112s encode_sjis 49.982 31.289 i/s - 100.000 times in 2.000736s 3.196026s Comparison: unquoted csv 3.1.1: 61.3 i/s csv 3.0.1: 38.1 i/s - 1.61x slower quoted csv 3.1.1: 30.6 i/s csv 3.0.1: 17.0 i/s - 1.80x slower mixed csv 3.1.1: 40.9 i/s csv 3.0.1: 23.0 i/s - 1.78x slower include_col_sep csv 3.1.1: 11.2 i/s csv 3.0.1: 10.7 i/s - 1.05x slower include_row_sep csv 3.1.1: 11.2 i/s csv 3.0.1: 4.3 i/s - 2.58x slower encode_utf-8 csv 3.1.1: 39.1 i/s csv 3.0.1: 31.5 i/s - 1.24x slower encode_sjis csv 3.1.1: 50.0 i/s csv 3.0.1: 31.3 i/s - 1.60x slower
  35. benchmark_driver-output-charty outputs as below. $ benchmark-driver examples/parse.yaml -o charty

  36. summary of Data Abstraction Layer currently supports Array Hash daru

    numo-narray nmatrix ActiveRecord benchmark_driver (Charty Adapter) It can output: image, HTML, JSON format
  37. Now Charty with JRuby is working This impression became my

    motivation. So, I implemented JFreeChart backend. Thanks to Charles, development progressed.
  38. Charty with JRuby red-data-tools/charty-backends-jfreechart Please use JRubyist. This is not

    yet complete. If you have important use cases, I can write code. I'm grad to hear everyone's thoughts.
  39. Future plan (we aim for these) Improvement interface (Continued) Support

    red-arrow for Data Abstraction Layer because Apache Arrow is great. Release stable version Add supported dataset (red-datasets) (e.g. titanic)
  40. We code && Have fun! If you are interested in

    Charty, feel free to talk me. Thank you!