Tweet
Tweet

Slide 1

Slide 1 text

@penelope_zone ✨Introducing✨ rubyfmt

Slide 2

Slide 2 text

@penelope_zone

Slide 3

Slide 3 text

@penelope_zone Updates

Slide 4

Slide 4 text

@penelope_zone This is gonna get real emotional

Slide 5

Slide 5 text

@penelope_zone Penelope She/Her Trans Woman

Slide 6

Slide 6 text

@penelope_zone Ruby Central Director

Slide 7

Slide 7 text

@penelope_zone No longer an RSpec maintainer

Slide 8

Slide 8 text

@penelope_zone Thanks to Jon, Myron, Andy, and David

Slide 9

Slide 9 text

@penelope_zone Rubyfmt is gonna be my open source focus for the next while

Slide 10

Slide 10 text

@penelope_zone What is Rubyfmt?

Slide 11

Slide 11 text

@penelope_zone @penelope_zone

Slide 12

Slide 12 text

@penelope_zone

Slide 13

Slide 13 text

@penelope_zone

Slide 14

Slide 14 text

@penelope_zone x

Slide 15

Slide 15 text

@penelope_zone I can’t unsee this

Slide 16

Slide 16 text

@penelope_zone Ruby autoformatter

Slide 17

Slide 17 text

@penelope_zone

Slide 18

Slide 18 text

@penelope_zone

Slide 19

Slide 19 text

@penelope_zone

Slide 20

Slide 20 text

@penelope_zone No formatting related configuration options

Slide 21

Slide 21 text

@penelope_zone “Unix tool”

Slide 22

Slide 22 text

@penelope_zone stdin/file name

Slide 23

Slide 23 text

@penelope_zone stdout/in place

Slide 24

Slide 24 text

@penelope_zone Not a gem!

Slide 25

Slide 25 text

@penelope_zone Probably > 6 months until something you can use daily

Slide 26

Slide 26 text

@penelope_zone Why are you doing this?

Slide 27

Slide 27 text

@penelope_zone Principle

Slide 28

Slide 28 text

@penelope_zone

Slide 29

Slide 29 text

@penelope_zone

Slide 30

Slide 30 text

@penelope_zone "TDD results in better code"

Slide 31

Slide 31 text

@penelope_zone This is another way of discussing tradeoffs

Slide 32

Slide 32 text

@penelope_zone My Principles

Slide 33

Slide 33 text

@penelope_zone Do one thing well Correctness beats simplicity Speed is needed at any cost

Slide 34

Slide 34 text

@penelope_zone Do one thing well

Slide 35

Slide 35 text

@penelope_zone

Slide 36

Slide 36 text

@penelope_zone I u Rubocop team

Slide 37

Slide 37 text

@penelope_zone Formatting Metrics Linting Naming Security Style Bundler Gemspec Rails

Slide 38

Slide 38 text

@penelope_zone

Slide 39

Slide 39 text

@penelope_zone Formatting Metrics Linting Naming Security Style Bundler Gemspec Rails

Slide 40

Slide 40 text

@penelope_zone Formatting Metrics Linting Naming Security Style Bundler Gemspec Rails

Slide 41

Slide 41 text

@penelope_zone

Slide 42

Slide 42 text

@penelope_zone $%&%'()*%%*+,)$%&%)%('- )'%$*($,%&%'%(&,-$()+$- +-%)$&+'')-%',*&)$&&'(% )$)-)&$*(+$%'-)&%+)*$'$ &*%)'-%$'-'$(*-(--$,)-- %&*,$+&*$*,))$)**,()'(* ,)++$-'&*'*'-*$)(-$($(+ '(%))$$),-$$+*$*(,%*(%- -&$(**,+&+'(&,,+)'&$)')

Slide 43

Slide 43 text

@penelope_zone

Slide 44

Slide 44 text

@penelope_zone Rubyformat is literally a different kind of tool

Slide 45

Slide 45 text

@penelope_zone

Slide 46

Slide 46 text

@penelope_zone Doing one thing extremely well

Slide 47

Slide 47 text

@penelope_zone By definition configuration options imply doing multiple things

Slide 48

Slide 48 text

@penelope_zone Support every possible combination

Slide 49

Slide 49 text

@penelope_zone Support every possible combination ❌

Slide 50

Slide 50 text

@penelope_zone Support one consistent style ✅

Slide 51

Slide 51 text

@penelope_zone bars = if foo baz else qux end

Slide 52

Slide 52 text

@penelope_zone bars = if foo baz else qux end

Slide 53

Slide 53 text

@penelope_zone bars = if foo baz else qux end

Slide 54

Slide 54 text

@penelope_zone bars = “one” \ “two”

Slide 55

Slide 55 text

@penelope_zone bars = “one” \ “two”

Slide 56

Slide 56 text

@penelope_zone bars = “one” \ “two”

Slide 57

Slide 57 text

@penelope_zone Rubyfmt always indents two relative to the parent construct

Slide 58

Slide 58 text

@penelope_zone bars = if foo baz else qux end

Slide 59

Slide 59 text

@penelope_zone bars = “one” \ “two”

Slide 60

Slide 60 text

@penelope_zone This is a tradeoff

Slide 61

Slide 61 text

@penelope_zone Correctness beats simplicity

Slide 62

Slide 62 text

@penelope_zone How do you parse Ruby?

Slide 63

Slide 63 text

@penelope_zone

Slide 64

Slide 64 text

@penelope_zone

Slide 65

Slide 65 text

@penelope_zone The parser output is really simple to deal with

Slide 66

Slide 66 text

@penelope_zone $ srb tc -p parse-tree -e 'a' Send { receiver = NULL method = args = [ ] }

Slide 67

Slide 67 text

@penelope_zone $ srb tc -p parse-tree -e ‘a(1)’ Send { receiver = NULL method = args = [ Integer { val = "1" } ] }

Slide 68

Slide 68 text

@penelope_zone

Slide 69

Slide 69 text

@penelope_zone

Slide 70

Slide 70 text

@penelope_zone For like 99% of Ruby programs this will never be a problem

Slide 71

Slide 71 text

@penelope_zone I refuse to build an auto formatter which could parse your program wrong

Slide 72

Slide 72 text

@penelope_zone

Slide 73

Slide 73 text

@penelope_zone Disadvantages

Slide 74

Slide 74 text

@penelope_zone Ripper requires a booted Ruby interpreter

Slide 75

Slide 75 text

@penelope_zone Running Ruby is slower than not running Ruby

Slide 76

Slide 76 text

@penelope_zone Ripper's output is painful

Slide 77

Slide 77 text

@penelope_zone $ rr 'a' [:program, [[:vcall, [:@ident, "a", [1, 0]]]]]

Slide 78

Slide 78 text

@penelope_zone $ rr 'a' [:program, [[:method_add_arg, [:fcall, [:@ident, "a", [1, 0]]], [:arg_paren, [:args_add_block, [[:@int, "1", [1, 2]]], false]]]]]

Slide 79

Slide 79 text

@penelope_zone Rubyfmt uses ripper to ensure compatibility

Slide 80

Slide 80 text

@penelope_zone Building against this is 90% of the dev time

Slide 81

Slide 81 text

@penelope_zone I would be done if I used whitequark

Slide 82

Slide 82 text

@penelope_zone I may end up separating Ripper from the Ruby interpreter as a general purpose tool

Slide 83

Slide 83 text

@penelope_zone It’s totally a good thing that Sorbet and Rubocop did not do this

Slide 84

Slide 84 text

@penelope_zone Speed is needed at any cost

Slide 85

Slide 85 text

@penelope_zone Gofmt executes on 25ms on even very large files

Slide 86

Slide 86 text

@penelope_zone

Slide 87

Slide 87 text

@penelope_zone

Slide 88

Slide 88 text

@penelope_zone How fast is fast enough?

Slide 89

Slide 89 text

@penelope_zone 16ms

Slide 90

Slide 90 text

@penelope_zone 16ms 16ms 16ms 16ms 16ms 16ms 16ms 100ms

Slide 91

Slide 91 text

@penelope_zone 100ms on 3k lines of Ruby code is the goal

Slide 92

Slide 92 text

@penelope_zone ruby —disable=gems -e ‘’

Slide 93

Slide 93 text

@penelope_zone 16ms 16ms Ruby boot 25ms

Slide 94

Slide 94 text

@penelope_zone ruby -e ‘'

Slide 95

Slide 95 text

@penelope_zone 16ms 16ms Ruby boot with gems 75ms 16ms 16ms 16ms

Slide 96

Slide 96 text

@penelope_zone We can’t run the parser in the remaining 25ms

Slide 97

Slide 97 text

@penelope_zone So we can't use gems

Slide 98

Slide 98 text

@penelope_zone Speed is needed at any cost

Slide 99

Slide 99 text

@penelope_zone 100 100 100 100 100 100 100 100

Slide 100

Slide 100 text

@penelope_zone bundle exec

Slide 101

Slide 101 text

@penelope_zone 100m 100m 100m 100m 100m 100m 100m 100m Bundler boot 100 100 100 100 100 100 100 100

Slide 102

Slide 102 text

@penelope_zone So we definitely can’t use bundler

Slide 103

Slide 103 text

@penelope_zone bundle exec rubocop <4 line file> 800ms 100m 100m 100m 100m 100m 100m 100m 100m Bundler boot Rubocop 100 100 100 100 100 100 100 100

Slide 104

Slide 104 text

@penelope_zone And we definitely can’t inherit from Rubocop

Slide 105

Slide 105 text

@penelope_zone So what does it look like today?

Slide 106

Slide 106 text

@penelope_zone

Slide 107

Slide 107 text

@penelope_zone

Slide 108

Slide 108 text

@penelope_zone

Slide 109

Slide 109 text

@penelope_zone

Slide 110

Slide 110 text

@penelope_zone So the complete Ruby version isn’t fast enough

Slide 111

Slide 111 text

@penelope_zone

Slide 112

Slide 112 text

@penelope_zone Parse.y isn’t really separable from the ruby interpreter

Slide 113

Slide 113 text

@penelope_zone ❤

Slide 114

Slide 114 text

@penelope_zone

Slide 115

Slide 115 text

@penelope_zone

Slide 116

Slide 116 text

@penelope_zone The initial test of the Rust version is that it is fast enough

Slide 117

Slide 117 text

@penelope_zone I am essentially building an entire set of tooling in Rust to work with Ruby source code

Slide 118

Slide 118 text

@penelope_zone Summary

Slide 119

Slide 119 text

@penelope_zone Do one thing well Correctness beats simplicity Speed is needed at any cost

Slide 120

Slide 120 text

@penelope_zone These principles imply a tonne of work

Slide 121

Slide 121 text

@penelope_zone Rubyfmt is the most technically complex project I have ever worked on

Slide 122

Slide 122 text

@penelope_zone Sweating the details

Slide 123

Slide 123 text

@penelope_zone Thanks to

Slide 124

Slide 124 text

@penelope_zone https://github.com/ penelopezone/rubyfmt

Slide 125

Slide 125 text

@penelope_zone That’s all @penelope_zone [email protected]