How does did_you_mean gem work?

Slides for RubyConf Philippines 2015

Transcript

1. How does did_you_mean gem work? Yuki Nishijima
2. yuki24   @yuki24 Yuki Nishijima
3. Yuki Nishijima Yuki != snow ʢ༔وʣ
7. What is did_you_mean gem?
8. "Yuki".starts_with?("Y")
9. "Yuki".starts_with?("Y") # => NoMethodError: undefined method # `starts_with?’ for "Yuki":String
10. "Yuki".starts_with?("Y") # => NoMethodError: undefined method # `starts_with?’ for "Yuki":String "Yuki".start_with?("Y") # => true
11. require "did_you_mean" "Yuki".starts_with?("Y") # => NoMethodError: undefined method # `starts_with?’ for "Yuki":String # # Did you mean? #start_with? # "Yuki".start_with?("Y") # => true
12. Why did I create it?
13. A friend of mine was trying to introduce kaminari
14. But accidentally wrote #pagenate
15. THIS TYPO IS KILLI- NG ME! A R R G H
18. const char *help_unknown_cmd(const char *cmd) { ... /* This abuses cmdname->len for levenshtein distance */ for (i = 0, n = 0; i < main_cmds.cnt; i++) { ... main_cmds.names[i]->len = levenshtein(cmd, candidate, 0, 2, 1, 3) + 1; } ... }
19. const char *help_unknown_cmd(const char *cmd) { ... /* This abuses cmdname->len for levenshtein distance */ for (i = 0, n = 0; i < main_cmds.cnt; i++) { ... main_cmds.names[i]->len = levenshtein(cmd, candidate, 0, 2, 1, 3) + 1; } ... }
20. start_with   starts_with Levenshtein Distance
21. start_with   starts_with Levenshtein Distance
22. start_with   starts_with The distance is 1 Levenshtein Distance 1
23. first_name   full_name Levenshtein Distance
24. first_name   full_name Levenshtein Distance
25. first_name   full_name Levenshtein Distance 1    1    1    1
26. first_name   full_name The distance is 4 Levenshtein Distance 1    1    1    1
27. Make NoMethodError correctable
28. module Levenshtein def self.distance(str1, str2) str1, str2 = str1.to_s, str2.to_s # calculates edit distance... end end str = "Yuki Nishijima" threshold = 2 begin str.starts_with?("Yuki") rescue NoMethodError => error suggestions = str.methods.select do |name| Levenshtein.distance(name, error.name) < threshold end end puts suggestions # => [:start_with?]
29. module Levenshtein def self.distance(str1, str2) str1, str2 = str1.to_s, str2.to_s # calculates edit distance... end end str = "Yuki Nishijima" threshold = 2 begin str.starts_with?("Yuki") rescue NoMethodError => error suggestions = str.methods.select do |name| Levenshtein.distance(name, error.name) < threshold end end puts suggestions # => [:start_with?]
30. module Levenshtein def self.distance(str1, str2) str1, str2 = str1.to_s, str2.to_s # calculates edit distance... end end str = "Yuki Nishijima" threshold = 2 begin str.starts_with?("Yuki") rescue NoMethodError => error suggestions = str.methods.select do |name| Levenshtein.distance(name, error.name) < threshold end end puts suggestions # => [:start_with?]
31. module Levenshtein def self.distance(str1, str2) str1, str2 = str1.to_s, str2.to_s # calculates edit distance... end end str = "Yuki Nishijima" threshold = 2 begin str.starts_with?("Yuki") rescue NoMethodError => error suggestions = str.methods.select do |name| Levenshtein.distance(name, error.name) < threshold end end puts suggestions # => [:start_with?]
32. module Levenshtein def self.distance(str1, str2) str1, str2 = str1.to_s, str2.to_s # calculates edit distance... end end str = "Yuki Nishijima" threshold = 2 begin str.starts_with?("Yuki") rescue NoMethodError => error suggestions = str.methods.select do |name| Levenshtein.distance(name, error.name) < threshold end end puts suggestions # => [:start_with?] [    :<=>,    :==,    ...    :include?,    :start_with?,    :end_with?,    ...    :instance_eval,    :instance_exec,    :__send__,    :__id__   ]
33. module Levenshtein def self.distance(str1, str2) str1, str2 = str1.to_s, str2.to_s # calculates edit distance... end end str = "Yuki Nishijima" threshold = 2 begin str.starts_with?("Yuki") rescue NoMethodError => error suggestions = str.methods.select do |name| Levenshtein.distance(name, error.name) < threshold end end puts suggestions # => [:start_with?] [    :<=>,    :==,    ...    :include?,    :start_with?,    :end_with?,    ...    :instance_eval,    :instance_exec,    :__send__,    :__id__   ] 12   12   11   1   6   10   10   11   10
34. module Levenshtein def self.distance(str1, str2) str1, str2 = str1.to_s, str2.to_s # calculates edit distance... end end str = "Yuki Nishijima" threshold = 2 begin str.starts_with?("Yuki") rescue NoMethodError => error suggestions = str.methods.select do |name| Levenshtein.distance(name, error.name) < threshold end end puts suggestions # => [:start_with?] [    :<=>,    :==,    ...    :include?,    :start_with?,    :end_with?,    ...    :instance_eval,    :instance_exec,    :__send__,    :__id__   ] 12   12   11   1   6   10   10   11   10
35. module Levenshtein def self.distance(str1, str2) str1, str2 = str1.to_s, str2.to_s # calculates edit distance... end end str = "Yuki Nishijima" threshold = 2 begin str.starts_with?("Yuki") rescue NoMethodError => error suggestions = str.methods.select do |name| Levenshtein.distance(name, error.name) < threshold end end puts suggestions # => [:start_with?]
36. require "did_you_mean" "Yuki".starts_with?("Y") # => NoMethodError: undefined method # `starts_with?’ for "Yuki":String # # Did you mean? #start_with? #
37. error = (boom rescue $!) puts error.to_s # => undefined local variable or method `boom' for main:Object puts error.message # => undefined local variable or method `boom' for main:Object
38. Exception#to_s Returns exception’s message (or the name of the exception if no message is set). http://ruby-doc.org/core-2.2.1/Exception.html#method-i-to_s Exception#message Returns the result of invoking exception.to_s. Normally this returns the exception’s message or name. http://ruby-doc.org/core-2.2.1/Exception.html#method-i-message
39. static VALUE exc_message(VALUE exc) { return rb_funcall(exc, rb_intern("to_s"), 0, 0); } Exception#message(in  C)
40. static VALUE exc_message(VALUE exc) { return rb_funcall(exc, rb_intern("to_s"), 0, 0); } Exception#message(in  C) It just calls #to_s
41. Override Exception#to_s
42. class NoMethodError def to_s msg = super msg << "\n\n" msg << " Did you mean? #{suggestions.join(", ")}\n" rescue super end end
43. This needs to be implemented class NoMethodError def to_s msg = super msg << "\n\n" msg << " Did you mean? #{suggestions.join(", ")}\n" rescue super end end
44. module Levenshtein def self.distance(str1, str2) str1, str2 = str1.to_s, str2.to_s # calculates edit distance... end end str = "Yuki Nishijima" threshold = 2 begin str.starts_with?("Yuki") rescue NoMethodError => error suggestions = str.methods.select do |name| Levenshtein.distance(name, error.name) < threshold end end puts suggestions # => [:start_with?]
45. class NoMethodError THRESHOLD = 2 ... def suggestions receiver.methods.select do |name| Levenshtein.distance(name, self.name) < THRESHOLD end end end
46. class NoMethodError THRESHOLD = 2 ... def suggestions receiver.methods.select do |name| Levenshtein.distance(name, self.name) < THRESHOLD end end end NoMethodError objects don’t have access to the receiver
47. error = (boom rescue $!) puts error.to_s # => undefined local variable or method `boom' for main:Object puts error.message # => undefined local variable or method `boom' for main:Object
48. error = (boom rescue $!) puts error.to_s # => undefined local variable or method `boom' for main:Object puts error.message # => undefined local variable or method `boom' for main:Object Isn’t that the receiver object?
49. How does Ruby generate an error message (in C)?
50. Read Ruby’s source code and write a (initial) C extension
51. static VALUE make_no_method_exception(VALUE exc, const char *format, VALUE obj, int argc, const VALUE *argv) { int n = 0; VALUE mesg; VALUE args[3]; if (!format) { format = "undefined method `%s' for %s"; } mesg = rb_const_get(exc, rb_intern("message")); if (rb_method_basic_definition_p(CLASS_OF(mesg), '!')) { args[n++] = rb_name_err_mesg_new(mesg, rb_str_new2(format), obj, argv[0]); } else { args[n++] = rb_funcall(mesg, '!', 3, rb_str_new2(format), obj, argv[0]); } args[n++] = argv[0]; if (exc == rb_eNoMethodError) { args[n++] = rb_ary_new4(argc - 1, argv + 1); } return rb_class_new_instance(n, args, exc); }
52. static VALUE make_no_method_exception(VALUE exc, const char *format, VALUE obj, int argc, const VALUE *argv) { int n = 0; VALUE mesg; VALUE args[3]; if (!format) { format = "undefined method `%s' for %s"; } mesg = rb_const_get(exc, rb_intern("message")); if (rb_method_basic_definition_p(CLASS_OF(mesg), '!')) { args[n++] = rb_name_err_mesg_new(mesg, rb_str_new2(format), obj, argv[0]); } else { args[n++] = rb_funcall(mesg, '!', 3, rb_str_new2(format), obj, argv[0]); } args[n++] = argv[0]; if (exc == rb_eNoMethodError) { args[n++] = rb_ary_new4(argc - 1, argv + 1); } return rb_class_new_instance(n, args, exc); } NameError or NoMethodError
53. static VALUE make_no_method_exception(VALUE exc, const char *format, VALUE obj, int argc, const VALUE *argv) { int n = 0; VALUE mesg; VALUE args[3]; if (!format) { format = "undefined method `%s' for %s"; } mesg = rb_const_get(exc, rb_intern("message")); if (rb_method_basic_definition_p(CLASS_OF(mesg), '!')) { args[n++] = rb_name_err_mesg_new(mesg, rb_str_new2(format), obj, argv[0]); } else { args[n++] = rb_funcall(mesg, '!', 3, rb_str_new2(format), obj, argv[0]); } args[n++] = argv[0]; if (exc == rb_eNoMethodError) { args[n++] = rb_ary_new4(argc - 1, argv + 1); } return rb_class_new_instance(n, args, exc); } NameError or NoMethodError
54. static VALUE make_no_method_exception(VALUE exc, const char *format, VALUE obj, int argc, const VALUE *argv) { int n = 0; VALUE mesg; VALUE args[3]; if (!format) { format = "undefined method `%s' for %s"; } mesg = rb_const_get(exc, rb_intern("message")); if (rb_method_basic_definition_p(CLASS_OF(mesg), '!')) { args[n++] = rb_name_err_mesg_new(mesg, rb_str_new2(format), obj, argv[0]); } else { args[n++] = rb_funcall(mesg, '!', 3, rb_str_new2(format), obj, argv[0]); } args[n++] = argv[0]; if (exc == rb_eNoMethodError) { args[n++] = rb_ary_new4(argc - 1, argv + 1); } return rb_class_new_instance(n, args, exc); } NameError or NoMethodError This is the receiver object!
55. } return rb_class_new_instance(n, args, exc); } static VALUE make_no_method_exception(VALUE exc, const char *format, VALUE obj, int argc, const VALUE *argv) { int n = 0; VALUE mesg; VALUE args[3]; if (!format) { format = "undefined method `%s' for %s"; } mesg = rb_const_get(exc, rb_intern("message")); if (rb_method_basic_definition_p(CLASS_OF(mesg), '!')) { args[n++] = rb_name_err_mesg_new(mesg, rb_str_new2(format), obj, argv[0]); } else { args[n++] = rb_funcall(mesg, '!', 3, rb_str_new2(format), obj, argv[0]); } args[n++] = argv[0]; if (exc == rb_eNoMethodError) { args[n++] = rb_ary_new4(argc - 1, argv + 1); NameError or NoMethodError // array[0] = obj # in Ruby rb_ary_store(args[n - 1], 0, obj); class NoMethodError def receiver args.first end end
56. Initial C extension • Really hacky and unstable • Ruby headers required, head to maintain • Doesn’t compile on all Windows machines • Sometimes doesn’t compile even on OS X • Breaks the default behaviour of NoMethodError#args • Breaks other behaviours of Ruby 2.1.x on OS X, other gems (e.g. letter_opener, arel) have been affected
57. NoMethodError objects internally have access to the receiver object. So it shouldn’t be so hard to implement in C. Do you think it’s possible to implement NoMethodError#receiver? @_ko1 @n0kada
58. https://twitter.com/schneems/status/580438456771473408
59. #include <ruby.h> static const rb_data_type_t *type; static VALUE name_err_receiver(VALUE self) { VALUE *ptr, mesg = rb_attr_get(self, rb_intern("mesg")); TypedData_Get_Struct(mesg, VALUE, type, ptr); return ptr[1]; } void Init_method_receiver() { VALUE err_mesg = rb_funcall(rb_cNameErrorMesg, '!', 3, Qnil, Qnil, Qnil); type = RTYPEDDATA(err_mesg)->type; rb_define_method(rb_eNameError, "receiver", name_err_receiver, 0); }
60. Better C extension • Simple and stable • Ruby headers not required, easy to maintain • Compiles on most platforms • Doesn’t affect existing behaviours
61. Use rake-­‐compiler gem
62. # ext/did_you_mean/extconf.rb require 'mkmf' create_makefile 'did_you_mean/method_receiver' # Rakefile require 'rake/extensiontask' Rake::ExtensionTask.new('did_you_mean') do |ext| ext.name = "method_receiver" ext.lib_dir = "lib/did_you_mean" end # From terminal $ rake compile require ‘did_you_mean/method_receiver’ # => actives the C extension
63. require 'did_you_mean/method_receiver' str = "receiver" error = (str.doesnt_exist rescue $!) str.equal?(error.receiver) # => true
64. Put it all together
65. require 'did_you_mean/method_receiver' require 'did_you_mean/levenshtein' class NoMethodError THRESHOLD = 2 def to_s msg << super msg << "\n\n" msg << " Did you mean? ##{suggestions.join(', #')}\n" rescue super end def suggestions receiver.methods.select do |name| DidYouMean::Levenshtein.distance(name, self.name) < THRESHOLD end end end "Yuki".starts_with?("Y") # => NoMethodError: undefined method `starts_with?' for "Yuki":String # # Did you mean? #start_with?
66. THIS TYPO IS KILLI- NG ME! A R R G H
68. NoMethodError isn’t the only exception
69. pust name # => NameError: undefined local variable or method `pust' for..
70. NameError should provide suggestions too
71. user = User.new("Yuki Nishijima") begin suer rescue NameError => error suggestions = (local_variables + methods + private_methods).select do |name| Levenshtein.distance(name, error.name) < 2 end end puts suggestions # => [:user]
72. user = User.new("Yuki Nishijima") begin suer rescue NameError => error suggestions = (local_variables + methods + private_methods).select do |name| Levenshtein.distance(name, error.name) < 2 end end puts suggestions # => [:user] This is the only difference. We can re-use most of the code from NoMethodError
73. require 'did_you_mean/levenshtein' class NameError THRESHOLD = 2 def to_s msg << super msg << "\n\n" msg << " Did you mean? ##{suggestions.join(', #')}\n" rescue super end def suggestions names.select do |name| DidYouMean::Levenshtein.distance(name, self.name) < THRESHOLD end end def names local_variables + methods + private_methods end end
74. require 'did_you_mean/levenshtein' class NameError THRESHOLD = 2 def to_s msg << super msg << "\n\n" msg << " Did you mean? ##{suggestions.join(', #')}\n" rescue super end def suggestions names.select do |name| DidYouMean::Levenshtein.distance(name, self.name) < THRESHOLD end end def names local_variables + methods + private_methods end end Same as NoMethodError
75. require 'did_you_mean/levenshtein' class NameError THRESHOLD = 2 def to_s msg << super msg << "\n\n" msg << " Did you mean? ##{suggestions.join(', #')}\n" rescue super end def suggestions names.select do |name| DidYouMean::Levenshtein.distance(name, self.name) < THRESHOLD end end def names local_variables + methods + private_methods end end
76. require 'did_you_mean/levenshtein' class NameError THRESHOLD = 2 def to_s msg << super msg << "\n\n" msg << " Did you mean? ##{suggestions.join(', #')}\n" rescue super end def suggestions names.select do |name| DidYouMean::Levenshtein.distance(name, self.name) < THRESHOLD end end def names local_variables + methods + private_methods end end They don’t return what is expected because the scope is different.
77. user = User.new("Yuki Nishijima") begin suer rescue NameError => error suggestions = (local_variables + methods + private_methods).select do |name| Levenshtein.distance(name, error.name) < 2 end end puts suggestions # => [:user]
78. user = User.new("Yuki Nishijima") begin suer rescue NameError => error suggestions = (local_variables + methods + private_methods).select do |name| Levenshtein.distance(name, error.name) < 2 end end puts suggestions # => [:user]
79. def create_user user = User.new("Yuki Nishijima") # do something... suer end begin create_user rescue NameError => error suggestions = ??????.select do |name| Levenshtein.distance(name, error.name) < 2 end end puts suggestions # => [:user]
80. def create_user user = User.new("Yuki Nishijima") # do something... suer end begin create_user rescue NameError => error suggestions = ??????.select do |name| Levenshtein.distance(name, error.name) < 2 end end puts suggestions # => [:user] 1st scope
81. def create_user user = User.new("Yuki Nishijima") # do something... suer end begin create_user rescue NameError => error suggestions = ??????.select do |name| Levenshtein.distance(name, error.name) < 2 end end puts suggestions # => [:user] 1st scope 2nd scope
82. Use binding
83. A binding is a closure without a function— that is, it’s just the referencing environment. Think of bindings as a pointer to a YARV stack frame. binding — Pat Shaughnessy, Ruby Under a Microscope
84. def create_user user = User.new("Yuki Nishijima") # do something... suer end begin create_user rescue NameError => error suggestions = ??????.select do |name| Levenshtein.distance(name, error.name) < 2 end end puts suggestions # => [:user] 1st scope 2nd scope
85. def create_user user = User.new("Yuki Nishijima") # do something... suer end begin create_user rescue NameError => error suggestions = ??????.select do |name| Levenshtein.distance(name, error.name) < 2 end end puts suggestions # => [:user] 1st scope 2nd scope binding binding
86. interception gem
87. interception gem • allows to define hooks that gets called whenever an exception is raised • Provides access to both the exception object, and the binding from which it was raised • Works on MRI 1.9.3 - 2.2.1, JRuby, and Rubinius
88. require 'interception' def create_user user = User.new("Yuki Nishijima") # do something... suer end Interception.listen do |exception, binding| exception.instance_variable_set(:@frame_binding, binding) end begin create_user rescue NameError => error bin = error.instance_variable_get(:@frame_binding) names = bin.eval("local_variables + methods + private_methods") suggestions = names.select do |name| Levenshtein.distance(name, error.name) < 2 end end puts suggestions # => [:user]
89. require 'interception' def create_user user = User.new("Yuki Nishijima") # do something... suer end Interception.listen do |exception, binding| exception.instance_variable_set(:@frame_binding, binding) end begin create_user rescue NameError => error bin = error.instance_variable_get(:@frame_binding) names = bin.eval("local_variables + methods + private_methods") suggestions = names.select do |name| Levenshtein.distance(name, error.name) < 2 end end puts suggestions # => [:user]
90. require 'interception' def create_user user = User.new("Yuki Nishijima") # do something... suer end Interception.listen do |exception, binding| exception.instance_variable_set(:@frame_binding, binding) end begin create_user rescue NameError => error bin = error.instance_variable_get(:@frame_binding) names = bin.eval("local_variables + methods + private_methods") suggestions = names.select do |name| Levenshtein.distance(name, error.name) < 2 end end puts suggestions # => [:user] NameError,  self.binding
91. require 'interception' def create_user user = User.new("Yuki Nishijima") # do something... suer end Interception.listen do |exception, binding| exception.instance_variable_set(:@frame_binding, binding) end begin create_user rescue NameError => error bin = error.instance_variable_get(:@frame_binding) names = bin.eval("local_variables + methods + private_methods") suggestions = names.select do |name| Levenshtein.distance(name, error.name) < 2 end end puts suggestions # => [:user]
92. Put it all together (again)
93. require 'did_you_mean/levenshtein' require 'interception' Interception.listen do |exception, binding| exception.instance_variable_set(:@frame_binding, binding) end class NameError THRESHOLD = 2 def to_s msg << super msg << "\n\n" msg << " Did you mean? #{suggestions.join(', ')}\n" rescue super end def suggestions names.select do |name| DidYouMean::Levenshtein.distance(name, self.name) < THRESHOLD end end def names @frame_binding.eval("local_variables + methods + private_methods") end end
94. user = "Yuki Nishijima" suer # => NameError: undefined local variable or method `suer' for ... # # Did you mean? user
96. did_you_mean   does even more
97. class User < ActiveRecord::Base end USer # => NameError: uninitialized constant USer # # Did you mean? User #
98. class User < ActiveRecord::Base end USer # => NameError: uninitialized constant USer # # Did you mean? User # User.new(nmee: "wrong flrst name") # => ActiveRecord::UnknownAttributeError: unknown attribute: nmee # # Did you mean? name: string #
99. edge
100. did_you_mean works on JRuby and Rubinius edge
101. Save Time Be More Productive
102. Thanks!