undefined method `to_str' for 🆔Symbol (NoMethodError) in 2.8.2

[johrstrom]

Hi, I'm getting this error when I've updated to 2.8.2 from 2.8.1. Here's a very simple test case in irb that I've confirmed works in 2.8.1.

irb(main):004:0> uri = Addressable::Template.new("https://example/some/api/{?query*}")
=> #<Addressable::Template:0x104 PATTERN:https://example/some/api/{?query*}>
irb(main):005:0> uri.expand({'query' => {id: 'abc123' } })
/home/jeff/.gems/ruby-3.0.3/gems/addressable-2.8.2/lib/addressable/template.rb:900:in `normalize_value': undefined method `to_str' for :id:Symbol (NoMethodError)
Did you mean?  to_se-2.8.2.gem
               to_symx-5.0.1.gem
Successffrom /home/jeff/.gems/ruby-3.0.3/gems/addressable-2.8.2/lib/addressable/template.rb:908:in `block in normalize_value'
Successffrom /home/jeff/.gems/ruby-3.0.3/gems/addressable-2.8.2/lib/addressable/template.rb:907:in `each'
Parsing from /home/jeff/.gems/ruby-3.0.3/gems/addressable-2.8.2/lib/addressable/template.rb:907:in `inject'
Installifrom /home/jeff/.gems/ruby-3.0.3/gems/addressable-2.8.2/lib/addressable/template.rb:907:in `normalize_value'
Parsing from /home/jeff/.gems/ruby-3.0.3/gems/addressable-2.8.2/lib/addressable/template.rb:774:in `block in transform_capture'
Installifrom /home/jeff/.gems/ruby-3.0.3/gems/addressable-2.8.2/lib/addressable/template.rb:756:in `each'
Done insfrom /home/jeff/.gems/ruby-3.0.3/gems/addressable-2.8.2/lib/addressable/template.rb:756:in `inject'
2 gems ifrom /home/jeff/.gems/ruby-3.0.3/gems/addressable-2.8.2/lib/addressable/template.rb:756:in `transform_capture'
[dashboafrom /home/jeff/.gems/ruby-3.0.3/gems/addressable-2.8.2/lib/addressable/template.rb:595:in `block in expand'
irb(mainfrom /home/jeff/.gems/ruby-3.0.3/gems/addressable-2.8.2/lib/addressable/template.rb:594:in `gsub!'
=> true from /home/jeff/.gems/ruby-3.0.3/gems/addressable-2.8.2/lib/addressable/template.rb:594:in `expand'
irb(mainfrom (irb):5:in `<main>'mplate.new("#{server[:host]}{/segments*}/{?query*}")
        from /home/jeff/apps/ruby/3.0.3/lib/ruby/gems/3.0.0/gems/irb-1.3.5/exe/irb:11:in `<top (required)>'
        from /home/jeff/apps/ruby/3.0.3/bin/irb:23:in `load'
        from /home/jeff/apps/ruby/3.0.3/bin/irb:23:in `<main>'

Here's a simple snippet that you may be able to just copy paste easier.

require 'addressable'

uri = Addressable::Template.new("https://example/some/api/{?query*}")
uri.expand({'query' => { id: 'abc123' }})

[johrstrom]

Obviously it's due to passing a hash with symbol keys instead of string keys. Just using string keys works, in case other folks run into this and need a work around.

require 'addressable'

uri = Addressable::Template.new("https://example/some/api/{?query*}")
uri.expand({'query' => { 'id' => 'abc123' }})

[dentarg]

@jarthod mind taking a look?

I guess the template tests does not use symbols?

[johrstrom]

I was also able to reproduce with integers as values, so I think it affects keys and values, just FYI.

require 'addressable'

uri = Addressable::Template.new("https://example/some/api/{?query*}")
uri.expand({'query' => { 'id' => 1 }})

[dentarg]

Looks like to_s was always called before #492 (https://github.com/sporkmonger/addressable/pull/492/files#diff-fb36d3dc67e6565ffde17e666a98697f48e76dac38fabf1bb9e97cdf3b583d76)

addressable/lib/addressable/idna/pure.rb

Lines 115 to 117 in ab4bc46

	# Unicode normalization form KC.
	def self.unicode_normalize_kc(input)
	input = input.to_s unless input.is_a?(String)

addressable/lib/addressable/idna/native.rb

Lines 32 to 34 in ab4bc46

	def self.unicode_normalize_kc(value)
	IDN::Stringprep.nfkc_normalize(value.to_s)
	end

[johrstrom]

I can submit a patch if here you like, though given it's worked for the past 11 years, I'm not sure what unintended side affects it may cause. Someone with more intimate knowledge of the callers may know more?

addressable/lib/addressable/template.rb

Line 900 in bcaf766

value = value.respond_to?(:to_ary) ? value.to_ary : value.to_str

[jarthod]

I am having a look, I saw specs checking both symbols, string, and integers so I thought this was covered but apparently they were only testing this at the first level, and it turns out there's a special treatment at this level, so I'll add a couple more specs testing this at the second level and provide a patch. From what I can see putting back a to_s at the right place will do the trick yes.

Edit: Fix merged in PR #499, sorry for the regression 🙇

[crevete]

@dentarg just want to know when the 2.8.3 tag will be cut? thanks

[dentarg]

2.8.3 is out now