JSON.parse Alters Encoding of Passed String Unexpectedly

[aeroastro]

Description

I am currently developing an application where string encoding is important, and I've discovered that JSON.parse unexpectedly alters the encoding of the passed string.

Here's a simple code snippet to reproduce the issue:

require 'json'

str = '{ "hoge": "fuga" }'.force_encoding('ASCII-8BIT')

puts 'JSON'
puts "Before: #{str.encoding}"
puts JSON.parse(str)
puts "After:  #{str.encoding}"

puts

str = '{ "hoge": "fuga" }'.force_encoding('ASCII-8BIT')

puts 'JSON::Ext::Parser'
puts "Before: #{str.encoding}"
puts JSON::Ext::Parser.new(str).parse
puts "After:  #{str.encoding}"

This code yields the following output:

JSON
Before: ASCII-8BIT
{"hoge"=>"fuga"}
After:  UTF-8

JSON::Ext::Parser
Before: ASCII-8BIT
{"hoge"=>"fuga"}
After:  UTF-8

As shown above, the encoding of the string str changes from 'ASCII-8BIT' to 'UTF-8' after the JSON.parse operation.

My current environment is as follows:

$ ruby --version
ruby 3.2.2 (2023-03-30 revision e51014f9c0) [x86_64-darwin22]
$ ruby -rjson -e 'puts JSON::VERSION'
2.6.3

Expected Behavior

I would expect JSON.parse not to alter the original encoding of the passed string, unless it is necessary for the parsing operation.

Current Workaround

Currently, I'm using the dup method to create a copy of the original string before parsing it to prevent the original string's encoding from being altered:

JSON.parse(str.dup)