Ruby Marshal.dump gives different results for what looks like the same thing

Question

I'm seeing slightly different results with Ruby's Marshal.dump depending on if I called .to_s on something or I typed in the characters. I'm really not clear on what's happening here:

»  Marshal.dump(1.to_s)
=> "\x04\bI\"\x061\x06:\x06EF"
»  Marshal.dump('1')
=> "\x04\bI\"\x061\x06:\x06ET"
»  1.to_s == '1'
=> true

So although it appears that 1.to_s == '1', they don't dump out to the same thing, but the only difference is in the very last byte. Any ideas why this is happening and how I can get both things to dump to the same byte sequence?

Answer 1

Marshal.load("\x04\bI\"\x061\x06:\x06EF").encoding
# => #<Encoding:US-ASCII> 
Marshal.load("\x04\bI\"\x061\x06:\x06ET").encoding
# => #<Encoding:UTF-8>

By default, 1.to_s.encoding is not the same as '1'.encoding. However, both strings are in 7-bit ASCII range, so they are comparable, and '1' == 1.to_s will be able to give you the result true, after some internal magic. But they are not the same thing.

Marshal.dump(1.to_s.force_encoding('utf-8'))
# => "\x04\bI\"\x061\x06:\x06ET"
Marshal.dump('1')
# => "\x04\bI\"\x061\x06:\x06ET"

(Assuming you run it on a newer Ruby, and don't mess with source encoding.)