Parse/Encode logfmt is not idempotent when values have double quotes

[srstrickland]

In general, any encode/decode function pairs should be idempotent; that is, this should always be true:

x == decode(encode(x))

It appears to not be the case with encode_logfmt and parse_logfmt, as demonstrated here, where result.key has additional escape characters.

Taken to the extreme:

.obj.key = s'this has a " quote'
parse_logfmt!(encode_logfmt(parse_logfmt!(encode_logfmt(parse_logfmt!(encode_logfmt(parse_logfmt!(encode_logfmt(parse_logfmt!(encode_logfmt(.obj))))))))))

yields:

{ "key": "this has a \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\" quote" }

This isn't really causing me any headaches, and the fact that it hasn't been raised yet probably indicates the same for other folks, just thought it was worth bringing up, as I stumbled on it today.

[drmason13]

Thanks for raising this, I think you're absolutely right about the principle. I would like to take a stab at fixing this.

[drmason13]

I got a test reproducing the error this evening. The problem is the same in both directions:
encode(decode())
decode(encode())

decoding does not remove escape characters.
encoding escapes " and \ characters. EDIT: also \n characters.
Either decoding should parse and remove the escape characters, or encoding should not add escape characters.

I've been testing the encode/decode value functions, so I don't think it's just the logfmt functions.

[drmason13]

Adding some parsing escape removal to parsing sounds like the right fix to me.

It would be a breaking change for people who are relying on escape characters being perserved when parsing.