Poison.decode/1 invalid method signature.

[KamilLelonek]

According to the documentation:

decode(iodata, options \\ [])
decode(iodata, Keyword.t) ::
  {:ok, Poison.Parser.t} |
  {:error, :invalid} |
  {:error, {:invalid, String.t}}

The invalid cases are either {:error, :invalid} or {:error, {:invalid, String.t}}.

However, have a look at this example:

iex(1)> Poison.decode "{"
{:error, :invalid, 1}

so we are getting {:error, :invalid, integer()}.

And that one:

iex(18)> Poison.decode "}"
{:error, {:invalid, "}", 0}}

which is basically: {:error, {:invalid, String.t(), integer()}}

Aren't docs up to date then?

[keyan]

I am seeing the same issue on 3.0.0 and 3.1.0. Looks like it is due to this commit: 50f25b7

This is acceptable because of the major version bump, and surely would have been mentioned in the CHANGELOG if there was one.

However, this change was then removed: a4208a6#diff-574cf330de0ee486ba021c9eaa1f0e7bL42 so it seems that 4.0.0 will have the original 2.X behavior.

@KamilLelonek -- Can you verify the version you were using when conducting the above test?

Perhaps @devinus can explain the reasoning for the change and subsequent revert?

[KamilLelonek]

From what I see, I was using 3.1.0 version when this issue occurred.

[alakra]

Bump.

[keyan]

@alakra -- the conclusion above was that this issue appears to be limited to 3.0.0 and 3.0.1. Please include what version you are using and also try upgrading first.

[KamilLelonek]

this issue appears to be limited to 3.0.0 and 3.0.1

Not really. I'm using 3.1.0.

[keyan]

Yeah sorry, my mistake, the fix is only on master. And creating a new release would be a major version bump.

[Qqwy]

So how is this going?

This kind of problem is the prime candidate to create a bugfix-release for 3.1.X that has updated documentation to match the actual behaviour that Poison has.