fix bind_* NIF typespecs

[ademenev]

No description provided.

[ademenev]

    now = :erlang.monotonic_time(:millisecond)
    {:ok, stmt} = Exqlite.Sqlite3.prepare(conn, "DELETE FROM queue WHERE expiration <= ?")
    :ok = Exqlite.Sqlite3.bind(stmt, [now])

dialyzer complains about the last line:

The function call will not succeed.
   
Exqlite.Sqlite3.bind(_stmt :: reference(), [integer(), ...])
   
will never return since the 2nd arguments differ  from the success typing arguments:
(reference(), nil | [])

and in exqlite source there is this:

  @spec bind_integer(statement, non_neg_integer, integer) :: :ok
  def bind_integer(stmt, index, integer) do
    case Sqlite3NIF.bind_integer(stmt, index, integer) do
      @sqlite_ok -> :ok
      rc -> raise Exqlite.Error, message: errmsg(stmt) || errstr(rc)
    end
  end

and this

  @spec bind_integer(statement, non_neg_integer, integer) :: :ok
  def bind_integer(_stmt, _index, _integer), do: :erlang.nif_error(:not_loaded)

The return value of Sqlite3NIF.bind_integer which is typed as :ok is matched against @sqlite_ok which is 0. That should always fail, so

The pattern can never match the type.
   
 Pattern:
 0
   
 Type:
 :ok

[warmwaffles]

😮‍💨 I could not figure out the dialyzer errors for the life of me. It didn't occur to me to actually check the return from the nif.

Can you go ahead and add the dialyzer to the CI as well now 😄

exqlite/.github/workflows/ci.yml

Line 38 in fd2492d

- run: mix credo --all

[warmwaffles]

I'm almost tempted to turn these integers into atoms in a separate PR. Much easier to decipher WTF.

[warmwaffles]

I'll chuck the ci check in later. Thanks a ton @ademenev.

[ademenev]

😮‍💨 I could not figure out the dialyzer errors for the life of me. It didn't occur to me to actually check the return from the nif.

Can you go ahead and add the dialyzer to the CI as well now 😄

exqlite/.github/workflows/ci.yml

Line 38 in fd2492d

- run: mix credo --all

It actually fails for me, so I guess adding it would just break CI

lib/mix/tasks/download_sqlite.ex:1:callback_info_missing
Callback info about the Mix.Task behaviour is not available.
________________________________________________________________________________
Please file a bug in https://github.com/jeremyjh/dialyxir/issues with this message.

Unknown error occurred: %Protocol.UndefinedError{protocol: String.Chars, value: {8, 34}, description: ""}


Legacy warning:
lib/mix/tasks/download_sqlite.ex:8:34: Unknown function 'Elixir.Exqlite.MixProject':sqlite_version/0
________________________________________________________________________________
lib/mix/tasks/test_sqlite_version.ex:1:callback_info_missing
Callback info about the Mix.Task behaviour is not available.
________________________________________________________________________________
lib/mix/tasks/test_sqlite_version.ex:17:unknown_function
Function Mix.raise/1 does not exist.
________________________________________________________________________________
Please file a bug in https://github.com/jeremyjh/dialyxir/issues with this message.

Unknown error occurred: %Protocol.UndefinedError{protocol: String.Chars, value: {9, 38}, description: ""}


Legacy warning:
lib/mix/tasks/test_sqlite_version.ex:9:38: Unknown function 'Elixir.Exqlite.MixProject':sqlite_version/0
________________________________________________________________________________
done (warnings were emitted)
Halting VM with exit status 2

Integestingly, ElixirLS does not report those for me, only mix dialyzer does

But al least for projects using the library dialyzer does is not complaining

[warmwaffles]

I stopped using ElixirLS for the time being. It is slow, eats memory, and causes so much pain. 😭