How to force NimbleParsec to use labels?

Questions / Help
1

I’m using NimbleParsec to parse a protocol and I would like to know if there is anyway to force a combinator to use labels.

For example, if I have this integer(min: 1) |> label("protocol"), the library will return an random error and not the protocol one in case I don’t send an integer. I would like to known if there is anyway to encapsulate a logic to guarantee that in case it fails, I can get a label to then pattern match later on the exact part it failed.

Thanks!

2

Using your example parser, I see:

iex> defmodule Parser do
...>  import NimbleParsec
...>
...>  defparsec :number,
...>    integer(min: 1) 
...>    |> label("protocol")
...> end
{:module, Parser, <<70, 79, 82, 49, 0, 0, 19, ...>>,.....

iex> Parser.number "3"
{:ok, [3], "", %{}, {1, 0}, 1

iex> Parser.number "a"
{:error, "expected protocol", "a", %{}, {1, 0}, 0}

Which seems pretty reasonable to me. I do find returning meaningful parsing errors a bit tricky at times - but typically not an issue in this simple kind of case.

What are you expecting, and what are you seeing?

1 Like
3

Sorry about the noise, I found the problem. I’m building a long chain of combinators, and the error was coming from the next one in the chain. I posted just a small snippet to make it easier to frame the question, but it masked out the problem.

This is the combinator now:

  device_id =
    integer(min: 1) |> unwrap_and_tag(:device_id) |> lookahead(string(",")) |> label("device id")

And it matches this message:

"*HQ,1234567890,V1,...

The *HQ, is being match by a previous combinator, and then the next part is the device ID, which is now correctly being parsed.

NimbleParsec is really good, but it can get quite tricky sometimes.

4

Glad you’re up and running. I typically try to make the grammar as explicit as possible and avoid lookahead unless really necessary. Perhaps thats just my preference. For example:

defmodule Parser do
  import NimbleParsec

  preamble =
    string("*HQ")
    |> label("preamble")
  
  device_id =
    integer(min: 1) 
    |> label("device id")

  version =
    string("V")
    |> integer(1)
    |> reduce({Enum, :join, []})
    |> label("version")

  separator =
    string(",")
    |> label("separator")
  
  defparsec :number,
    preamble
    |> ignore(separator)
    |> concat(device_id)
    |> ignore(separator)
    |> concat(version)
  
end

Examples

iex> Parser.number "*HQ,1234567890,V1"
{:ok, ["*HQ", 1234567890, "V1"], "", %{}, {1, 0}, 17}

iex> Parser.number "*HQ,A1234567890,V1"
{:error, "expected device id", "A1234567890,V1", %{}, {1, 0}, 4}

Which still gives an appropriate error message and makes it easy to expand the grammar imo.

2 Likes