Recreate Sprache Csv Parser in Superpower

Hi @nblumhardt, I have been trying to recreate the Sprache Csv Parser in the test folder over in it’s repository with the Superpower library.

1. I have successfully made a simple version to parse my Csv data by row but I wanted to try and recreate the slightly more comprehensive Csv parser from Sprache’s tests but in Superpower.

class Program
{
    static void Main(string[] args)
    {
        var input1 = "a,b,c";
        var input2 = "d,e,f" + Environment.NewLine + "a,b,c";

        var values = CsvParser.Record.Parse(input1).ToList(); // Works for parsing individual records

        Console.WriteLine($"1:{values[0]}, 2:{values[1]}, 3:{values[2]}");

        values = CsvParser.Record.Parse(input2).ToList(); // Works just grabbing the first row

        Console.WriteLine($"1:{values[0]}, 2:{values[1]}, 3:{values[2]}");

        var test = CsvParser.Csv.Parse(input2); // Exception
    }
}

public class CsvParser
{
    private static readonly TextParser<char> CellSeparator = Character.EqualTo(',');

    private static readonly TextParser<char> QuotedCellDelimiter = Character.EqualTo('"');

    private static readonly TextParser<char> QuoteEscape = Character.EqualTo('"');

    private static TextParser<T> Escaped<T>(TextParser<T> following)
    {
        return from escape in QuoteEscape
                from f in following
                select f;
    }

    private static readonly TextParser<char> QuotedCellContent =
        Character.AnyChar.Between(QuotedCellDelimiter, Escaped(QuotedCellDelimiter));

    static readonly TextParser<char> LiteralCellContent =
        Character.ExceptIn(',', '\r', '\n');

    static readonly TextParser<string> QuotedCell =
        from open in QuotedCellDelimiter
        from content in QuotedCellContent.Try().Many()
        from end in QuotedCellDelimiter
        select new string(content);

    static readonly TextParser<string> NewLine =
        Parse.Return(Environment.NewLine);

    static readonly TextParser<string> RecordTerminator =
        Parse.Return("").AtEnd().Or(
        NewLine.AtEnd()).Try().Or(
        NewLine);

    static readonly TextParser<string> Cell =
        QuotedCell.Or(
        LiteralCellContent.Many().Select(x => new string(x)));

    public static readonly TextParser<IEnumerable<string>> Record =
        from leading in Cell
        from rest in CellSeparator.Then(_ => Cell).Try().Many()
        from terminator in RecordTerminator
        select Cons(leading, rest);

    static IEnumerable<T> Cons<T>(T head, IEnumerable<T> rest)
    {
        yield return head;
        foreach (var item in rest)
            yield return item;
    }

    public static readonly TextParser<IEnumerable<IEnumerable<string>>> Csv =
        Record.Many().AtEnd().Select(x => x.AsEnumerable());
}

I’ve tried following the Sprache Conversion very carefully, but, I cannot get the TextParser<IEnumerable<IEnumerable<string>>> Csv to work.

Exception:

Exception has occurred: CLR/Superpower.ParseException
An unhandled exception of type 'Superpower.ParseException' occurred in Superpower.dll: 'Many() cannot be applied to zero-width parsers; value ParserConcept.CsvParser+<Cons>d__11`1[System.String] at position 5 (line 1, column 6).'

I have a feeling this is going to be a very simple problem and hoping you can provide some insight. Anything seem wrong to you? I may try and just implement one from scratch instead of converting.

P.S. Thanks for the awesome library, it’s the first combinator I’ve used in C# and falling in love with it. It is making parsing crazy custom files I have much easier and with less code.

[nblumhardt]

Closing this as stale since we haven’t looped back to it in a while, but hope all’s going well!

[nblumhardt]

Thanks! 😄

I might have overstated the suggestion to start with token parsers; for simple things that stay simple, text parsers are probably just as good from the perspective of this library. Token parsers are able to make use of more features of Superpower, though, including the nicer error reporting and the easy handling of whitespace and comments. Not a big difference, but probably enough that I’d try starting with a token parser in most cases.

Keen to hear how you go, making the API as usable/approachable as Sprache’s is an ongoing project 😃 Cheers!