Migrating to v28.0.0

CsvHelper 28 changes ShouldSkipRecord: the function gets the reader row instead of the record's fields, and it is null by default, so it isn't called at all unless you set it. CsvParser and CsvReader take your own IParserConfiguration and IReaderConfiguration implementations instead of only CsvConfiguration, so IParserConfiguration gets the CultureInfo and the Validate method that the parser needs. The parser also caches the fields it processes, and delimiter detection looks at every line it reads. The target frameworks don't change.

Skipping records

ShouldSkipRecordArgs.Record, the record's fields as a string[], is replaced by ShouldSkipRecordArgs.Row, the IReaderRow being read. args.Row.Parser.Record has the fields that args.Record had, and you can also read single fields from the row, such as args.Row[0].

var config = new CsvConfiguration(CultureInfo.InvariantCulture)
{
    ShouldSkipRecord = args => args.Record.All(string.IsNullOrWhiteSpace),
};

var config = new CsvConfiguration(CultureInfo.InvariantCulture)
{
    ShouldSkipRecord = args => args.Row.Parser.Record.All(string.IsNullOrWhiteSpace),
};

ConfigurationFunctions.ShouldSkipRecord, the default function that returned false, is removed. Where you assigned it, to reset the setting, assign null, which is the default now.

Custom configuration classes

A class that implements IParserConfiguration can be passed to new CsvParser(reader, configuration), and one that implements IReaderConfiguration to new CsvReader(reader, configuration); 27 took only a CsvConfiguration. The parser needs two more members for that:

public class ParserConfiguration : IParserConfiguration
{
    public bool CacheFields => false;
    public bool LeaveOpen => false;
    public string NewLine => "\r\n";
    public bool IsNewLineSet => false;
    public CsvMode Mode => CsvMode.RFC4180;
    public int BufferSize => 0x1000;
    public int ProcessFieldBufferSize => 1024;
    public bool CountBytes => false;
    public Encoding Encoding => Encoding.UTF8;
    public BadDataFound BadDataFound => ConfigurationFunctions.BadDataFound;
    public bool LineBreakInQuotedFieldIsBadData => false;
    public char Comment => '#';
    public bool AllowComments => false;
    public bool IgnoreBlankLines => true;
    public char Quote => '"';
    public string Delimiter => ",";
    public bool DetectDelimiter => false;
    public string[] DetectDelimiterValues => new[] { ",", ";", "|", "\t" };
    public char Escape => '"';
    public TrimOptions TrimOptions => TrimOptions.None;
    public char[] WhiteSpaceChars => new[] { ' ' };
    public bool ExceptionMessagesContainRawData => true;
}

public class ParserConfiguration : IParserConfiguration
{
    public CultureInfo CultureInfo => CultureInfo.InvariantCulture; // New in 28.
    public bool CacheFields => false;
    public bool LeaveOpen => false;
    public string NewLine => "\r\n";
    public bool IsNewLineSet => false;
    public CsvMode Mode => CsvMode.RFC4180;
    public int BufferSize => 0x1000;
    public int ProcessFieldBufferSize => 1024;
    public bool CountBytes => false;
    public Encoding Encoding => Encoding.UTF8;
    public BadDataFound BadDataFound => ConfigurationFunctions.BadDataFound;
    public bool LineBreakInQuotedFieldIsBadData => false;
    public char Comment => '#';
    public bool AllowComments => false;
    public bool IgnoreBlankLines => true;
    public char Quote => '"';
    public string Delimiter => ",";
    public bool DetectDelimiter => false;
    public string[] DetectDelimiterValues => new[] { ",", ";", "|", "\t" };
    public char Escape => '"';
    public TrimOptions TrimOptions => TrimOptions.None;
    public char[] WhiteSpaceChars => new[] { ' ' };
    public bool ExceptionMessagesContainRawData => true;

    public void Validate() // New in 28.
    {
        if (Delimiter == Quote.ToString())
        {
            throw new ConfigurationException("The delimiter and the quote can't be the same.");
        }
    }
}

using var parser = new CsvParser(reader, new ParserConfiguration());

Behavior changes

• IParser.Record is cached: reading it twice for the same row gives the same array, until the next Read. If you change that array, copy it first, or the next read of Record for that row returns your changes.
• BadDataFound is called once per bad field. 27 called it each time the field was read, for example once for GetField and again for Record.
• Delimiter detection (DetectDelimiter = true) ignores text in quotes (or escaped text in CsvMode.Escape), only considers delimiters that are on every line it reads, and picks the culture's list separator when that is one of them. It can choose a different delimiter than 27 did for the same file.
• No header, mapped names: with HasHeaderRecord = false, a member mapped with Name is read by its index (its position in the map, unless you set Index). 27 threw a ReaderException saying there is no header record to find the name in.
• Nullable annotations in 28.0.0 only: 28.0.0 is built with nullable reference types enabled. In a project that enables them too, you get warnings for values such as HeaderRecord and Parser.Record, which are string[]?, and for BadDataFound = null. 28.0.1 builds without the annotations again, as 27 did.