CsvHelper从抽象类中保存派生类对象

Question

[Serializable]
public abstract class AbstractModel : ObservableObject
{
    // nothing.
}

public class RealModel : AbstractModel 
{
    public string PropertyA {get; set;}
    public string PropertyB {get; set;}
}

Note that ObservableObject is from Mvvm-light.

With above models, I used CsvHelper as below.

AbstractModel instance = new RealModel()
                             {
                                 PropertyA = "foo",
                                 PropertyA = "bar"
                             };

using (TextWriter file = new StreamWriter("path"))
using (var csv = new CsvWriter(file))
{
    csv.WriteRecord(instance);
}

It throws error as below;

No properties are mapped for type 'AbstractModel'

It works fine when I set RealModel instance = new RealModel(); instead. But, I have various derived classes and want to save them in a single Save method.

How can I do?

Answer 1

Unfortunately I would say, this is the way it is supposed to be. The output is not an XML file, but a CSV file. And CSV file has only a single header (when it has any; not all CSV files have a header).

The parser can not determine class by data in line.

Standard CSV data example 1 (names):

ID,GivenName,FamilyName
1,John,A
2,Mike,B

Standard CSV data example 2 ("anniversaries"):

Date,Anniversary
1.1.,New year

When you save to CSV file, you can not have two "header lines"; how should the parser know, to which child class decode the output? Names or anniversaries? Why is the first line "name" and the second "anniversary"? Why is it not another way around?

This approach is not good. At least not for a CSV file. There should be always a table with exact count of column for each line. When the count of data is always exact, you can me converters from your class to a universal class and from universal class to specific data classes (for parser).

Non standard data example: (names and anniversaries; no header line):

1,John,A
1.1.,New year
2,Mike,B

Try to ask your self, how should the parser handle this file? Yes the data are stored in "the CSV way", but I don't think, these data are correct. This is not the way it should be.

I know there are situations, when you need to use this approach (formatted data; first item specify type, all other values in line are data for given type). But for this usage you need to make a different approach to store and parse the data.

A example of different approach:

public abstract class AbstractModel : ObservableObject
{
    // no property

    protected string Prefix(object data)
    {
       if (ReferenceEquals(data, null))
         return string.Empty;
       string result = data.ToString();
       if (data.Contains(",") || data.Contains("\""))
         return $"\"{data}\"";
       return data;
    }

}

public class RealModel : AbstractModel 
{
    public string PropertyA {get; set;}
    public string PropertyB {get; set;}

    public override string ToString()
    {
       return $"{Prefix(PropertyA)},{Prefix(PropertyB)}";
    }
}

And save with:

using (StreamWriter file = new StreamWriter("path"))
{
  foreach (AbstractModel instance in allInstances)
    file.WriteLine(instance);
}

Answer 2

I know this is three years old, but I was attempting to do the same thing, and I've found a hack to get it to work.

The offending code in csvHelper is the following:

public virtual Type GetTypeForRecord<T>(T record)
{
    var type = typeof(T);
    if (type == typeof(object))
    {
        type = record.GetType();
    }

    return type;
}

The C# generic that is passed down to this method is determined at compile time, so T is always going to be the base class, not the derived class, but if you cast to to type object before calling WriteRecord , the method will use GetType() , which will return the derived type.

AbstractModel instance = new RealModel()
                             {
                                 PropertyA = "foo",
                                 PropertyA = "bar"
                             };

using (TextWriter file = new StreamWriter("path"))
using (var csv = new CsvWriter(file))
{
    csv.WriteRecord((object)instance);
}