I actually recently created something like this for my job. I was writing a custom CSV serializer class to support an ill-conceived CSV file specification.
public CSVSerializer Serialize<T>(T data, Func<T, object> map)
{
if (map == null) throw new ArgumentNullException("map");
object mappedData = map(data);
if (mappedData == null) throw new NullReferenceException("Mapped data produced null value");
// Iterate over public members of `mappedData`
MemberInfo[] members = mappedData.GetType().GetMembers(BindingFlags.Instance | BindingFlags.Public);
List<string> values = new List<string>();
foreach (MemberInfo member in members)
{
// Skip events and methods
if (!(member is FieldInfo || member is PropertyInfo)) continue;
// Value of `mappedData`
object memberVal = MemberInfoValue(member, mappedData);
if (memberVal == null)
{
// If the actual value stored by `memberVal` is null, store string.Empty and continue
values.Add(string.Empty);
continue;
}
// Check if `memberVal` contains a member named "map"
MemberInfo[] maps = memberVal.GetType().GetMember("map");
MemberInfo memberMap = maps.Length > 0 ? maps[0] : null;
string val = MapToString(memberVal, o => o.ToString());
if (map != null) // map is present
{
// Get first property other than map
MemberInfo dataVal = memberVal.GetType().GetMembers(BindingFlags.Instance | BindingFlags.Public)
.Where(mi => mi is FieldInfo || mi is PropertyInfo)
.Except(new MemberInfo[] { memberMap })
.DefaultIfEmpty(memberMap)
.FirstOrDefault();
object tmp = MemberInfoValue(memberMap, memberVal);
if (dataVal == memberMap)
{
// map is only property, so serialize it
val = MapToString(tmp, o => o.ToString());
}
else
{
// try to serialize map(dataVal), or use empty string if it fails
Delegate dlg = tmp as Delegate;
if (dlg != null)
{
object param = MemberInfoValue(dataVal, memberVal);
try { val = MapToString(dlg, d => d.DynamicInvoke(param).ToString()); }
catch (Exception ex)
{
// exception should only occur with parameter count/type mismatch
throw new SerializationException(string.Format("Poorly formatted map function in {0}", member.Name), ex);
}
}
else
{
// map is not a delegate (!!)
throw new SerializationException(string.Format("map member in {0} is not a delegate type", member.Name));
}
}
}
// Handle quotes and the separator string
val = val.Trim('"');
if (val.Contains("\""))
{
val = val.Replace("\"", "\\\"");
}
if (val.Contains(Separator))
{
val = string.Format("\"{0}\"", val);
}
values.Add(val);
}
string line = string.Join(Separator, values);
Writer.WriteLine(line);
return this;
}
This function serializes the public fields and properties of data
in the order they're defined; the names of those members are ignored.
If you're only interested in properties, you can use GetProperties
instead of GetMembers
(using PropertyInfo
instead of MemberInfo
). If you're not creating a CSV, you can obviously ignore the CSV file formatting portion.
MemberInfoValue(info, parent)
returns ((FieldInfo)info).GetValue(parent)
or ((PropertyInfo)info).GetValue(parent, null)
, as appropriate. MapToString
is just a null-guard function.
The presence of the map
parameter and hunting for a member named "map" is part of the necessities of my particular situation, although it may be useful to you. The mapping allows for things such as:
mySerializer.Serialize(myPersonObject,
p => new {
name = string.Format("{0}, {1}", p.LastName, p.FirstName),
address = p.address
});
I have an overload of Serialize
which calls Serialize(data, d => d);
. Hunting for the "map" member allows things like:
mySerializer.Serialize(new {
employeeID = new { val = myEmployeeObject.ID, map = d => d.ToUpperCase() },
employeeName = myEmployeeObject.Name
});