Creating a Predicate Builder extension method

Question

I have a Kendo UI Grid that I am currently allowing filtering on multiple columns. I am wondering if there is a an alternative approach removing the outer switch statement?

Basically I want to able to create an extension method so I can filter on a IQueryable<T> and I want to drop the outer case statement so I don't have to switch column names.

    private static IQueryable<Contact> FilterContactList(FilterDescriptor filter, IQueryable<Contact> contactList)
    {
        switch (filter.Member)
        {
            case "Name":
                switch (filter.Operator)
                {
                    case FilterOperator.StartsWith:
                        contactList = contactList.Where(w => w.Firstname.StartsWith(filter.Value.ToString()) || w.Lastname.StartsWith(filter.Value.ToString()) || (w.Firstname + " " + w.Lastname).StartsWith(filter.Value.ToString()));
                        break;
                    case FilterOperator.Contains:
                        contactList = contactList.Where(w => w.Firstname.Contains(filter.Value.ToString()) || w.Lastname.Contains(filter.Value.ToString()) || (w.Firstname + " " + w.Lastname).Contains( filter.Value.ToString()));
                        break;
                    case FilterOperator.IsEqualTo:
                        contactList = contactList.Where(w => w.Firstname == filter.Value.ToString() || w.Lastname == filter.Value.ToString() || (w.Firstname + " " + w.Lastname) == filter.Value.ToString());
                        break;
                }
                break;
            case "Company":
                switch (filter.Operator)
                {
                    case FilterOperator.StartsWith:
                        contactList = contactList.Where(w => w.Company.StartsWith(filter.Value.ToString()));
                        break;
                    case FilterOperator.Contains:
                        contactList = contactList.Where(w => w.Company.Contains(filter.Value.ToString()));
                        break;
                    case FilterOperator.IsEqualTo:
                        contactList = contactList.Where(w => w.Company == filter.Value.ToString());
                        break;
                }

                break;
        }
        return contactList;
    }

Some additional information, I am using NHibernate Linq. Also another problem is that the "Name" column on my grid is actually "Firstname" + " " + "LastName" on my contact entity. We can also assume that all filterable columns will be strings.

EDIT Remember this needs to work with NHibernate Linq and AST.

Answer 1

Answering your specific question,

private static IQueryable<Contact> FilterContactList(
    FilterDescriptor filter,
    IQueryable<Contact> contactList,
    Func<Contact, IEnumerable<string>> selector,
    Predicate<string> predicate)
{
    return from contact in contactList
           where selector(contract).Any(predicate)
           select contact;
}

In the case of "Name", you call it as;

FilterContactList(
    filter,
    contactList,
    (contact) => new []
        {
            contact.FirstName,
            contact.LastName,
            contact.FirstName + " " + contact.LastName
        },
    string.StartWith);

You should add an overload as,

private static IQueryable<Contact> FilterContactList(
    FilterDescriptor filter,
    IQueryable<Contact> contactList,
    Func<Contact, string> selector,
    Predicate<string> predicate)
{
    return from contact in contactList
           where predicate(selector(contract))
           select contact;
}

So you can call it like this for "Company" field.

FilterContactList(
    filter,
    contactList,
    (contact) => contact.Company,
    string.StartWith);

This prevents the overhead of forcing the caller to create an array when they only intend to seleect one Field/Property.

What you're probably after is something as follows

To remove that logic completely around defining the selector and predicate need more info on how filter is constructed. If possible filter should have the selector and predicate as properties for FilterContactList to use that get automatically constructed.

Expanding on that a little,

public class FilterDescriptor
{
    public FilterDescriptor(
        string columnName,
        FilterOperator filterOperator,
        string value)
    {
        switch (columnName)
        {
            case "Name":
                Selector = contact => new []
                               {
                                   contact.FirstName,
                                   contact.LastName,
                                   contact.FirstName + " " + contact.LastName
                               };
                break;
            default :
                // some code that uses reflection, avoids having
                // a case for every column name

                // Retrieve the public instance property of a matching name
                // (case sensetive) and its type is string.
                var property = typeof(Contact)
                    .GetProperties(BindingFlags.Public | BindingFlags.Instance)
                    .FirstOrDefault(prop =>
                        string.Equals(prop.Name, columnName) &&
                        prop.PropertyType == typeof(string));

                if (property == null)
                {
                    throw new InvalidOperationException(
                        "Column name does not exist");
                }

                Selector = contact => new[]
                {
                    (string)property.GetValue(contact, null)
                };
                break;
        }

        switch (filterOperator)
        {
            case FilterOperator.StartsWith:
                Predicate = s => s.StartsWith(filter.Value);
                break;
            case FilterOperator.Contains:
                Predicate = s => s.Contains(filter.Value);
                break;
            case FilterOperator.IsEqualTo:
                Predicate = s => s.Equals(filter.Value);
                break;
        }
    }

    public Func<Contact, IEnumerable<string>> Selector { get; private set; }
    public Func<string, bool> Predicate { get; private set; }
}

Your FilterContactList would then become

private static IQueryable<Contact> FilterContactList(
    FilterDescriptor filter,
    IQueryable<Contact> contactList)
{
    return from contact in contactList
           where filter.Selector(contract).Any(filter.Predicate)
           select contact;
}

Answer 2

I think a simple way to do this would be to create a map of property names to Func's:

e.g.

private static Dictionary<string, Func<Contact, IEnumerable<string>>> propertyLookup = new Dictionary<string, Func<Contact, IEnumerable<string>>>();

static ClassName() 
{
   propertyLookup["Name"] = c => new [] { c.FirstName, c.LastName, c.FirstName + " " c.LastName };
   propertyLookup["Company"] = c => new [] { c.Company }; 
}

And change your code to:

 var propertyFunc = propertyLookup(filter.Member);

 case FilterOperator.StartsWith:
          contactList = contactList.Where(c => propertyFunc(c).Any(s => s.StartsWith(filter.Value));

You could also eliminate the switch altogether by creating a lookup for the matching function as well:

matchFuncLookup[FilterOperator.StartsWith] = (c, f) => c.StartsWith(f);
matchFuncLookup[FilterOperator.Contains] = (c, f) => c.Contains(f);

var matchFunc = matchFuncLookup[filter.Operator];

contactList = contactList.Where(c => propertyFunc(c).Any(s => matchFunc(s, filter.Value));

So, to put it all together:

public class ClassName
{
    private static readonly Dictionary<string, Func<Contact, IEnumerable<string>>> PropertyLookup
        = new Dictionary<string, Func<Contact, IEnumerable<string>>>();
    private static readonly Dictionary<FilterOperator, Func<string, string, bool>> MatchFuncLookup
        = new Dictionary<FilterOperator, Func<string, string, bool>>();

    static ClassName()
    {
        PropertyLookup["Name"] = c => new[] { c.FirstName, c.LastName, c.FirstName + " " + c.LastName };
        PropertyLookup["Company"] = c => new[] { c.Company };
        MatchFuncLookup[FilterOperator.StartsWith] = (c, f) => c.StartsWith(f);
        MatchFuncLookup[FilterOperator.Contains] = (c, f) => c.Contains(f);
        MatchFuncLookup[FilterOperator.IsEqualTo] = (c, f) => c == f;
    }

    private static IQueryable<Contact> FilterContactList(FilterDescriptor filter, IQueryable<Contact> contactList)
    {
        var propertyLookup = PropertyLookup[filter.Member];
        var matchFunc = MatchFuncLookup[filter.Operator];
        return contactList.Where(c => propertyLookup(c).Any(v => matchFunc(v, filter.Value)));
    }
} 

NB - Is it not redundant to check c.FirstName if you are also checking (c.FirstName + " " c.LastName) ?