The art of Deconstructing

In this article with code samples provides additional ways to return data from methods, iterating foreach statements, how to work with date time objects to have cleaner code.

An important consideration is every developer has their own style of coding and with that keep an open mind, if something learned looks inviting but not your style than modify the code to suite you’re or a team’s style.

Note
Code sample are done in console and Windows Forms projects as there is no need to do this in web or other desktop projects as they simple create too much noise which lessens the learning experience.

What is deconstruct/unpacking in C#?

Deconstruction is a process of splitting a variable value into parts and storing them into new variables.

This could be useful when a variable stores multiple values such as a Tuple and different ways to work with foreach statements or create language extensions for returning smaller sets of information from third party methods.

Most developers know about how to return informaton with Tuples but many don't know about other benefits which will be gone over in this article.

Microsoft documentation Deconstructing tuples and other types.

Warm and fuzzy

Pretty much at one time or another a developer needs to break apart a DateTime or DateTimeOffset as shown below.

var date = new DateTimeOffset(Now.Year, Now.Month, Now.Day, 0, 0, 0, 0, TimeSpan.Zero);
int year = date.Year;

We can deconstruct a DateTime or a DateTimeOffset using an extension method.

public static class Extensions
{
    public static void Deconstruct(this DateTimeOffset date, out int day, out int month, out int year) 
        => (day, month, year) = (date.Day, date.Month, date.Year);
}

Example usage

var date = new DateTimeOffset(Now.Year, Now.Month, Now.Day, 0, 0, 0, 0, TimeSpan.Zero);

date.Deconstruct(out int day, out int month, out int year);

Console.WriteLine($"Year: {year} Month: {month} Day: {day}");

How about a DateOnly? Yes we can do this also along with TimeOnly

public static void Deconstruct(this DateOnly date, out int day, out int month, out int year) =>
    (day, month, year) = (date.Day, date.Month, date.Year);

This means a developer who works with dates can have cleaner code. But wait, why not a language extension method like below? Because this article focuses on deconstructing 😉

public static class Extensions
{
    public static (int day, int month, int year) Chunk(this DateTimeOffset sender) 
        => (sender.Day, sender.Month, sender.Year);
}

Basics

The following shows different ways to return data from a method.

In conventional coding returning information typically looks like the following which is devoid of any exception handling.

public class DataOperations
{
    public static async Task<List<Customers>> GetCustomersList()
    {
        await using NorthContext context = new();
        return await context
            .Customers
            .Include(c => c.ContactTypeIdentifierNavigation)
            .ToListAsync();
    }
}

But we all know that at anytime an exception may be thrown, in the above case perhaps the server where the database resides is not available.

So what can be done? Write to a log file and return null when an exception is thrown.

public class DataOperations
{
    public static async Task<List<Customers>> GetCustomersList()
    {
        try
        {
            await using NorthContext context = new();
            return await context
                .Customers
                .Include(c => c.ContactTypeIdentifierNavigation)
                .ToListAsync();
        }
        catch (Exception ex)
        {
            // write to error log
            return null;
        }
    }
}

The above although works is not a nice solution, let's try using deconstruction.

public class DataOperations
{
    public static async Task<(List<Customers> customers, Exception exception)> GetCustomersList()
    {
        try
        {
            await using NorthContext context = new();
            return 
                (
                    await context.
                        Customers
                        .Include(c => c.ContactTypeIdentifierNavigation)
                        .ToListAsync(), 
                    null
                );
        }
        catch (Exception ex)
        {
            // write to error log
            return (null, ex);
        }
    }
}

Usage

var (customers, exception) = await DataOperations.GetCustomersList();
if (exception is null)
{
    // use customers
}
else
{
    // we have an exception object to see what happened
}

In this case we have returned either a list of customers and null or null and an exception.

Can we do better? Yes and no, it depends on your programming style. Here an the first value is a bool which represents success or failure.

public class DataOperations
{
    public static async Task<(bool success, List<Customers> customers, Exception exception)> GetCustomersList()
    {
        try
        {
            await using NorthContext context = new();
            return 
                (
                    true,
                    await context.
                        Customers
                        .Include(c => c.ContactTypeIdentifierNavigation)
                        .ToListAsync(), 
                    null
                );
        }
        catch (Exception ex)
        {
            // write to error log
            return (false,null, ex);
        }
    }
}

Personally, it's clearer adding the bool.

var (success, customers, exception) = await DataOperations.GetCustomersList();
if (success)
{
    // use customers
}
else
{
    // check out exception
}

Each of the above examples perform the same operations with varying ways to return data. The first should be avoided unless it’s never going to fail. To be honest, the second is what an uneducated coder might come up with and is not recommended. The final two are personal choices along with showing we can return information and name them like local variables.

Suppose we want to simply check success and not access the exception in regards to the last example? We can use a discard.

Deconstructing non-tuples for classes/models

This feature is great, but it is actually not limited to just tuples - you can add deconstructors to all your classes. Using the following syntax we can return and deconstruct only the information needed for an operation. The benefit is not needing to return all properties of a model and that there may be several different operations needing different sets of data.

public class PersonEntity
{
    public int PersonID { get; set; }
    public string FirstName { get; set; }
    public string LastName { get; set; }
    public string FullName => $"{FirstName} {LastName}";
    public ICollection<StudentGrade> Grades { get; set; }
    public override string ToString() => $"{FirstName} {LastName}";

    public void Deconstruct(out int id, out string firstName, out string lastName)
    {
        id = PersonID;
        firstName = FirstName;
        lastName = LastName;
    }
    public void Deconstruct(out int id, out string fullName)
    {
        id = PersonID;
        fullName = FullName;
    }
    public void Deconstruct(out int id, out string firstName, string lastName, ICollection<StudentGrade> grades)
    {
        id = PersonID;
        firstName = FirstName;
        lastName = LastName;
        grades =Grades;
    }
}

Usage using a discard which means we don't want first name, only the primary key and last name.

var (id, _, last) = (PersonEntity)SomeControl.SelectedItem;

Caveat, the above can only be done with classes you have access too. Suppose there is a class you don't have source code too? The solution is to create a language extension.

Here is a model in a third party library

namespace SomeThirdPartyLibrary.Classes
{
    public class Customer
    {
        public int CustomerIdentifier { get; set; }
        public string CompanyName { get; set; }
        public int? ContactId { get; set; }
        public string Street { get; set; }
        public string City { get; set; }
        public string PostalCode { get; set; }
        public int? CountryIdentifier { get; set; }
        public override string ToString() => CompanyName;
    }
}

In your project we create an extension method for Customer.

using SomeThirdPartyLibrary.Classes;

namespace DeconstructCodeSamples.Extensions
{
    public static class ThirdPartyExtensions
    {
        /// <summary>
        ///  Extension for third party class
        /// </summary>
        /// <param name="customer"></param>
        /// <param name="id">customer key</param>
        /// <param name="companyName">customer's company name</param>
        /// <param name="contactIdentifier">customer's contact identifier</param>
        /// <param name="countryIdentifier">country identifier for customer</param>
        public static void Deconstruct(this Customer customer, out int id, out string companyName, out int? contactIdentifier, out int? countryIdentifier)
        {
            id = customer.CustomerIdentifier;
            companyName = customer.CompanyName;
            contactIdentifier = customer.ContactId;
            countryIdentifier = customer.CountryIdentifier;
        }
    }
}

Deconstucting other code

Let's look at a simple dictionary

var peopleDictionary = new Dictionary<string, int>
{
    ["Mary"] = 32, 
    ["Frank"] = 17
};

Convental method to iterate the key and values.

foreach (var pair in peopleDictionary)
{
    Console.WriteLine($"{pair.Key} is {pair.Value} years old");
}

Minor issue when looking at the code is say the dictionary definition is not visible, it can be difficult to tell what key and value are, and in a little bit I will drive this home with a more complex example using a switch and grouping.

Next level, deconstruct and provide meaningful variable names.

foreach (var (name, age) in peopleDictionary)
{
    Console.WriteLine($"{name} is {age} years old");
}

The above is easy to understand. Now for those who like the cool factor here you go.

foreach (var (name, age) in peopleDictionary.Select(x => (x.Key, x.Value)))
{
    Console.WriteLine($"{name} is {age} years old.");
}

❓ Which to use? the second example, it's clear and easy to read.

Dictionary/IGrouping deconstruct

In this example the model where the task is to group by price, Cheap, Medium and Expensive.

public partial class Book
{
    public int Id { get; set; }
    public string Title { get; set; }
    public decimal? Price { get; set; }
}

Code to read and group.

var books = await context.Book.ToListAsync();

Dictionary<string, IGrouping<string, Book>> results = books
    .GroupBy(book => book.Price switch
    {
        <= 10 => "Cheap",
        > 10 and <= 20 => "Medium",
        _ => "Expensive"
    })
    .ToDictionary(gb =>
            gb.Key,
        g => g);

Conventional foreach where like the prior conventional foreach we have is Key and Value which can make understand the code hard.

foreach (KeyValuePair<string, IGrouping<string, Book>> pair in results)
{
    Console.WriteLine(pair.Key);
    foreach (var book in pair.Value)
    {
        Console.WriteLine($"\t{book.Title,-25}{book.Price:C0}");
    }
}

Using Deconstruct the code is much easier to read.

foreach (var (pricingCategory, bookGrouping) in results)
{
    Console.WriteLine(pricingCategory);
    foreach (var book in bookGrouping)
    {
        Console.WriteLine($"\t{book.Title, -25}{book.Price:C0}");
    }
}

And we can specify each variable type too.

foreach ((string pricingCategory, IGrouping<string, Book> bookGrouping) in results)
{
    Console.WriteLine(pricingCategory);
    foreach (var book in bookGrouping)
    {
        Console.WriteLine($"\t{book.Title, -25}{book.Price:C0}");
    }
}

Records

We can deconstruct records also, given the following record Deconstruct provides us a way to return all information in a compact form.

public record Person()
{
    public int Id { get; init; }
    public string Firstname { get; init; }
    public string Lastname { get; init; }
    public string FullName => $"{Firstname} {Lastname}";

    public override string ToString() => $"{Id} {FullName}";

    public void Deconstruct(out int id, out string fullName)
        => (id, fullName) = (Id, FullName);

}

Usage with mocked data using Bogus NuGet package.

namespace ForWritingArticle.Classes
{
    public class Operations
    {
        public static List<Person> PeopleList(int count = 5)
        {
            var faker = new Faker<Person>()
                .WithRecord()
                .RuleFor(p => p.Id, f => f.IndexVariable++)
                .RuleFor(p => p.Lastname, f => f.Person.LastName)
                .RuleFor(p => p.Firstname, f => f.Person.FirstName);

            return faker.Generate(count);
        }

        public static void PeopleDeconstruct()
        {
            var list = PeopleList();

            foreach (Person person in list)
            {
                var (id, fullName) = person;
                Console.WriteLine($"{id,-4}{fullName}");
            }

        }
    }
}

Beginning in C# 10

You can mix variable declaration and assignment in a deconstruction.

public static void MixingDeclarationAndAssignment()
{
        string city = "Raleigh";
        int population = 458880;

        (city, population, double area) = QueryCityData("New York City");

        Console.WriteLine();

}
private static (string, int, double) QueryCityData(string name) 
    => name == "New York City" ? (name, 8175133, 468.48) : ("", 0, 0);

Resharper

Generate Deconstructors ReSharper_GenerateDeconstructor

Before generation

public class Version
{
    public int Major { get; }
    public int Minor { get; }
}

After generation

public class Version
{
    public int Major { get; }
    public int Minor { get; }

    public void Deconstruct(out int major, out int minor)
    {
        major = this.Major;
        minor = this.Minor;
    }
}

In the above example the developer may choice all properties or select only those they want.

Without Resharper as an extension method.

public static void Deconstruct(

    this Version version, 

    out int major, 

    out int minor, 

    out int build, 

    out int revision)

    => 

        (

            major, minor, 

            build, revision

        ) = 

        (

            version.Major, 

            version.Minor, 

            version.Build, 

            version.Revision

        );

Notes

• If using a framework prior to 4.8, you will need the following NuGet package System.ValueTuple, otherwise the package is included in new projects

Source code

Clone the following GitHub repository which requires Visual Studio 2022 or higher.