I'm working with some C# code that's using .Net 4 Lazy loads and I'm not super familiar with it. I'm trying to figure out if this particular code is useless or not.
Originally the property and code below where in the same class, but now I've moved the code to an external class that no longer has access to the private "lazyRecords" property. I'm wondering what the point of checking "lazyRecords.IsValueCreated" is since the lazyRecords.Value has not been invoked yet, wouldn't it always be false? Or is it checking to see if another thread somehow invoked the Value? Or is it doing this in case of a thread exception that resulted in not loading the object?
Property:
private Lazy<List<Record>> lazyRecords;
public List<Record> Records
{
get
{
return lazyRecords.Value;
}
set
{
lazyRecords = new Lazy<List<Record>>(() => value);
}
}
Code:
public Category LoadCategory(BaseClient client)
{
Category category = new Category();
category.Records = client.RecordClient.GetRecordsByCategoryID(category.ID);
if (lazyRecords.IsValueCreated)
{
category.WorldRecord = category.Records.FirstOrDefault();
}
else
{
category.WorldRecord = client.RecordClient.GetWorldRecord(category.ID);
}
}
The code is pretty useless, yes. To help you understand why, consider this very minimal version of Lazy (the real class has more options and logic to take care of multiple threads, but this is the rough idea):
public class Lazy<T>
{
private readonly Func<T> _creator;
private T _cachedValue;
public Lazy(Func<T> creator) => _creator = creator;
public bool IsValueCreated { get; private set; }
public T Value
{
get
{
if (!IsValueCreated)
{
_cachedValue = _creator();
IsValueCreated = true;
}
return _cachedValue;
}
}
}
The delegate passed to the constructor is called on demand, the first time the Value is requested. In the code you've posted there is no point to this because the delegate simply returns the value passed into the setter.
As to the LoadCategory method, the code you posted is hard to decipher. It directly accesses lazyRecords, implying it's a method of the same class. But then it accesses Records on a different object.
I have a property which getter should load its value only the first time. The second time it returns the loaded value without loading it again:
private Object _MemberValue;
public Object MemberValue
{
get
{
if(_MemberValue == null)
{
_MemberValue = LoadMember();
}
return _MemberValue;
}
}
In VB.NET there is the Static keyword. With it you don't have to declare a class wide member.
Public Property MemberValue as Object
Get
Static value as Object = Nothing
If (value is Nothing) Then
value = LoadMember()
End If
Return value
End Get
End Property
In C# there isn't such a keyword.
Are there better C# implementations of this problem or other patterns?
Are there better C# implementations of this problem or other patterns?
Probably not. You can use Lazy<T> as an replacement if you like, but basically it is the same as your first example. Using Static in VB.NET has some serious drawbacks, so I wouldn't use it either way.
If you prefer Lazy<T>, this is what I would use:
private Lazy<object> _MemberLazy = new Lazy<object>(LoadMember);
public object MemberValue
{
get
{
return _MemberLazy.Value;
}
}
Your initial approach seems appropriate, I have never had reason to do something different. That said if your goal here is to avoid a class level field that could potentially be written to outside the getter, perhaps something like this would work. There are a number of other ReadOnly, WriteOnce, SetOnce implementations that would also work similarly.
ReadOnlyField.cs
public class ReadOnlyField<T>
{
private bool _frozen;
private T _value;
public T Value
{
get { return _value; }
set
{
if (_frozen)
throw new InvalidOperationException();
_value = value;
}
}
public void Freeze()
{
_frozen = true;
}
}
YourObject.cs
public class YourObject
{
private readonly ReadOnlyField<object> _someMember;
public object MemberValue
{
get
{
if(_someMember.Value == null)
{
_someMember.Value = LoadMember();
_someMember.Freeze();
}
return _someMember.Value;
}
}
public YourObject()
{
_someMember = new ReadOnlyField<object>();
}
}
It's not perfect. Unlike your VB.Net example; code outside of the getter could write to the field first, but at least you're protected from it being overwritten after Freeze is called.
When .NET 4.5 was released i started using such great Attribute as CallerMemberName. It's easier to understand code, developers can write it faster also. It's like a snippet, not only a feature for debug/test purposes.
So I have a question. Is it normal to create and use something like this?
public class PropertyStore
{
Dictionary<string, object> data = new Dictionary<string,object>();
ViewModelBase modelBase;
internal PropertyStore(ViewModelBase _base)
{
modelBase = _base;
}
public void SetValue<T>(T value = default(T), [CallerMemberName] string prop = "")
{
T prev = GetValue<T>(prop);
if ((prev == null && value == null) || (prev != null && prev.Equals(value))) return;
data[prop] = value;
modelBase.OnPropertyChanged(prop);
}
public T GetValue<T>([CallerMemberName] string prop = "")
{
if (!data.ContainsKey(prop))
data[prop] = default(T);
return (T)data[prop];
}
}
Class-helper, that makes other class more readable, and also we have list of our properties without need to use Reflection.
The usage is:
public class SampleClass : ViewModelBase
{
PropertyStore PropertyStore;
public SampleClass ()
{
PropertyStore = new PropertyStore(this);
}
public string Key
{
get { return PropertyStore.GetValue<string>(); }
set { PropertyStore.SetValue(value); }
}
public DateTime Date
{
get { return PropertyStore.GetValue<DateTime>(); }
set { PropertyStore.SetValue(value); }
}
public bool IsSelected
{
get { return PropertyStore.GetValue<bool>(); }
set { PropertyStore.SetValue(value); }
}
}
The class ViewModelBase here simply implements INotifyPropertyChanged interface.
As I understand, this approach is something like Microsoft Dependency Properties, but I don't need all power of DependencyObject class, and I don't want inherit it.
With something like this I can use Binding, because it's enough to implement INotifyPropertyChanged, also we have no fields (as for me, i try to use properties smarter, than using fields directly (however, there is no problem to use Dictionary directly ^_^))
Sorry for my bad English... Not main language and not much practice.
Another Sample (after moving Methods to base class)
public class SampleClass : ViewModelBase
{
public string Key
{
get { return GetValue<string>(); }
set { SetValue(value); }
}
public DateTime Date
{
get { return GetValue<DateTime>(); }
set { SetValue(value); }
}
public bool IsSelected
{
get { return GetValue<bool>(); }
set { SetValue(value); }
}
}
No diff with Microsoft's WPF Property System.
Only feature you'll get with it is an ability to access property values via Dictionary.Get|Set methods.
You can get this ability with field based implementation of INotifyPropertyChanged. You can access property values by its name using dictionary, with property name to precompiled delegate mapping like it done in Yappi project.
var dateValue= Property<SampleClass>.Get<DateTime>(this,"Date");
Property<SampleClass>.Set<DateTime>(this,"Date",DateTime.Now);
Both can be rewritten as extension methods.
Nice idea, property bag without reflection and it will even work with obfuscation.
I don't see major problems with it but you may consider the following:
The prop parameter is optional so potentially a bug can be introduced by given a value in the call.
Value types will get boxed.
Access to the fields is relatively more expensive, can be a factor more expensive as you have much more code in a simple get (especially with boxing).
Dictionary takes more space than the number of properties you keep in (especially with boxing).
Each property also stores a string of the property name adding to the overhead.
I have a class with like 20 fields which get populated from SQL database on load. Currently I am calling load data method right after the constructor, which calls SQL proc and populate all the required fields. At times, I may not access the these 20 fields at all, I am adding additional cost of SQL call even though it was not required. So I changed all the properties to have an associated private property and when the program calls the public property, first I check the private property and if it is null that means we need to load data from sql so I call the load method. It works great, but when I see the code, there is a repeated pattern of null check and load the sql query. Is there a better way of doing this?
private string _name;
public string Name
{
get {
if (_name == null)
LoadData(); //this popultes not just but all the properties
return _name;
}
}
Btw C# now has default lazy-loaders implementation. Why not to use it, instead of providing isSomethingLoaded flags? :)
public class Bar
{
private Lazy<string> _name = new Lazy<string>(() => LoadString());
public string Name
{
get { return _name.Value; }
}
}
In case of non-static LoadString method, lazy-loader should be initialized in constructor;
Nope, this is right. Here is the wikipedia article. The overhead of the null check will be very minimal compared to unnecessary database calls. Now, if the users of the program actually use the values 99% of the time, then I would say this pattern is not needed.
Just one note of caution: If any of your values could possibly be null, then you will make unnecessary database calls. It might be better to do something like this (which will be an even quicker check since it is just a bit check):
//Constructor default to not loaded
bool isLoaded = false;
private string _name;
public string Name
{
get {
if (!isLoaded)
LoadData(); //this popultes not just but all the properties
return _name;
}
}
private LoadData()
{
//Load Data
isLoaded = true;
}
Well you could change it to:
if (!initialized)
LoadData();
And in your LoadData set initialized to true, but that really doesn't change the semantics of it.
One thing you can do is to extract if into separate method so each property contains just one additional call:
void EnsureData()
{
if (!dataLoaded)
LoadData(); //this populates all the properties
}
public string Name {
get {
EnsureData();
return _name;
}
}
I think you should consider your application structure. Why would you even instantiate the class if you are not going to be using the properties? I believe it's actually cleaner for you to call the SQL after your constructor code but only create the objects of your class if you are going to be using it. The other more flexible solution is making the LoadData public and calling it as needed from the object instance as needed.
I am in the learning process of design patterns. i have one suggestion if you load data only once you can try with singleton design pattern.
public class Singleton123
{
private static readonly string _property1 = ClassLoadData.LoadData();
public static string MyProperty1
{
get
{
return _property1;
}
}
}
public class ClassLoadData
{
public static string LoadData()
{
// any logic to load data
return "test";
}
}
Call property as below
Singleton123 obj = new Singleton123();
string stra = Singleton123.MyProperty1;
string strb = Singleton123.MyProperty1;
this property will be loaded only once.
The following is a simple example of an enum which defines the state of an object and a class which shows the implementation of this enum.
public enum StatusEnum
{
Clean = 0,
Dirty = 1,
New = 2,
Deleted = 3,
Purged = 4
}
public class Example_Class
{
private StatusEnum _Status = StatusEnum.New;
private long _ID;
private string _Name;
public StatusEnum Status
{
get { return _Status; }
set { _Status = value; }
}
public long ID
{
get { return _ID; }
set { _ID = value; }
}
public string Name
{
get { return _Name; }
set { _Name = value; }
}
}
when populating the class object with data from the database, we set the enum value to "clean". with the goal of keeping most of the logic out of the presentation layer, how can we set the enum value to "dirty" when a property is changed.
i was thinking something along the lines of;
public string Name
{
get { return _Name; }
set
{
if (value != _Name)
{
_Name = value;
_Status = StatusEnum.Dirty;
}
}
}
in the setter of each property of the class.
does this sound like a good idea, does anyone have any better ideas on how the dirty flag can be assigned without doing so in the presentation layer.
When you really do want a dirty flag at the class level (or, for that matter, notifications) - you can use tricks like below to minimise the clutter in your properties (here showing both IsDirty and PropertyChanged, just for fun).
Obviously it is a trivial matter to use the enum approach (the only reason I didn't was to keep the example simple):
class SomeType : INotifyPropertyChanged {
private int foo;
public int Foo {
get { return foo; }
set { SetField(ref foo, value, "Foo"); }
}
private string bar;
public string Bar {
get { return bar; }
set { SetField(ref bar, value, "Bar"); }
}
public bool IsDirty { get; private set; }
public event PropertyChangedEventHandler PropertyChanged;
protected void SetField<T>(ref T field, T value, string propertyName) {
if (!EqualityComparer<T>.Default.Equals(field, value)) {
field = value;
IsDirty = true;
OnPropertyChanged(propertyName);
}
}
protected virtual void OnPropertyChanged(string propertyName) {
var handler = PropertyChanged;
if (handler != null) {
handler(this, new PropertyChangedEventArgs(propertyName));
}
}
}
You might also choose to push some of that into an abstract base class, but that is a separate discussion
One option is to change it on write; another is to keep a copy of all the original values and compute the dirtiness when anyone asks for it. That has the added benefit that you can tell exactly which fields have changed (and in what way) which means you can issue minimal update statements and make merge conflict resolution slightly easier.
You also get to put all the dirtiness-checking in one place, so it doesn't pollute the rest of your code.
I'm not saying it's perfect, but it's an option worth considering.
If you want to implement it in this way, and you want to reduce the amount of code, you might consider applying Aspect Oriented Programming.
You can for instance use a compile-time weaver like PostSharp , and create an 'aspect' that can be applied to properties. This aspect then makes sure that your dirty flag is set when appropriate.
The aspect can look like this:
[Serializable]
[AttributeUsage(AttributeTargets.Property)]
public class ChangeTrackingAttribute : OnMethodInvocationAspect
{
public override void OnInvocation( MethodInvocationEventArgs e )
{
if( e.Delegate.Method.ReturnParameter.ParameterType == typeof(void) )
{
// we're in the setter
IChangeTrackable target = e.Delegate.Target as IChangeTrackable;
// Implement some logic to retrieve the current value of
// the property
if( currentValue != e.GetArgumentArray()[0] )
{
target.Status = Status.Dirty;
}
base.OnInvocation (e);
}
}
}
Offcourse, this means that the classes for which you want to implement ChangeTracking, should implement the IChangeTrackable interface (custom interface), which has at least the 'Status' property.
You can also create a custom attribute ChangeTrackingProperty, and make sure that the aspect that has been created above, is only applied to properties that are decorated with this ChangeTrackingProperty attribute.
For instance:
public class Customer : IChangeTrackable
{
public DirtyState Status
{
get; set;
}
[ChangeTrackingProperty]
public string Name
{ get; set; }
}
This is a little bit how I see it.
You can even make sure that PostSharp checks at compile-time whether classes that have properties that are decorated with the ChangeTrackingProperty attribute, implement the IChangeTrackable interface.
This method is based on a set of different concepts provided in this thread. I thought i'd put it out there for anyone that is looking for a way to do this cleanly and efficiently, as i was myself.
The key of this hybrid concept is that:
You don't want to duplicate the data to avoid bloating and resource hogging;
You want to know when the object's properties have changed from a given original/clean state;
You want to have the IsDirty flag be both accurate, and require little processing time/power to return the value; and
You want to be able to tell the object when to consider itself clean again. This is especially useful when building/working within the UI.
Given those requirements, this is what i came up with, and it seems to be working perfectly for me, and has become very useful when working against UIs and capturing user changes accurately. I have also posted an "How to use" below to show you how I use this in the UI.
The Object
public class MySmartObject
{
public string Name { get; set; }
public int Number { get; set; }
private int clean_hashcode { get; set; }
public bool IsDirty { get { return !(this.clean_hashcode == this.GetHashCode()); } }
public MySmartObject()
{
this.Name = "";
this.Number = -1;
MakeMeClean();
}
public MySmartObject(string name, int number)
{
this.Name = name;
this.Number = number;
MakeMeClean();
}
public void MakeMeClean()
{
this.clean_hashcode = this.Name.GetHashCode() ^ this.Number.GetHashCode();
}
public override int GetHashCode()
{
return this.Name.GetHashCode() ^ this.Number.GetHashCode();
}
}
It's simple enough and addresses all of our requirements:
The data is NOT duplicated for the dirty check...
This takes into account all property changes scenarios (see scenarios below)...
When you call the IsDirty property, a very simple and small Equals operation is performed and it is fully customizable via the GetHashCode override...
By calling the MakeMeClean method, you now have a clean object again!
Of course you can adapt this to encompass a bunch of different states... it's really up to you. This example only shows how to have a proper IsDirty flag operation.
Scenarios
Let's go over some scenarios for this and see what comes back:
Scenario 1
New object is created using empty constructor,
Property Name changes from "" to "James",
call to IsDirty returns True! Accurate.
Scenario 2
New object is created using paramters of "John" and 12345,
Property Name changes from "John" to "James",
Property Name changes back from "James" to "John",
Call to IsDirty returns False. Accurate, and we didn't have to duplicate the data to do it either!
How to use, a WinForms UI example
This is only an example, you can use this in many different ways from a UI.
Let's say you have a two forms ([A] and [B]).
The first([A]) is your main form, and the second([B]) is a form that allows the user to change the values within the MySmartObject.
Both the [A] and the [B] form have the following property declared:
public MySmartObject UserKey { get; set; }
When the user clicks a button on the [A] form, an instance of the [B] form is created, its property is set and it is displayed as a dialog.
After form [B] returns, the [A] form updates its property based on the [B] form's IsDirty check. Like this:
private void btn_Expand_Click(object sender, EventArgs e)
{
SmartForm form = new SmartForm();
form.UserKey = this.UserKey;
if(form.ShowDialog() == DialogResult.OK && form.UserKey.IsDirty)
{
this.UserKey = form.UserKey;
//now that we have saved the "new" version, mark it as clean!
this.UserKey.MakeMeClean();
}
}
Also, in [B], when it is closing, you can check and prompt the user if they are closing the form with unsaved changes in it, like so:
private void BForm_FormClosing(object sender, FormClosingEventArgs e)
{
//If the user is closing the form via another means than the OK button, or the Cancel button (e.g.: Top-Right-X, Alt+F4, etc).
if (this.DialogResult != DialogResult.OK && this.DialogResult != DialogResult.Ignore)
{
//check if dirty first...
if (this.UserKey.IsDirty)
{
if (MessageBox.Show("You have unsaved changes. Close and lose changes?", "Unsaved Changes", MessageBoxButtons.YesNo, MessageBoxIcon.Warning) == DialogResult.No)
e.Cancel = true;
}
}
}
As you can see from the examples above, this can be a very useful thing to have since it really streamlines the UI.
Caveats
Every time you implement this, you have to customize it to the object you're using. E.g.: there's no "easy" generic way of doing this without using reflection... and if you use reflection, you lose efficiency, especially in large and complex objects.
Hopefully this helps someone.
Take a look at PostSharp (http://www.postsharp.org/).
You can easily create a Attribute which marks it as dirty you can add the attrubute to each property that needs it and it keeps all your code in one place.
Roughly speaking Create an interface which has your status in make the class implement it.
Create an attribute which can be applied on properties and cast to your interface in order to set the value when something changes one of the marked properties.
Your approach is basically how I would do it. I would just
remove the setter for the Status property:
public StatusEnum Status
{
get { return _Status; }
// set { _Status = value; }
}
and instead add a function
public SetStatusClean()
{
_Status = StatusEnum.Clean;
}
As well as SetStatusDeleted() and SetStatusPurged(), because I find it better indicates the intention.
Edit
Having read the answer by Jon Skeet, I need to reconsider my approach ;-) For simple objects I would stick with my way, but if it gets more complex, his proposal would lead to much better organised code.
If your Example_Class is lightweight, consider storing the original state and then comparing the current state to the original in order to determine the changes. If not your approach is the best because stroing the original state consumes a lot of system resources in this case.
Apart from the advice of 'consider making your type immutable', here's something I wrote up (and got Jon and Marc to teach me something along the way)
public class Example_Class
{ // snip
// all properties are public get and private set
private Dictionary<string, Delegate> m_PropertySetterMap;
public Example_Class()
{
m_PropertySetterMap = new Dictionary<string, Delegate>();
InitializeSettableProperties();
}
public Example_Class(long id, string name):this()
{ this.ID = id; this.Name = name; }
private void InitializeSettableProperties()
{
AddToPropertyMap<long>("ID", value => { this.ID = value; });
AddToPropertyMap<string>("Name", value => { this.Name = value; });
}
// jump thru a hoop because it won't let me cast an anonymous method to an Action<T>/Delegate
private void AddToPropertyMap<T>(string sPropertyName, Action<T> setterAction)
{ m_PropertySetterMap.Add(sPropertyName, setterAction); }
public void SetProperty<T>(string propertyName, T value)
{
(m_PropertySetterMap[propertyName] as Action<T>).Invoke(value);
this.Status = StatusEnum.Dirty;
}
}
You get the idea.. possible improvements: Use constants for PropertyNames & check if property has really changed.
One drawback here is that
obj.SetProperty("ID", 700); // will blow up int instead of long
obj.SetProperty<long>("ID", 700); // be explicit or use 700L
Here is how i do it.
In cases where i do not need to test for specific fields being dirty,
I have an abstract class:
public abstract class SmartWrap : ISmartWrap
{
private int orig_hashcode { get; set; }
private bool _isInterimDirty;
public bool IsDirty
{
get { return !(this.orig_hashcode == this.GetClassHashCode()); }
set
{
if (value)
this.orig_hashcode = this.orig_hashcode ^ 108.GetHashCode();
else
MakeClean();
}
}
public void MakeClean()
{
this.orig_hashcode = GetClassHashCode();
this._isInterimDirty = false;
}
// must be overridden to return combined hashcodes of fields testing for
// example Field1.GetHashCode() ^ Field2.GetHashCode()
protected abstract int GetClassHashCode();
public bool IsInterimDirty
{
get { return _isInterimDirty; }
}
public void SetIterimDirtyState()
{
_isInterimDirty = this.IsDirty;
}
public void MakeCleanIfInterimClean()
{
if (!IsInterimDirty)
MakeClean();
}
/// <summary>
/// Must be overridden with whatever valid tests are needed to make sure required field values are present.
/// </summary>
public abstract bool IsValid { get; }
}
}
As well as an interface
public interface ISmartWrap
{
bool IsDirty { get; set; }
void MakeClean();
bool IsInterimDirty { get; }
void SetIterimDirtyState();
void MakeCleanIfInterimClean();
}
This allows me to do partial saves, and preserve the IsDirty state if there is other details to save. Not perfect, but covers a lot of ground.
Example of usage with interim IsDirty State (Error wrapping and validation removed for clarity):
area.SetIterimDirtyState();
if (!UpdateClaimAndStatus(area))
return false;
area.MakeCleanIfInterimClean();
return true;
This is good for most scenarios, however for some classes i want to test for each field with a backing field of original data, and either return a list of changes or at least an enum of fields changed.
With an enum of fields changed i can then push that up through a message chain for selective update of fields in remote caches.
You could also think about boxing your variables, which comes at a performance cost, but also has its merits. It is pretty consise and you cannot accidentally change a value without setting your dirty status.
public class Variable<T>
{
private T _value;
private readonly Action<T> _onValueChangedCallback;
public Variable(Action<T> onValueChangedCallback, T value = default)
{
_value = value;
_onValueChangedCallback = onValueChangedCallback;
}
public void SetValue(T value)
{
if (!EqualityComparer<T>.Default.Equals(_value, value))
{
_value = value;
_onValueChangedCallback?.Invoke(value);
}
}
public T GetValue()
{
return _value;
}
public static implicit operator T(Variable<T> variable)
{
return variable.GetValue();
}
}
and then hook in a callback that marks your class as dirty.
public class Example_Class
{
private StatusEnum _Status = StatusEnum.New;
private Variable<long> _ID;
private Variable<string> _Name;
public StatusEnum Status
{
get { return _Status; }
set { _Status = value; }
}
public long ID => _ID;
public string Name => _Name;
public Example_Class()
{
_ID = new Variable<long>(l => Status = StatusEnum.Dirty);
_Name = new Variable<string>(s => Status = StatusEnum.Dirty);
}
}
Another method is to override the GetHashCode() method to somthing like this:
public override int GetHashCode() // or call it GetChangeHash or somthing if you dont want to override the GetHashCode function...
{
var sb = new System.Text.StringBuilder();
sb.Append(_dateOfBirth);
sb.Append(_marital);
sb.Append(_gender);
sb.Append(_notes);
sb.Append(_firstName);
sb.Append(_lastName);
return sb.ToString.GetHashCode();
}
Once loaded from the database, get the hash code of the object. Then just before you save check if the current hash code is equal to the previous hash code. if they are the same, don't save.
Edit:
As people have pointed out this causes the hash code to change - as i use Guids to identify my objects, i don't mind if the hashcode changes.
Edit2:
Since people are adverse to changing the hash code, instead of overriding the GetHashCode method, just call the method something else. The point is detecting a change not whether i use guids or hashcodes for object identification.