When is my C# property initialized? - c#

I'm a little confused about when exactly my Property is being initialized.
Suppose I have a property declared like this:
private Dictionary<string, Dictionary<string,string>> MessageLookup
{
get
{
return messages ?? doSomething();
}
}
The doSomething method populates the messages Dictionary and returns it.
My question is, when is this code run?
If I place a breakpoint into the doSomething code, it isn't hit, but the MessageLookup property is holding data (this is the only place it is initialized) when I view it in the debugger.
Is this code run at construction? does the debugger run it automatically when I hover over the variable name? If so, why isn't the breakpoint hit?

That code is run whenever anyone refers to the property, and not before.
If you use the debugger, you'll see it because the debugger tries to fetch property values automatically (so you can see the state of the object). I don't know whether the debugger ignores breakpoints while it's evaluating properties for itself - that would explain everything.
Try running your code not in a debugger, and make some code access your property:
var lookup = someObject.MessageLookup;
Make doSomething() dump a stack trace and you'll see what's going on.

It is run when your property is first evaluated. No background stuff going on.
I'm guessing you're not seeing this because you use Quickwatch to inspect your object. At that point it will get executed and your breakpoint will be skipped.

Property getters (and ToString() for that matter) are assumed to be pure, which basically means evaluating it has no side effects. You should rewrite the code to adhere to that assumption or you'll face nasty consequences. If must use lazy initialization at access time, use a GetMessageLookup() method instead.

Related

Does .NET CLR precalculate property values?

When I run C# code through the debugger, calculated property values always seem to be displayed when I inspect an object.
e.g.
I am wondering if this is only done in the debugger, or if .Net does this as an optimization and can detect when a property changes, because this would affect how I use and access such properties to avoid performing calculations multiple times.
I have never seen it not display the value in the debugger even with complex calculations.
The debugger is calling .ToString() on all the objects. Whatever code is implemented in that function for each object is executed, which may mean it looks like the debugger has some insider information, but it really doesn't.
You can confirm this by writing your own .ToString() function in a class and see what happens.
The value of your auto-property will be computed when something calls the getter. It just so happens that the debugger will call the getter to display the property value, which is what you are seeing.
I am wondering if ... .Net does this as an optimization and can detect when a property changes, because this would affect how I use and access such properties to avoid performing calculations multiple times.
There is no built-in property value caching. Since the get method will be executed whenever you "get" the property, the calculation will be executed every time. If you want to cache the value, you could add a backing field, but you'll need to detect if/when to recalculate the value. One way would be to add logic to the setter of Value to either invalidate the cache or recompute the dependent properties (like MessageCode) at that time.
My opinion is that such a simple calculation is safe to run millions of times rather then adding the overhead of change detection.

C# control not going into the method

I am trying to call a private method from another private method like this
UploadFeeScheduleToDb(147, finalPath);
Method definition:
void UploadFeeScheduleToDb(int UploadID, string UploadFilePath)
{
DataSet CSVData = CSVToDataSet(UploadFilePath);
}
The problem is that the C# control is coming to the method call but not going inside it. I added breakpoints like this:
As you can see, the control is reaching the breakpoint but it's not reaching to the second breakpoint inside that method. It's just skipping to lblMsg.Text... statement without any exceptions in output window.
I tried cleaning solution and rebuilding. Also, I passed constants or magic values to the method. But no luck. I don't know what is happening?
As #Silvermind and #HenkHolterman said, the UploadFeeScheduleToDb method is not doing anything productive except assigning the value to its local variable, the C# compiler will ignore this method when Code Optimization feature is turned on. I think this is called Dead Code Optimization. Correct me if I am wrong.

How to use Environment.StackTrace in a conditional expression for a breakpoint in VS2015

I'm trying to debug an issue and it would really help if I could have a breakpoint in the getter of a property, but I don't need to break every time it is called from a CanExecute call for some button in my UI. My thought was that I could simply condition the breakpoint on the call stack not containing that string, but that seems to not work. The best way I can think of to explain this is with an image of the settings and the output from the breakpoint being hit.
As you can see in the image, the Conditional settings for the breakpoint is outputted directly when it is hit, and in the output you can see it being hit a few times .get(): true - correctly. However, when it took this screenshot it was being hit with the condition being false, as seen in the output. The program halted as the breakpoint was hit, incorrectly.
Am I doing something wrong - is this possible at all? Seems to me like a bug in VS2015, the output can evaluate the bool correctly, why can't the breakpoint condition do that?
Edit, to annotate the image in case it gets lost sometime.
I have a breakpoint in the get method of a property that simply returns the underlying field. The Breakpoint settings show that there is a Condition on the breakpoint that defines it such that it should only be hit when the expression !Environment.StackTrace.Contains("CanExecute") returns true, i.e. only break when the Stack Trace does not contain the "CanExecute" string.
The Action part of the breakpoint settings simply outputs the function name and the conditional expression using $FUNCTION: {!Environment.StackTrace.Contains("CanExecute")}. The action is set to not continue the execution.
I don't know exactly why the stack trace information is not available inside breakpoint conditions, but there's a crazy workaround/hack you can do as an extension to a technique outlined in another answer
It won't work as "cleanly" if you use auto properties as it requires two breakpoints, and auto properties only have a single point to bind a breakpoint to but I'd imagine this isn't as useful for an auto property anyway.
Place the first breakpoint on the opening curly brace of the get method (set your cursor on the curly brace and hit F9). Set an action on this breakpoint; set "Log a message to Output Window:" to
{System.AppDomain.CurrentDomain.SetData("break", !Environment.StackTrace.Contains("CanExecute"))}
and ensure Continue execution is checked.
Place the second breakpoint on the return statement (set your cursor on the curly brace and hit F9). Set the condition on this breakpoint:
(bool)System.AppDomain.CurrentDomain.GetData("break")
Depending on your formatting, this may require the aid of the "Breakpoints" window Debug > Windows > Breakpoints (Ctrl+Alt+B) in case the curly brace and the return statement are on the same line. You can edit the action/condition by right clicking on the breakpoint in the "Breakpoints" window and selecting Settings
The cast to bool is required as GetData() returns an object and the conditional breakpoint won't do the cast for you.
It isn't pretty, and it doesn't work very well in a multi-threaded environment due to the global state, but it can be useful in a pinch. If you need multiple "conditional breakpoints" in this manner, make sure to use a different key in SetData()/GetData().
However, if you can, it's usually (depending on your compile time and access to the code vs debug symbols) quicker/easier to just temporarily edit the code to put the condition you want to break on.
e.g.
public Foo Selected
{
get
{
if (!Environment.StackTrace.Contains("CanExecute"))
System.Diagnostics.Debugger.Break();
return _selected;
}
}
*In the case of auto properties, you can use a pair of action breakpoints in the CanExecute() method before the property access to set "break" to true, and then after to set "break" to false, and for the condition use
(bool?)System.AppDomain.CurrentDomain.GetData("break") != false
to ensure it still breaks before and after CanExecute() but not during.

Bug tracking by debugging when a variable changes

I'm a beginner at programming and having a very difficult time tracking down bugs, this happens because usually place a variable on watch, and keep pressing f5 until i notice some change. I'm on visual c# 2010 and i have 18000 lines of code, so only with some luck i do get to catch the problem.
Is there a way to instantly go to the line of code when a variable changes?
You can change your variable into a property and put a breakpoint on the setter. Then you can have a single breakpoint that will hit each time a piece of code changes its value.
So if you have:
int myVariable;
Change it to:
int myVariable {
get;
set; // <-- Put your breakpoint here
}
This sounds like a design problem. Ideally, you should limit the places where a variable changes to very specific locations in your code. This is one reason to avoid global variables and static variables unless you have very good reasons to use them. Even then, you should define accessor methods as an interface for these variables rather than changing them directly.
As you debug your code, I suggest you look for ways you can improve it so that debugging isn't so difficult in the future.
You can have conditional breakpoints, that will only be hit when a condition is met.
So, assuming you have an index variable indx, you can put a conditinal break point, saying only stop when value = 7, and then it'll stop there when you're condition changes...
Have a look at this msdn page
and at this youtube tutorial.
In native code, your best bet would be to setup a data breakpoint. A data breakpoint fires when the data changes, irrespective of where the change comes from.
You can't do this for .NET however. You can't ask the debugger to break when the value of a variable changes. But, not all hope is lost. Do a "Find Usages" or "Find References" on the variable in question to find all places in your code that make use of the variable. Then set a breakpoint at each of those locations to see when the value of the variable changes.

break whenever a file (or class) is entered

In Visual Studio, is there any way to make the debugger break whenever a certain file (or class) is entered? Please don't answer "just set a breakpoint at the beginning of every method" :)
I am using C#.
Macros can be your friend. Here is a macro that will add a breakpoint to every method in the current class (put the cursor somewhere in the class before running it).
Public Module ClassBreak
Public Sub BreakOnAnyMember()
Dim debugger As EnvDTE.Debugger = DTE.Debugger
Dim sel As EnvDTE.TextSelection = DTE.ActiveDocument.Selection
Dim editPoint As EnvDTE.EditPoint = sel.ActivePoint.CreateEditPoint()
Dim classElem As EnvDTE.CodeElement = editPoint.CodeElement(vsCMElement.vsCMElementClass)
If Not classElem Is Nothing Then
For Each member As EnvDTE.CodeElement In classElem.Children
If member.Kind = vsCMElement.vsCMElementFunction Then
debugger.Breakpoints.Add(member.FullName)
End If
Next
End If
End Sub
End Module
Edit: Updated to add breakpoint by function name, rather than file/line number. It 'feels' better and will be easier to recognise in the breakpoints window.
You could start by introducing some sort of Aspect-Oriented Programming - see for instance
this explanation - and then put a breakpoint in the single OnEnter method.
Depending on which AOP framework you choose, it'd require a little decoration in your code and introduce a little overhead (that you can remove later) but at least you won't need to set breakpoints everywhere. In some frameworks you might even be able to introduce it with no code change at all, just an XML file on the side?
Maybe you could use an AOP framework such as PostSharp to break into the debugger whenever a method is entered. Have a look at the very short tutorial on this page for an example, how you can log/trace whenever a method is entered.
Instead of logging, in your case you could put the Debugger.Break() statement into the OnEntry-handler. Although, the debugger would not stop in your methods, but in the OnEntry-handler (so I'm not sure if this really helps).
Here's a very basic sample:
The aspect class defines an OnEntry handler, which calls Debugger.Break():
[Serializable]
public sealed class DebugBreakAttribute : PostSharp.Laos.OnMethodBoundaryAspect
{
public DebugBreakAttribute() {}
public DebugBreakAttribute(string category) {}
public string Category { get { return "DebugBreak"; } }
public override void OnEntry(PostSharp.Laos.MethodExecutionEventArgs eventArgs)
{
base.OnEntry(eventArgs);
// debugger will break here. Press F10 to continue to the "real" method
System.Diagnostics.Debugger.Break();
}
}
I can then apply this aspect to my class, where I want the debugger to break whenever a method is called:
[DebugBreak("DebugBreak")]
public class MyClass
{
public MyClass()
{
// ...
}
public void Test()
{
// ...
}
}
Now if I build and run the application, the debugger will stop in the OnEntry() handler whenever one of the methods of MyClass is called. All I have to do then, is to press F10, and I'm in the method of MyClass.
Well, as everyone is saying, it involves setting a breakpoint at the beginning of every method. But you're not seeing the bigger picture.
For this to work at all, a breakpoint has to be set at the beginning of every method. Whether you do it manually, or the debugger does it automatically, those breakpoints must be set for this to work.
So, the question really becomes, "If there enough of a need for this functionality, that it is worth building into the debugger an automatic means of setting all those breakpoints?". And the answer is, "Not Really".
This feature is implemented in VS for native C++. crtl-B and specify the 'function' as "Classname::*", this sets a breakpoint at the beginning of every method on the class. The breakpoints set are grouped together in the breakpoints window (ctrl-alt-B) so they can be enabled, disabled, and removed as a group.
Sadly the macro is likely the best bet for managed code.
This works fine in WinDbg:
bm exename!CSomeClass::*
(Just to clarify, the above line sets a breakpoint on all functions in the class, just like the OP is asking for, without resorting to CRT hacking or macro silliness)
You could write a Visual Studio macro that obtained a list of all of the class methods (say, by reading the .map file produced alongside the executable and searching it for the proper symbol names (and then demangling those names)), and then used Breakpoints.add() to programmatically add breakpoints to those functions.
System.Diagnostics.Debugger.Break();
(at the beginning of every method)
No. Or rather, yes, but it involves setting a breakpoint at the beginning of every method.
Use Debugger.Break(); (from the System.Diagnostics namespace)
Put it at the top of each function you wish to have "broken"
void MyFunction()
{
Debugger.Break();
Console.WriteLine("More stuff...");
}
Isn't the simplest method to get closest to this to simply set a break point in the constructor (assuming you have only one - or each of them in the case of multiple constructors) ?
This will break into debugging when the class is first instantiated in the case of a non-static constructor, and in the case of a static constructor/class, you'll break into debugging as soon as Visual Studio decides to initialize your class.
This certainly prevents you from having to set a breakpoint in every method within the class.
Of course, you won't continue to break into debugging on subsequent re-entry to the class's code (assuming you're using the same instantiated object the next time around), however, if you re-instantiate a new object each time from within the calling code, you could simulate this.
However, in conventional terms, there's no simple way to set a single break point in one place (for example) and have it break into debugging every time a class's code (from whichever method) is entered (as far as I know).
Assuming that you're only interested in public methods i.e. when the class methods are called "from outside", I will plug Design by Contract once more.
You can get into the habit of writing your public functions like this:
public int Whatever(int blah, bool duh)
{
// INVARIANT (i)
// PRECONDITION CHECK (ii)
// BODY (iii)
// POSTCONDITION CHECK (iv)
// INVARIANT (v)
}
Then you can use the Invariant() function that you will call in (i) and set a breakpoint in it. Then inspect the call stack to know where you're coming from. Of course you will call it in (v), too; if you're really interested in only entry points, you could use a helper function to call Invariant from (i) and another one from (v).
Of course this is extra code but
It's useful code anyway, and the structure is boilerplate if you use Design by Contract.
Sometimes you want breakpoints to investigate some incorrect behaviour eg invalid object state, in that case invariants might be priceless.
For an object which is always valid, the Invariant() function just has a body that returns true. You can still put a breakpoint there.
It's just an idea, it admittedly has a footstep, so just consider it and use it if you like it.
Joel, the answer seems to be "no". There isn't a way without a breakpoint at every method.
To remove the breakpoints set by the accepted answer add another macro with the following code
Public Sub RemoveBreakOnAnyMember()
Dim debugger As EnvDTE.Debugger = DTE.Debugger
Dim bps As Breakpoints
bps = debugger.Breakpoints
If (bps.Count > 0) Then
Dim bp As Breakpoint
For Each bp In bps
Dim split As String() = bp.File.Split(New [Char]() {"\"c})
If (split.Length > 0) Then
Dim strName = split(split.Length - 1)
If (strName.Equals(DTE.ActiveDocument.Name)) Then
bp.Delete()
End If
End If
Next
End If
End Sub
Not that I'm aware of. The best you can do is to put a breakpoint in every method in the file or class. What are you trying to do? Are you trying to figure out what method is causing something to change? If so, perhaps a data breakpoint will be more appropriate.
You could write a wrapper method through which you make EVERY call in your app. Then you set a breakpoint in that single method. But... you'd be crazy to do such a thing.
You could put a memory break point on this, and set it to on read. I think there should be a read most of the time you call a member function. I'm not sure about static functions.
you can use the following macro:
#ifdef _DEBUG
#define DEBUG_METHOD(x) x DebugBreak();
#else
#define DEBUG_METHOD(x) x
#endif
#include <windows.h>
DEBUG_METHOD(int func(int arg) {)
return 0;
}
on function enter it will break into the debugger
IF this is C++ you are talking about, then you could probably get away with, (a hell of a lot of work) setting a break point in the preamble code in the CRT, or writing code that modifies the preamble code to stick INT 3's in there only for functions generated from the class in question... This, BTW, CAN be done at runtime... You'd have to have the PE file that's generated modify itself, possibly before relocation, to stick all the break's in there...
My only other suggestion would be to write a Macro that uses the predefined macro __FUNCTION__, in which you look for any function that's part of the class in question, and if necessary, stick a
__asm { int 3 }
in your macro to make VS break... This will prevent you from having to set break points at the start of every function, but you'd still have to stick a macro call, which is a lot better, if you ask me. I think I read somewhere on how you can define, or redefine the preamble code that's called per function.. I'll see what I can find.
I would think I similar hack could be used to detect which FILE you enter, but you STILL have to place YOUR function macro's all over your code, or it will never get called, and, well, that's pretty much what you didn't want to do.
If you are willing to use a macro then the accepted answer from this question
Should be trivially convertible to you needs by making the search function searching for methods, properties and constructors (as desired), there is also quite possibly a way to get the same information from the the ide/symbols which will be more stable (though perhaps a little more complex).
You can use Debugger.Launch() and Debugger.Break() in the assembly System.Diagnostics
Files have no existence at runtime (consider that partial classes are no different -- in terms of code -- from putting everything in a single file). Therefore a macro approach (or code in every method) is required.
To do the same with a type (which does exist at runtime) may be able to be done, but likely to be highly intrusive, creating more potential for heisenbugs. The "easiest" route to this is likely to be making use of .NET remoting's proxy infrastructure (see MOQ's implementation for an example of using transparent proxy).
Summary: use a macro, or select all followed by set breakpoint (ctrl-A, F9).
Mad method using reflection. See the documentation for MethodRental.SwapMethodBody for details. In pseudocode:
void SetBreakpointsForAllMethodsAndConstructorsInClass (string classname)
{
find type information for class classname
for each constructor and method
get MSIL bytes
prepend call to System.Diagnostics.Debugger.Break to MSIL bytes
fix up MSIL code (I'm not familiar with the MSIL spec. Generally, absolute jump targets need fixing up)
call SwapMethodBody with new MSIL
}
You can then pass in classname as a runtime argument (via the command line if you want) to set breakpoints on all methods and constructors of the given class.

Categories