Comments by Patrick Skelton (Top 200 by date)

Thank you, Richard. Wow! That is one lump of code. I'll have a play with it.

I'd still love to know how Windows knows the file is open on another machine. Or maybe it doesn't. Maybe even deep in the OS the rename simply fails with an exception. I should have asked before I sold the idea to the client. :(

Thank you for the suggestion. It does look interesting. I'd have to do a bit of logic to code around the fact that I don't want one application to be able to send a message that is some-time-later processed by the second application (i.e. I'm only interested in the case when both applications are currently running simultaneously) but I can't see that being too difficult. Definitely an option.

Thank you!

Thank you for that. Definitely a step in the right direction. Is it not possible to go one step further and just use nested curly braces that the compiler interprets as an implicit call to Add()? Is there any way to do that? (I'm just curious; your code is clear and concise and will do just fine.)

Jeez... that is so simple when you see how it should be done! I was *so* close to that yesterday afternoon but was getting a crash that lead me down a garden path. Thank you for all your help, Richard. For my personal edification, can you suggest a place I could put a call to Debug.WriteLine(TaskStatusString()) that would show any of them as Running?

Ah, I see! I didn't spot that I was implicitly creating an async void. One step closer to the penny dropping... I hope...

PS - Thanks for the tip about the analyser. I don't like what it tells me about my code but that's undoubtedly a good thing.

That is a good explanation, Richard. Thank you. I feel like I understand all that you said with the exception that I'm still not sure why the await Tasks.WhenAll() back in the main function doesn't 'seem' to do what I expect it to do. If you forgive the vague terminology, doesn't that, in effect, make the top-level thread aware of the sub-threads?

It's strange but when I use 'old fashioned' ways of threading (I worked a while using C++ and RTOS), I had no problems conceptualising all this. I just seem to have a mental block on how await actually works, no matter how much I read about it. I have updated my question to show code that behaves as I expected it to behave but am still struggling to see precisely how it differs fundamentally from my original example. The penny might drop one day. Thank you for the explanation.

Hi, Richard. I'm not sure I have phrased my question very well. What I don't understand is how any code can be running after the Task.WhenAll(). It seems to be doing nothing.

That is an excellent solution, sir! Thank you! (This is the first time I've passed a cancellation token into a Task.Delay(). I was surprised to get a TimeoutException. I'm sure there are practical reasons for this but semantically it seems a bit odd to me that I tell it to cancel and it responds with an error.)

Sorry, forgot that little detail. It's a Xamarin.Forms application. The platform-independent chunk of the code is .NETStandard2.0 but the target framework in the iOS project - where this audio player is implemented - is blank. I'm guessing the same?

Perhaps if you asked a specific question, instead of ranting, I might be able to improve the question to your liking. I'd hazard a guess no one else is having much problem understanding how my code looks. It's textbook MVVM. If there is no such thing, take it up with authors of websites such as this one: https://bookauthority.org/books/best-mvvm-books.

I didn't add this because it couldn't be simpler but I have now improved the question, hopefully to clarify the points you raise.

I've added it, plus a long-shot explanation of what might be happening.

You mean for the whole screen?

Thank you for the heads-up! - Patrick

Thank you for the answer. I'm using ASP.NET, not Core, so I'm not sure it will be the answer, but I will take a look next time I get time to work on the project.

Even my question was confused. I was having a good day. :/

Thanks for the reply, Richard. I believe I am using the correct function, since I am wanting only files in one specific directory. By adding some logging, I have identified what I believe is a simple failure of the UI layer to be refreshed and therefore the containing path passed into the DirectoryInfo constructor could indeed have vanished from existence. I am now wearing a hat with a 'D' on the front. At least it's Friday.

That does sound like a candidate. However, it does say 'The connection pooler removes a connection from the pool after it has been idle for approximately 4-8 minutes'. I do find it odd that any should be open on a weekend.

Point taken about the using statement though. I'd feel happier if I refactored the code to use that approach.

Thanks for the information, Richard.

Thanks for the help, Richard. Sorry to have wasted your time on what turned out to be a pretty dumb mistake not to have spotted. I've posted the solution I found.

Mmm... code is not particularly complex but is pretty long. I'll try commenting bits out, to see if I can spot what might be causing the problem.

You are right about using 'this' instead of 'Application'. I've modified the code but there is no change in behaviour. StartupUri is null when I call Shutdown() but does get set later (in code that shouldn't be hit but is unless I do Environment.Exit()). Not sure if it means anything, but a breakpoint just after a call to Shutdown() shows this.Dispatcher.HasShutdownStarted to be false.

Hi, Richard! No, it is being called from Application.Startup(), on the main thread.

I'm just getting the general 500 error 'Internal Server Error'. Thing is the files are not missing. I've checked and treble-checked that they are in the same place as the HTML reference, and since these are all relative to site-root, I really don't see how the files can cause a problem.

Ideally, I would be able to use this in both an ASP.NET project and a WPF desktop application. I face the same problem in both types of project.

I have added a note to the original question, as, looking at replies, I think I may not have been sufficiently clear that any way to solve this problem does not need to be a part of the build process.

I have added a note to the original question, as, looking at replies, I think I may not have been sufficiently clear that any way to solve this problem does not need to be a part of the build process.

Thanks for the reply. I have done this before and it works well when the commands are tied to the build configurations. I couldn't think of an easy way to implement these in such a way that I can copy the set of files I want irrespective of whether the build is set to Debug or Release. Since these commands run every build, ideally I'd only like to do the copy when the source changes, though I guess copying a handful of files would be fast enough to live without this last bit.

An example of not being able to Google something 'cos I honestly didn't know it was referred to as 'auto-versioning', nor did I realise the bundling process solves this problem. So I've learnt three things! Thank you!

Sorry - MVC.

Thank you for the information. Do you know how to do it in ASP.NET, not .Core?

I'd have thought IIS might have had a command-line option or config file option that VS might be able to make use of, though I guess forcing it to always reload resources might make the already-slow deployment to local IIS even slower.

Thanks for the reply. I'm not sure what you mean when you say to make sure BundleConfig.cs *has* the minimised versions of CSS. I'm still new to this whole MVC thing (last did Web.Forms) so things like bundling and routing are all still new to me.

Thanks for the information. The project is the out-of-the-box Microsoft site, so don’t really know where to start with fixing that.

Hope no one minds but I've accepted this as solution because the provided link points to extremely useful information, and therefore I feel that anyone who might look at my question would best be helped by reading these excellent illustrations of various patterns implemented in C#.

That is a very clear way of putting it! Is it still a problem if A and B are themselves subclasses of a common ancestor Z and all I need to do is store a single reference somewhere to X<Z>, while any code specific to X<A> and X<B> is contained privately inside those classes?

Thanks for the reply. I really do seem to have done a poor job of asking this question. I put the concrete functions in just to 'test' that my two subclasses are capable of being different; in reality, the only place the code relies on its specific type is inside that specific type. The caller generally doesn't need to know the specific type. Mainly this whole thing was an exercise in avoiding casting. That said, the fact that all of the answers don't seem to line-up with my intent implies I have indeed been over-complicating things. I'll work out something a bit cleaner on Monday and post it if it looks remotely relevant to my original question. Not sure what to do about accepting solutions here. All supply useful information but it my question is poor so none of the solutions seems to be exactly what I am looking for. Any advice on protocol here anyone?

Mmm... I think your suggestion of this as a solution maybe indicates I haven't worded the title of the question very well. I don't think what I'm trying to do is that complicated. I'm not wanting to be able to swap out factories for example. I think all I am wanting is a factory that can create an instance of a generic class.

Sorry - I've (hopefully) improved the question.

@F-ES Sitecore I clearly haven't got a solid understanding of this. I can sort-of see why the following code won't compile, but part of me feels like I'm missing a key brick here. public static class GraphicsFactory
{
public static GraphicsBase<logicbase> Create( string whichOne )
{
return new GraphicsSubclassB( new LogicSubclassB() );
}
}

Error: Can't implicitly convert GraphicsSubclassB to GraphicsBase<logicbase>.

Jeez! Easy when you know how! Thank you very much.

Doh! Sorry, I forgot to mention that I have added to the 'What I have tried' part of the question my latest code. I did it there so I'd get proper formatting.

Mmm... I've bumped into problems with this. I am clearly seriously misunderstanding something. I'm hoping you (or someone else) can tell me what I am doing wrong.

Ooo... I had forgotten reading about that many moons ago. I like it. (Only thing my OCD is struggling with is the fact that these are the first FnAsync() methods I've come across that are not awaitable, but I guess that would kind-of lead me full-circle.) Thank you!

Thank you for the reply, Richard. I have improved my question to be a bit closer to what I am actually trying to do. I tried to follow your code but still get the error mentioned in the question. (You are right that, in the example I give, the FunctionRunner adds nothing. In the real code, it receives three delegates and decides which one to run depending upon certain conditions. I chopped it down to give the shortest possible code example.)

I have just discovered this rather obvious fact for myself. I discovered the initialisation for the static variable was being hit more than once, which is exactly what you would expect given what you have pointed out.

The clue should have been in the rather horrible code I was using to call the static functions: SessionStateBase<ISession>.InitiateTimeSeek();

I have moved the variable to a class that is not generic and is actually a much more natural home. Now everything works fine.

That will teach me to make comments like 'this class is an abstract generic class, though I struggle to see why this should be relevant'!

Thank you for the reply, Richard.

So sorry! I did indeed cut the code clumsily. I have corrected it and added a couple of lines to make it more complete. I've also removed the 'async' bits because they were just confusing things. I should also point out that this class is an abstract generic class, though I struggle to see why this should be relevant (a view I may end up retracting, when I have exhausted all other possibilities).

Mmm... that's strange. It does indeed work fine now. I guess I must have had a syntax error or something. I'll leave the question up because it may provide a nice quick answer for anyone else who may come along. Thank you.

That is an excellent way of making it clear! I don't know why but seeing it written like that has helped a lot! Thank you. I'm still uncertain if what I am doing is inherently bad, though. I'm always wary of effectively forcing the compiler to do something it doesn't want to.

Thank you for the reply. I'm still struggling to get this. I've read your reply several times and I think I see what you mean, but when you say '...you are allowed to implicitly cast to a type that is derived from it' did you mean to say '...you are allowed to implicitly cast to a type that it is derived from'?

Thanks for the reply. I'll give it a try when I get back onto that part of the project.

Thanks for the reply. You are, of course, right. I realised soon after posting that what I was attempting just didn't make sense. I held back on a reply until I had got something that works, which I have now posted.

I'll get my coat. :-|

(Thank you!)

I think this 'the code that calls the function defines the type' is the key piece of information I was failing to understand.

Thank you both!

Thank you for the answer. That is an interesting point I had overlooked but it doesn't seem to fix the problem for me. The text is always left-aligned when I apply Stretch to the TextBlock.

Maybe I should post my entire XAML for this part of the window. Maybe something I'm doing higher up the visual tree is breaking things.

Modification to my previous comment to edit out brain-in-neutral bits: In the Live Visual Tree, I've been reading 'AlternationCount' as 'AlternationIndex'. is it possible to see the latter in the Live Visual Tree?

Thank you for the reply. I will mark it as a solution but am delaying while I experiment because so far I haven't got it working. I'm struggling to see the difference between what I have and the example in the second link you gave, yet, when I look in the Live Visual Tree, I always get an AlternationIndex of 2, regardless of any changes I make to the AlternationCount in either the parent TreeView or its root HierarchicalDataTemplate. Any ideas what might be going wrong? (I'm puzzled that it is always 2; it is as if it is marking the items according to depth in the visual tree.)

I chose the screenshot simply to illustrate roughly what I want to achieve. But there is different behaviour for each row. The coloured titles cannot, for example, be edited in the same way as the individual costed items. You could, if you wanted, implement my example image as a two-level tree. I'd argue that these items are, therefore, organised into sections.

Not fond of turning off warnings (since in my case, the warning is usually right and I've got my brain in neutral), but thanks for the reply.

Wasn't aware of ValueTask at all, so I've learnt two things! Thank you, Richard.

Doesn't that mean you are forever stuck with the compiler warning telling you that the task will complete synchronously?

The solution I found below seems to work without this, but I have to admit I wasn't aware it was one-way by default, so that would probably have crawled out and bit me at some point. Thanks for the suggestion.

I'm not sure what you mean by the first part of your answer, the bit about 'knows about', but the 'Display Items: Numbers' did the trick. I must have missed it. I do think this list is getting stupidly long and needs a completely new interface. I also need to learn to read, maybe. :-|

Anyway, thank you! Problem solved and my eyes already feel better!

How absolutely bizarre! It works now for me, too. And I cannot for the life of me recreate the broken version.

ToString() calls removed. Thanks for the tip!

Sorry, I did of course mean LINQ lamda functions. Thank you for the correction and the answer.

I see. At least I now understand why it doesn't work. Still, it seems like a maintenance problem to me. My understanding of CSS variables is that avoidance of repeated constants is precisely the problem they were intended to solve. Guess I can get round it when I convert my current static HTML/CSS site to MVC/C# but I would like to see one of the proposed solutions in the GitHub thread on SO implemented.

Thank you for the answer, Richard.

Thank you for the reply. I haven't checked network activity, as you suggest, for the simple bizarre reason that it is suddenly working. I moved on to working on something else, so I guess I must have fixed it by accident, which more than likely means I had some CSS rules in the wrong order. I'm also guilty of not committing to source control frequently enough, so I'm not even sure how I fixed it. (Is there a delete button for poor questions?) :~

I'm sorry I didn't get it first time around. When you understand something as well as you obviously do, it can be easy to forget what it is like to be coming at this as a newbie. Trust me, I'm no strange to (mostly non-web) development but this wep api landscape Microsoft has created is truly confusing from my newbie viewpoint.

Thank you for your longer answer. It has made things *much* clearer. I think I can now make a sensible choice about the path I should follow.

- Patrick

@RickZeeland @SuperLloyd

I definitely could have been more specific in my question. I get that MVC is just a pattern. I've done a bit more reading and I think I've arrived at three questions I am still struggling to answer:

1. Can I use Razor syntax with standard ASP.NET?
2. By doing so, can I use the GET/POST web mechanism directly (perhaps by building a RESTful web API, which seems intuitively like a nice way to work to me) and remove all reliance on the WebForms controls and view-state?
3. How does Entity Framework fit in? Is this just a good way to implement the model part of the architecture?

I think I mostly understand all the different bits I am asking about here. I'm just struggling to see which ones I need and which I don't and how the pieces fit together.

- Patrick

That is indeed an excellent explanation, making good use of direct quotes from the language spec. Thank you!

Your code fixes my problem but I have marked a different reply as solution because it offers an explanation, which should help anyone else who stumbles across this thread.

I will in fact probably use the explicit cast you mention above, just to tell anyone else reading my code that I have not made a mistake.

An excellent explanation, sir! Thank you!

Sorry, I'm not sure I understand the question. When I use 0x00 in a ternary operator, the compiler does indeed tell me I cannot convert an int to a byte, which is an unambiguous, concise statement of the problem. I have attempted to clarify my question.

Of course, you are correct. I guess my question should really have been why does the plain assignment (_bytes[ 3 ] = 0x00;) compile okay without a cast but requires one when the expression is part of a ternary operator?

Thanks for the answer. Unfortunately that will not work for me as I am not using GitHub. I am using a local Git installation and pushing to a remote BitBucket account.

Sorry, I should have been clearer. I will modify my original question.

Yes, I saw that. It does appear to be the closest to what I had in mind. I'll give it a try when I get chance.

Thanks for the suggestion.

:)
Is it just me who has a problem with PowerShell? The syntax seems highly convoluted to me. The verb-noun bit is a good idea but it seems to go quickly downhill from there. I actually prefer the old DOS shell commands and I NEVER thought I'd say that!

I think I googled 'what is the advantage of c# is operator'.

So, I think the short answer is, yes, I am being paranoid! :)

Thanks for the explanation. Isn't this hopelessly inefficient? If I don't cache the result of the IEnumerable in a List somewhere, what will happen if I data bind a UI element to the enumeration? Won't the IEnumerable (which involves reflection) be executed every time the UI element is displayed? Maybe I am being paranoid. I have worked in video games for a lot of years and things like reflection being called repeatedly make me nervous.

So don't I need to put it back to get a list of all the enumeration values or is there a better way?

Mmm... I can't get it to work without explicitly adding a public parameter-less constructor. Doesn't the declaration of the protected constructor with parameters remove the default one?

Thanks for the explanation. You did state very clearly something I hadn't quite realised. I chose a different answer as solution because your reply didn't mention the public parameterless constructor that appears to be required. Thank you.

That is indeed all that was needed. the compiler told me as such but I didn't add one because I foolishly thought, "If it needed a public parameterless constructor, surely Microsoft's example code would have one."

Now all of the following compile fine:

var AllFruits1 = Fruit.GetAll<fruit>();
IEnumerable<fruit> AllFruits2 = Fruit.GetAll<fruit>();
List<fruit> AllFruits3 = Fruit.GetAll<fruit>().ToList();

Thank you!

Cherry picking your replies, this seems to do what I want: Get-ChildItem -Path C:\Temp -Recurse | Select-Object FullName, @{n='l';e={$_.FullName.Length}} | Sort-Object -Property l

Thanks, Richard. I've learnt something but it doesn't seem to quite work for me. I've shown a simplified version of your code below together with its output. As you can see, it seems to 'chop off' the final subdirectory when displaying the length.

PS C:\WINDOWS\system32> Get-ChildItem -Recurse -Directory C:\Temp\Debug | Select-Object FullName, @{n='l';e={$_.Parent.FullName.Length}} | Sort-Object -Property
l

FullName l
-------- -
C:\Temp\Debug\x64 13
C:\Temp\Debug\x86 13
C:\Temp\Debug\Resources 13
C:\Temp\Debug\WorkingFolders 13
C:\Temp\Debug\Resources\Graphics 23
C:\Temp\Debug\WorkingFolders\Temp 28
C:\Temp\Debug\WorkingFolders\Reports 28
C:\Temp\Debug\WorkingFolders\Logs 28

I ended up simply bypassing the Microsoft settings system. I know rolling your own solution is often considered bad-form but it so often seems to be the case that out-of-the-box solutions don't work the way you want and are not easy to modify. I even ended up writing a load of code to modify the functionality of the wonderful NLog recently because it simply would not do what I needed it to do.

Thanks for the replies, Richard. Looks like the VS team are missing a trick from Code team. Hope they do it soon because they do have a very nice difference viewer.

SourceTree undoubtedly does do some 'magic' behind the scenes, but from the user's point-of-view it allows individual lines of code to be committed, skipped or changed back to the previous version. It is so useful, I now struggle to remember how I ever got by without it. I hoped Visual Studio might allow something similar.

Sorry, I should have checked the documentation. I just thought I was dealing with a fundamental principle in the language. I am honestly surprised to see such a clear, to-the-point answer right there in the documentation.

I wanted to learn something and I have. The lesson is: don't have so little faith in the documentation.

Ooo! That fixed it! I've never had a huge amount of faith in repair options. Silly, I know, but historically I just seem to have had such a low success-rate with them, not just in VS but in many other applications. This time it worked. Sometimes the obvious suggestions are the best. Thank you!

I still don't have a .suo file though. Is that normal for VS2017?

Thanks for the answers.

WPF support is installed. I have just updated to the latest VS 15.5.6, but it hasn't fixed the problems.

Interestingly, the Solution doesn't have a .suo file, at least not anywhere I can find.

The nesting issue appears to be related to the way I fix my solution-level folders. Initially, .xaml.cs files are properly nested. Then I click 'Solutions and Folders' at the top of Solution Explorer, then click 'Show All Files', then click 'Show All Files' again. This gets my solution-level folders displaying correctly but un-nests the .xaml.cs files.

I take your point about removing stuff from the project file. I did take a backup first.

I suspect I will indeed end up recreating the whole solution manually when I get time. I will try to remember to revisit this question to report whether or not this fixes it.

Thanks for the suggestions. Unfortunately, neither is the answer. I've rebuilt and the project is set to Debug.

I've used Visual Studio since the late 90s and never seen anything like this.

Thanks for the answer.

It is not a static constructor. The debugger is attached because I can step normally except for the problem of F11 refusing to step-into the constructor.

As suggested by Original Griff, I created a new WPF project, added a button and then created a POCO class of my own in the click-handler. I can step into this constructor fine, so I'm guessing my problem has something to do with being in the main App() constructor. Though, as I said, I had no problem with this on VS2015. Odd.

Fair comment. I can't get on the target system right now. Next time I can, I will try to write a small chunk of code to illustrate the failure. I'll add it to my question in the 'What I have tried' section.

Thank you for the reply, Richard.

If 'C:\Users\<username>\My Documents' is a link to an existing physical folder 'C:\Users\<username>\Documents' then shouldn't my code be able to use the actual physical path (i.e. 'C:\Users\<username>\Documents') to access this folder? In other words, why is my code failing at all?

That is correct, yes. The actual folder is 'My Documents' whereas in my code I get back just 'Documents'. Oddly, if I ask for the user's profile directory and then manually append '\Documents' in my code, I still don't seem to be able to access anything in 'C:\Users\<username>\Documents'. I need to do more testing around this, though. This was a last minute result I got last night before I had to quickly log out of the target machine.

Sorry - I should have been more clear.

I haven't tried this (because I am temporarily sans development environment - different story), but won't this be more-or-less the same as doing my file work synchronously? By this I mean that the main window will not load until the file work (which could take a while) completes?

Got to admit that I need to refresh my memory about task continuation, so I may be completely misunderstanding here.

Yes, I get a string value returned with both versions of the function, regardless of whether I put the call to the function before or after the setting of the start-up URI.

You are correct - this is called via a C#.net .DLL API to SQLite.

I guess I'll have to live with the exception approach. I can reduce the chances of the exception occurring virtually to zero by doing a quick check in C# code before I attempt the insert. I realise this is not thread-safe, but, as you say, the worst that can happen is that the subsequent insert throws an exception, which my code then handles okay.

I am surprised that my seemingly simple requirement is not easier to implement though.

Thank you for the answer.

Currently, my code works as you describe. I rely on the uniqueness of the primary key (PK_id) and catch any exception. This is, however, a foreseeable error. It happens in normal operation. I'm not fond of exceptions for this situation. I was hoping to change the SQL script so that it always runs to completion but somehow returns a result. Perhaps the number of rows affected might be zero instead of one?

The Guid is difficult to explain. It is possible (though unlikely) for locks in this table to become orphaned, perhaps as a result of a power cut for example. The Guid simply adds an extra layer of protection, allowing the calling thread or application instance to know beyond any doubt whether or not it is the owner of any given lock record.

You are right, Jochen. If I take out the workaround I mentioned above and I make both parts of if{}else{} return the None effect, I still get the can-drop over the ListBox. So my event handler is being ignored. Something else is simply using the AllowDrop="True" property on the ListBox to set the can-drop symbol.

The drag-source is Explorer. The cursor does change back to Stop-symbol when leaving the ListBox. Thank you for the extra information.

A workaround is to bind the IsEnabled property of the ListBox to the boolean property ValidProjectsDirectory, which is set true only when the file pane is showing the contents of a directory.

Yes, Jochen, that is exactly what happens. I get the No-Drop symbol when I drag over any area of my desktop that won't accept the drop. When I move over my main application window, I still get the No-Drop. But as soon as I go over the ListBox, I get a Can-Drop symbol.

I am guessing that my else{} is doing what it is supposed to but maybe something higher or lower in the XAML visual tree is overriding the behaviour.

I'll keep poking around, to see if I can find anything.

Thank you for your answer.

In your Google search, I see the word 'manually'. I confess I didn't try this because this seemed like the exact opposite of what I was trying to do.

The search link seems like a ridiculous amount of code for something so simple. The second contained reference to the very promisingly named 'ValidatesOnTargetUpdated'. This seems to work but I get a null reference exception in the XAML design view when I add it to the validation rule. I tried putting it in the constructor for the ValidationRule, which again works but again I get the same error in the XAML designer.

Very good point. I have to say that I found the whole thing - including the documentation - to be less than intuitive.

This is weird. I did have something close to your code, though I have created a dedicated private class to hold the items instead of using tuples. I am obviously doing something different because I get a compiler error where I try to add the item to the dictionary saying that the derived class cannot be converted to the base class.

private class BusinessObjectCacheItem<t> where T : BusinessObjectBase
{
public BusinessObjectCacheItem( List<t> businessObjectList, DataCommunicator<t> dataCommunicator )
{
_businessObjectList = businessObjectList;
_dataCommunicator = dataCommunicator;
}

public List<t> BusinessObjectList { get { return _businessObjectList; } }
public DataCommunicator<t> DataCommunicator { get { return _dataCommunicator; } }

private List<t> _businessObjectList = null;
private DataCommunicator<t> _dataCommunicator = null;

}

private Dictionary<BusinessObjectName,BusinessObjectCacheItem<businessobjectbase>> _cacheItems = null;

private void InitialiseCacheItems()
{
_cacheItems = new Dictionary<BusinessObjectName,BusinessObjectCacheItem<businessobjectbase>>();
_cacheItems.Add( BusinessObjectName.Contact, new BusinessObjectCacheItem<businessobjectbase>( _contactMasterList, _contactDataReaderWriter ) );
_cacheItems.Add( BusinessObjectName.Contact, new BusinessObjectCacheItem<contact>( _contactMasterList, _contactDataReaderWriter ) );
// Neither of the last two lines will compile even though Contact is derived from BusinessObjectBase.
}

I did mention in my question that this might be the way to go. The code would be more readable.

Thanks for the reply.

Actually, I think this highlights where my thinking is muddy. I can't use this code because each tuple I add has a different type parameter. That said, all the types do share a common base class, but I still can't persuade the compiler to see my side of things.

Thank you, Richard! Your phrase 'It should be as simple as...' brought a smile to my face. Even after a year, some aspects of XAML still seem like a black art to me, and judging by the number of wild goose chases I follow on the internet while looking for solutions, it seems I am not alone. I still seem often to take an approach of randomly setting properties, just to see if I can get anything to move on the screen. Anyway, thanks again.

The expression above does indeed work as expected using the excellent Debuggex (thank you for the link). And, when I go back into code, I find that it does indeed now work as expected. Maybe I was somehow testing a different piece of code to the one I was editing. Doh! I take the earlier point by Tomas, that I should have included some test code, which would quickly have indicated that something other than the regular expression was to blame. Thank you all for the help.

Well spotted, Richard. Sorry, it should indeed be as you say. It does match in the real code. The real code has a property IsPlaceholder but I changed it for the example just to get rid of the need for a negation converter in the XAML. I will update the question now.

Arghh! How come I didn't know about that? (Please don't answer that.) Thank you so much. Screen now works perfectly!

For a while, I couldn't get this to work. Turned out to be my trying to use colons in the date format, which I did escape with a backslash as mentioned in the docs. Still didn't work. Anyway, switching to hyphens got it working.

I still don't really see why the archiving needs to be mutually exclusive with having any kind of a time-stamp in the main file, given how easy I found it to write C# code to do what I wanted it to do. But thank you very much for the help.

Thank you for the answer.

That does indeed only keep the required number of log files. One oddity is that, despite what the documentation says, archiveNumbering="Rolling" and archiveNumbering="Sequence" seem identical - i.e. the higher-numbered archived file contains later messages than the lower-numbered archived file. This seems like too big a bug to have gone unnoticed, so I guess I am doing something else wrong.

Thanks for those suggestions.

None of your suggestions proved fruitful when I investigated. In fact, the mystery only deepened. A few days ago, I had experimented with creating multiple settings files (in the Designer) to hold groups of settings. I then ditched this and yet these namespaces were still appearing in the App.exe.config file. I even used a brute-force find-in-files tool capable of displaying hex, and I could not find anywhere these namespaces were being generated.

I have now fixed this. I've posted what I did an an answer.

Thanks for your help.

Thanks for the reply.

I'm not clear on where you suggest putting a breakpoint. The code is auto-generated. I simply create the settings using the Settings Designer. The rest is Microsoft magic (or it would be magic if I understood half of what it is doing).

Thanks for the reply. I have done something similar - deleting the file entirely. When the application recreates it during the build process, the entries are once more duplicated.

That is what I was looking for, Richard! Somewhere, buried in my fuzzy memory, was something close to what you've posted above. I didn't go ahead because I had been confused by posts talking about exactly the problem you mention, namely the client being able to cast to Dictionary<K,V>.

Thank you very much!

The constructor was just an example. The need for this particular expression (which simply sets a value based on the run-time platform) is required in several places.

I guess a more general version of the problem is: can I write a lambda expression that compares two strings and returns a different value depending upon the result of the comparison?

You, sir, are a star! I just didn't think of that. Works a treat (and has saved me hours over the horrible alternatives I was looking at). Thank you!

I'd accept this as the answer, but I don't know how to do that with a comment.

I am accepting this as a solution, but the solution by Peter Vetger seems neater in my particular use-case.

Thank you.

Excellent. Actually, since all I am doing is a bit of debug logging inside the 'add' and 'remove', I don't even need the count-variable. Which is nice because it means I don't need a single #if DEBUG, and my final release code is almost the same as if I had used the simpler property-like event declaration syntax.

Thank you very much.

I am simply trying to log the count in the Debug build, to keep an eye on potential memory leaks. I will use a separate count variable as you suggest. It does seem a shame, though, to be introducing a variable and associated code that will only exist in the Debug build and will therefore have to be wrapped in #if DEBUG ... #endif.

Thank you all.

The 'this' keyword appears to refer to the enclosing class, so I get a message saying that the class does not contain a method called GetInvocationList().

I'm guessing maybe it isn't possible because what I am trying to do is get a reference to the class instance that the C# compiler generates automatically in response to the 'event' keyword.

It won't compile. It says something like 'MyEvent can only appear on the left hand side of += or -='.

That's weird. Something else must be breaking it for me, because I can't edit the text boxes. I have a solution based on switching templates, which has got me most of the way there. I will post it when I am sure it works. Thanks again.

Wow! Looks good but (I'm hoping) a bit of overkill for now. I will have a play with it when I (a) have a bit more time or (b) have finally torn my last piece of hair out.

Bit confused here. The tutorial above says at the end, "If you want the user to be able to select items from the list, then you're better off with one of the other controls, e.g. the ListBox or the ListView." I do need to keep the notion of having a selected item. Sorry if that wasn't clear. I will amend the original question.

Also, I'm sure an ItemsControl will do what I want, but I'm rather pressed for time (as I'm sure we all are) and was hoping to avoid such a learning curve.

I thought I had a found a simple solution, which I have added to my question, and wonder if you (or anyone) can spot why it isn't working?

There is always the possibility that I might need the same functionality elsewhere, but for now this is a one-off in this application. If I need to reuse it, I'll worry about it later. Hopefully by then I will have a better understanding of all this.

Hi.

I'm afraid my choice was based simply on the fact that I didn't know there was an alternative. The list is not part of a master-detail view or anything like that. The screen displays a single customer on the left and a list of their contacts on the right. Making sure the correct list of contacts is displayed for the customer is already working fine, so the actual list of contacts can be thought of as existing in isolation if that makes sense.

Thank you, Matt. That is very helpful.

The project is WPF, and I have tried to keep reasonably closely to the MVVM pattern, but I am not using any specific framework. Hope that answers your question.

You comment this is not a particularly good case for comparing the two syntax variants. Is it true to say then that, if I want consistency, I might do better staying with the plain C# syntax? Is it more flexible than the pure SQL-ish form?

Arrrgghhh... think my brain has already broken up for Christmas. It is so frustratingly obvious now you have explained it. Thank you!

PS - Any chance of an up-vote on my question? Surely it will be of use to other learners, and I did put some effort into it being written reasonably well.

Okay, this is maybe not too elegant but it appears to work. Thank you again.

sqlCommand.Parameters.AddWithValue( "FK_Contact", ( ContactID == -1 ) ? (object)DBNull.Value : ContactID );

Thank you very much for that. I'm very much an SQL-Learner and it would have taken me a while to remember the section I read on LEFT JOIN in my book.

I should have mentioned that I am using SQLite. According to the docs, SQLite foreign keys can be null by default, unless you add the NOT NULL to the column definition, so my tables should be okay as they stand.

My problem now is that I can't figure out how to get DBNull to be accepted as a parameter from C# into a SQLiteCommand. Various casts on DBNull.Value don't work.

I'll keep trying and post back if I find something that works.

Oh, hang on. I've mixed up two use cases. When the value being returned is a valid number, its type is 'long', which explains the invalid cast. Doh! My stupid mistake. Thank you for letting me bounce this off you. I'd have spent a lot longer getting to the result otherwise.

It shows (with each line nested inside the one above):
result {} object {System.DBNull}
Static members
Value {} System.DBNull

Actually, now you have got me to write it out, I've suddenly realised how the hell is it getting past my if() statement if the value is DBNull? It implies I have two different values for DBNull. Is it okay to test for DBNull with the standard '==' operator?

It shows a valid positive integer.

Unfortunately, I spoke too soon. I now have the belt-and-braces approach shown below. When control gets past my over-the-top if() statement, the debugger shows a valid positive integer; however, when I perform the cast I get an invalid cast exception. Anyone have any ideas what could be going on? Think I might have to try the ExecuteNonQuery and named column approach instead. I just don't appear to be able to do anything reliable with the result from ExecuteScalar.

SQLiteDataStore.Connection.Open();
string sqlScript = "SELECT max( enquiry_job_number ) FROM EnquiryJobs;";
SQLiteCommand sqlCommand = new SQLiteCommand( sqlScript, SQLiteDataStore.Connection );
object result = sqlCommand.ExecuteScalar();
if( result != null && result.ToString() != String.Empty && result != DBNull.Value )
{
// At least one record in DB; debugger shows positive int but the following cast gives an exception. :(
int resultAsInteger = (int)result;
}
else
{
// No records in DB.
}

The code is actually split amongst a few helper functions, but pulling it together and stripping out error checking, it is this:

SQLiteDataStore.Connection.Open();
string sqlScript = "SELECT max( enquiry_job_number ) FROM EnquiryJobs;";
SQLiteCommand sqlCommand = new SQLiteCommand( sqlScript, SQLiteDataStore.Connection );
object result = sqlCommand.ExecuteScalar();
if( result != null && result != DBNull.Value )
{
// At least one record in DB.
}
else
{
// No records in DB.
}

Ah... the following works:

if( result != DBNull.Value )
{
...
}
Thank you for your help.

Using the following code:

if( result != null )
{
string s = (string)result;
int x = 10;
}

...I can't get as far as a breakpoint on the assignment to x because the cast to a string causes an exception.

I'm afraid that doesn't appear to work. I test against null, get false, but then get an exception when I try to cast the result to an int. Do I need to test for DbNull or something? (As you've no doubt gathered, I'm no SQL expert.)

One problem I am having with ExecuteScalar() is that when it is first run on an empty database, I am not getting null. I get an object that simply shows an empty set of brackets in the debugger. I'm not sure what this is or, more importantly, what if() statement I can write to catch this first run.

Thank you for that. I do normally name return values. I wrongly assumed that the name would be the same as the column name.

Outstanding! That works perfectly. That might have occurred to me before Christmas if I were lucky. Thank you.

Ah, your comment about AppDomains got me thinking. It turns out to be the simple fact that I was running it in the debugger. I could kick myself for not trying it using a simple double-click on the .exe.

Thank you very much for your help.

Yes, again on both the control and the containing window.

No. I've added event handlers for DragOver and PreviewDragOver, to both the control and the containing window. None are ever hit.

I'll have to play with the library above, to try to work out what it is doing that I am not. I'd simply use it, but it seems like overkill for my simple requirement.

Thank you for the information. I wouldn't have worked that out any time soon.

I still get a no-entry sign when I drag over the application. I have tried adding the event handlers to the window itself, but still it doesn't work, so I am obviously missing a key piece of the jigsaw somewhere.

I guess I'm being thick here, but I can't seem to find any setting that changes the default grey colour.

Thank you for that.

Thank you for that link. Jeez... could it be any more complicated? That seems like a monstrous amount of work to change a single property of the standard ComboBox. Is there any way to get a copy of the in-built control, so I could just change the bit I want?

Thank you for those replies. I have learnt a lot. One thing I have learnt is to choose less-confusing names for example classes. I should have chosen something different to 'Car' and 'Vehicle' because, as I mention in my question, the two types don't have a base-subclass relationship. I will know what to do if I do encounter that situation though. :)

It is a set of TextBox controls inside a Grid. It just would feel more natural if pressing Esc or clicking on a 'blank' area of the window removed selection from whichever TextBox currently has keyboard focus.

Thank you for a very thorough explanation.

Thanks for the tips. I'll have a poke about. I am a bit worried though because I'm pretty sure I only have the controls I have mentioned. I must be mistaken.

Higher up the tree, I have DockPanel, Border and other instances of Grid. Your explanation is one of the clearest and most concise I have read on WPF layout, and hence the 'Accept Answer' and 5-stars because I am sure it will lead to a solution when I get time to play with it. Thank you.

Just spotted it. The class is a custom collection and it looks like it has not been fully implemented - i.e. it does not implement ICollection. My bad for not checking the type of it. Since it was using methods like Add() and RemoveAt(), I just assumed it was a List. :-|

Thank you for the reply.

It must be strange indeed. I just put the following inside the class that contains the actual instance of the List<t></t>.

int a = _recentDevices.Entries.Count;
_recentDevices.RemoveAt( index );
int b = _recentDevices.Entries.Count;

And the Count remains unchanged. I am running debug code on an Android emulator. This is also the C# that comes with Xamarin. Wonder if any of these have anything to do with it? Sigh... I feel a whole bunch of time-consuming experiments coming on...

Perfect! Thank you.

You are exactly right, sir. The MSDN example works fine when moved to the MainWindow. Seems like a pretty major flaw to me. My requirement to access a setting prior to displaying the first window doesn't seem unreasonable. What I am wanting to do in fact is change the StartupUri depending on a setting, so I will have to work around this. Thank you for your help.

Thanks for the reply. Do you have any idea why the Properties.Settings member is not even present? Do you think the docs are wrong?

I see (I think). Thank you for your help, Richard.

That is pretty much it. Excellent! My only remaining problem is setting the initial size of the Canvas. Surely there is a better way than just basing my initial guess on the set window size minus the other stuff I've already put in there?

That has helped, Richard. Unfortunately, what I really want is exactly what you said will not happen. Is there a way to get the Canvas to start at the maximum size of its containing element? I tried setting the Binding Mode to OneTime but my ellipses then vanish for reasons I won't pretend to understand.

I guess I was more concerned about the logic of it, rather than it causing any physical problems. I did think TrySetResult() would fail safely (as you say, the clue is in the name). Now I think about it on a new day, I can see that leaving the result in an undefined state is logically okay because the calling code will have received the time-out exception and therefore _should_ assume it is not safe to try to use the result.

I was aware of the CancellationToken stuff, but I have no access to the code of some of the tasks to which I need to apply a time-out and, guess what, they do not have a CancellationToken parameter.

Thank you for the helpful reply.

I've up-voted this to 5* because this discussion really got me thinking about this, and by playing about with the options above, I discovered that part of the problem was a bug I was introducing. One of the pitfalls of trying to post the smallest code example that demonstrates the problem. Think I need to put my hands up and say that I must not have actually run the _exact_ same code I posted above. Sorry, guys, if I have wasted anyone's time.

So, to be clear, I can get $"x:00" to work as expected using the Xamarin.iOS compiler. And the tool suggested above (https://code.msdn.microsoft.com/net-framework-4-formatting-9c4dae8d) is a god-send for playing with this stuff quickly.

Thank you very much for the detailed assistance. - Patrick

This pads the integer, but using a space character, not a leading zero (for me anyway - see comment about compiler above).

Mmm... maybe it is a compiler issue. I am actually in a Xamarin.iOS application, which I think may well use the Roslyn compiler.

If I try this:

int x = 3;
string s = String.Format( "{0:D2}", x );
Console.WriteLine( s );

I still don't get a padded integer.

Still using something like the loop above, I have noticed that the the thread number keeps going up in the debugger. I get messages like:

Thread started: <Thread Pool> #8
Thread started: <Thread Pool> #9
Thread started: <Thread Pool> #10
Thread started: <Thread Pool> #11
Thread finished: <Thread Pool> #6
Thread finished: <Thread Pool> #8
Thread started: <Thread Pool> #12
Thread finished: <Thread Pool> #10

Note that I don't know if the total number of threads is rising, but the ID number does keep going up. Is this normal? (I can attempt to match start-finish messages, take all these away, and see how many threads are left, but I just wanted to ask here first, to see if I am worrying over nothing.)

I see. Think I need to do some restructuring of my code.

Ooo... seems like I suddenly have more reading to do. I've tried to do thorough research on this. I've even bought a couple of books, which admittedly I have only just started reading. But I have not read anything to indicate why putting Tasks in a list and starting them later is a bad idea. I ~think~ I understand why, but it is also obvious there is a big hole in my knowledge. Thank you.

It is a nice article, but I don't have the System.Windows.Forms namespace available, because I am in a Xamarin.Forms application. Is there a 'more generic' alternative?

Sorry, so are you saying that the async method will be triggered as soon as I add it to the List?

I cannot believe I have lasted this long without knowing about http://referencesource.microsoft.com/.

Thank you - both for the answer above and learning about this new (to me) resource.

That is indeed a great article. It also lead me to find the Channel 9 stuff, which is an excellent introduction to the most common pitfalls of using async/await. Thank you.

Thank you for the explanation.

That's pretty much what I thought. Maybe it just looks strange to me BECAUSE I come from a C++ background.

Outstanding, sir! Thank you very much.

In these days of colour-coding editors that take account of which compiler directives are active, I don't actually see the need to stick to the ancient notion that compiler directives should always go against the left margin, but that's a subject for another thread.

Thank you very much for the information.

Kind wishes ~ Patrick

Now I see it, you are right - it seems dead simple. The compiler message was throwing me off, I think - something about not being able to cast to an Action<T>.

Thank you.

Thank you for that, Sergey (and also your patience). I am actually using this XAML in a Xamarin.Forms Portable Class Library. I am very new to this, but it appears that not all the features found in WPF have been implemented. As I said, I'm really just experimenting, to try to gain more understanding. I will bear your comments in mind.

Hi, Sergey.

I am simply experimenting with XAML, to try to improve my understanding. I thought that, in general, if you had a UI element with a property then you could set the property one of two ways. You can either do this:

<mycontrol myproperty="Value-which-may-be-a-binding-expression"/>

Or you could achieve exactly the same thing like this:

<mycontrol>
<mycontrol.myproperty>Value-goes-here</mycontrol.myproperty>
</mycontrol>

I can get this to work with various properties but can't work out how to do it with the Button.Command property. But I think my problem is not specifically the Command property; I think it is the fact that I don't know how to write a binding expression using the second syntax above.

My specific question is: how do you write a binding expression where I have written 'Value-goes-here'?

Patrick

Sorry the code was not complete. I was trying (too hard) to be concise. Slapped hands for me.

Thank you for all the replies. I think I understand your answer, Sergey. The part that really confused me was the (Action<object>) where this() chains to the first constructor. Is this a plain cast? I've never seen a cast in this context before.

Thank you for the reply. I was actually doing something similar to your suggestion, but I must have missed one or two references. I've now done a full search-and-replace to fix the package path and the whole solution builds without error.