|
johannesnestler wrote: couldn't let your "object x =" stand
In no way did I intend to imply it was equivalent. I did mean to imply they are equivalent in the sense of "I'm too lazy to look up or care what object I am dealing with." I have to admit I am too lazy to look it up now, but if you are passed an object field and set "var x = field". I know the true class isn't ever lost, but is x now the true object type?
|
|
|
|
|
The "var" thing in C# is often miss-interpreted. Because everyone thinks of runtime type identification (VB...). But in C# it's not - it's only syntactic sugar if you don't want to retype complicated Type-names like: Dictionary<mykeyclass<int>, List<list<double>>>, but the compiler always knows what type it is - it's normal strong compile-time type checking. Even the IntelliSense knows the type.
But there are situations it is neccessary: you - as the programmer - sometimes do not know which type an expression will give back (anonymouse types), but the compiler will create a class for the anonymouse type during compilation and can fill in the missing type instead of the var keywords - but again: AT COMPILE TIME.
So in your example: object field = new SomeClass(); var x = field; will result in x interpreted as object not SomeClass. Don't mix that up with GetType() or typeof operator in c# - you can always ask an object for it's type - AT RUNTIME.
This is an better example: object field1 = new SomeClass(); var field2 = new SomeClass(); Where for field1 you are holding a SomeClass-instance as a base class reference (everything derives from object) where field2 is totally equivalent to: SomeClass field2 = new SomeClass();
If you want my personal opinion: I just use it where needed (anonymouse types) or for LINQ (there you often have those ugly long typenames). For me it had another nice side-effect I didn't think of in the first place: If you have specific algorithms or snippets and used the var keyword instead of the real type, you get code that is great for copy-pasting arround. This sometimes helps to focus on the underlaying constructs - and it helped me "to see" how repetitive code can be unified.
var regards = from reg in AllRegards
where reg.Quality = "best"
select new { Regards=reg, Quality=reg.Quality };
(this would be valid C# (LINQ) giving back an anonymouse type - you can not know it's Name, so use of var keyword is mandatory...)
|
|
|
|
|
Seems about the same as doing this in C or C++:
/* What could possibly go wrong */
#define int double
Since the substitutions are done before the tokens are interpreted.
|
|
|
|
|
What's old is new. I found the following:
#define void int in some C code when I started my current job.
Software Zen: delete this;
|
|
|
|
|
Another two (three?) hours spent on this kind of typo:
public int LoadedSampleCount
{
get { return (int)GetValue(LoadedSampleCountProperty); }
set { SetValue(LoadedSampleCountProperty, value); }
}
public static readonly DependencyProperty LoadedSampleCountProperty =
DependencyProperty.Register("LoadedSampleCount", typeof(int), typeof(TreeWindowVM SampleSetLoaderVM), new PropertyMetadata(0));
This is a result of a "GUI refactoring" -- moving parts of code to an user control. And forgetting to set an owner of a dependency property to a new model view again. Awww... 
Greetings - Jacek
|
|
|
|
|
Seems like some people like to make things in C# more complex than they need to be. Makes it much easier to generate bugs. They must have a great desire to become an entomologist.
One of the things that confused me was that people were changing readonly objects. Then the oh, duh, moment. the address and definition of the object can't change. Internal changes don't count as changing the object.
|
|
|
|
|
The owner of a dep roperty has to be changed to a new ViewModel, otherwise it wouldn't be visible in the XAML, in which a DataContext had been changed to that new ViewModel. Or I get it all wrong.
Greetings - Jacek
|
|
|
|
|
Sorry for late comment... but I have to ask: You think "readonly" should mean the same like "const"? Did you never had the need to "calculate" a later constant value? I like the "readonly" construct (keyword) and think it's a good thing to have. (and you can only reset in in constructors and not everywhere)
To the related topic you mentioned - that changing the through a readonly variable represented object doesn't count as changing the variable, and isn't forbidden -> it's completly clear for me that a pointer-variable change means to change it's content - the address it is representing not the object it is pointing to. That's also true for const pointers in C++ and all other languages I know - have you seen a language where a constant pointer means that the pointed object is immutable?
|
|
|
|
|
johannesnestler wrote: I have to ask: You think "readonly" should mean the same like "const"?
Not really, I don't recall talking with the instructor about either const or readonly. And I didn't read anything in the books I'd bought for the training. I just recall seeing it for the first time on the job and my amazement that the values in the readonly object were being changed and the compiler wasn't complaining. Then an "Oh, duh" moment. You can't change the instance, but you can change the contents of the instance. This code is totally legal:
public class vals
{
public int age;
public DateTime date;
}
class x
{
public readonly vals val = new vals();
}
static void Main(string[] args)
{
x val = new x();
val.val.age = 20;
val.val.date = DateTime.Now;
...const has it's own peculiarities, basically it supports value types. When you read the documentation it says the only non-value type is string. I was taught that string was also a value type. I haven't read any documentation (other than const) that contradicts string is a value type. I got an eye opener several years ago when I read that you shouldn't send a string to unmanaged C++ code. Before that, I was totally confused by the statement that strings were IMMUTABLE. I can change the value of the string whenever I want! After I read that article I understood what was meant by that and it makes me believe that you can change a const string variable's value by passing it to that C++ code I mentioned. Much later I learned about Intern and IsInterned.
Because I am not dedicated to this language I don't consider myself near an expert in C#. More knowledgeable than some C# programmers, most dedicated C# developers know more than me.
|
|
|
|
|
I think the idea for the readonly keyword in C# was stolen from JAVA (final keyword). And solves the problem I mentioned. (What to do if a later constant value must be calculated?)
In the example you have given, you can see what I mean: What did you declare as readonly? The age or date properties? -> No, You declared the val variable of type vals as readonly.
Because vals is a reference type (you created it with the class keyword) the variable val is now holding the address of the new vals instance on the heap. - so val is a pointer. This is the variable that is immutable, you can not change this pointer (the address it is holding) in an other way but the rules for readonly keyword usage are dictating.
So if you don't know any language where a "constant pointer" means the object pointed to is immutable (nor do I), how would you expect it to behave in an imaginary language? What if the pointed object is holding a lot of other references (pointers). Should they be then immutable too? I think if you go that thought to the end, you will come to the conclusion that the current behaviour is the only logical consistent one - please correct me if I'm wrong...
greetings from Austria ![Rose | [Rose]](https://www.codeproject.com/script/Forums/Images/rose.gif)
|
|
|
|
|
The only thing wrong with the way readonly works is my initial impression of it when I first encountered it. I got that impression before I read any documentation on the property. First impressions can be exactly right, but they are usually wrong. As soon as I got it through my head that the object I was looking at was a reference type it came clear how readonly worked. Doesn't even have to be a reference type, everything has a reference location in memory, that's how const can work as well.
|
|
|
|
|
Just came across the following in some code I maintain:
foreach (Object item in group)
{
int i = 1;
string key = "SomeString_" + i.ToString();
}
I suspect the idea was to use i to help make the dictionary keys unique.
|
|
|
|
|
try
{
account.Save();
}
catch { ; }
|
|
|
|
|
At least there are comments.
|
|
|
|
|
I can think of a few comments myself...
The universe is composed of electrons, neutrons, protons and......morons. (ThePhantomUpvoter)
|
|
|
|
|
Guess programmer was hungry thats why he "ate" the exception 
Always Keep Smiling.
Yours Pankaj Nikam
|
|
|
|
|
Relax.
This is just sweeping under the rug.
It could be worse.
Mislim, dakle jeo sam.
|
|
|
|
|
I'm impressed they wasted the energy on a surplus semicolon.
Software Zen: delete this;
|
|
|
|
|
They might have a style rule that enforces:
"catch block cannot be empty"
Warning: Empty catch block.
ftfy: ";"
|
|
|
|
|
... which just demonstrates how misleading style enforcement can be.
Software Zen: delete this;
|
|
|
|
|
What kills me is the comment. It would have taken the same effort to say Logger.Log(ex);
|
|
|
|
|
I'm trying (in vain so far) to think up any scenario where the context around this might make this o.k. For example, something like:
int pageIndex = 0;
try {
pageIndex = int.TryParse(Request[pageNum])
} catch (Exception e) {
Logger.info("Page Index of " + Request[pageNum] + " invalid, using " + pageIndex.ToString());
}
(Of course, I'd use Int.TryParse instead and null-check the request variable)
But, you get the idea--some scenario where you can continue to operate with default data given an unexpected condition. I can't think of any situation where the missing context makes it o.k. to swallow an exception trying whatever a Save operation does.
|
|
|
|
|
In this case it's just plain sloppy, lazy, and wrong. It should log and throw, allowing the error to propagate to the global error handler. The object's been validated so if the save fails it's a critical failure somewhere in the system, like the database is down. Eating the error when a save fails is never the right thing to do. The user blissfully goes about their business because the save "worked" but it didn't. Wrong, wrong, wrong.
|
|
|
|
|
I can think of situations where it might be appropriate to do something like that.
For instance, suppose you are implementing a "like" button or something similar. It's not critical that it works and let's assume that it's unreliable for reasons beyond the programmer's control, like maybe it depends on an external service which is not always available.
So in that case maybe it's OK to swallow the exception since it's not unusual and nothing will really break if it fails, so you just silently fail and the user can try again if they want. There aren't many situations like this in programming though, and you still should probably log the exceptions.
I'm pretty sure that something like account.Save() is a bit too important to treat this way, though.
|
|
|
|
|
StatementTerminator wrote: implementing a "like" button or something similar. It's not critical that it works
Man, people split up because of a 'like' button. You don't 'like' in time -- you loose. It IS crtical. 
Greetings - Jacek
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
