In all examples of F# classes and records I see that records an classes are instantiated/created via new keyword or simply by the type name.
So for example if I have this record:
type MyRecord = {field1: int; field2:int}
let myvar = {new MyRecord with field1 = 3 and field2 = 3}
let myvar2 = {MyRecord with field1 = 1 and field2 = 2}
let myvar3 = {field1 = 34; field2 = 23}
They seem to be all the same.
It is the same for this example too:
type MyClass(x: int) =
let mutable xx = x
let myvar = MyClass(12)
let myvar2 = new MyClassw(234)
Well what is new for?? Thnkyou
Ah... I know that there are answers for the question of new regarding classes, but there is no mention for records. Furthermore, I do not understand why new should be an optional and abosultely indifferent keywork when constructing classes? Is it possible that using new or not using it for classes does change nothing?
The new keyword is typically not used to instantiate F# records. The first example is the OCaml-ish syntax for record declarations:
let myvar = {new MyRecord with field1 = 3 and field2 = 3}
// Normal F#: let myvar = { MyRecord.field1 = 3; field2 = 3}
The second example doesn't compile any more:
// Not F#: let myvar2 = {MyRecord with field1 = 1 and field2 = 2}
For classes, you can always omit new. However, you get a compiler warning if you instantiate an IDisposable class without using the new keyword.
// These two are the same
let myvar = MyClass(12)
let myvar2 = new MyClass(234)
// Compiler warning
let f = FileStream("hello.txt", FileMode.Open)
// No warning
use f = new FileStream("hello.txt", FileMode.Open)
Edit: In response to your comment -
isn't there a rule or s rationale why new is so "flexible"? Why is it the same using or not using new for classes
new is optional - I'm not aware of a rationale for making it flexible; it would be nice if there was some guidance one way or the other. (My preference is to never use new, except in response to the compiler warning.)
why do I get a compiler warning when implementing IDisposable
Since I never use new normally, I take this to be the compiler's way of reminding me to write use or using when I allocate an IDisposable.
and why records do not use the second syntax I wrote before?
{ new MyRecord with ... } is the Haskell syntax. It may have been valid in one of the F# compiler betas, but it doesn't parse on F# 2.0. (Why do you think this ought to be valid?)
I'm trying to find a general logic rule/guideline rather than learning byheart a bunch or scattered syntax rules...
I know how you feel -- F# is a little like this, particularly when you come from a language like C#, where there's a right way and a wrong way to do something. My advice is to pick one of the alternatives and stick to it.
This may help: the F# Component Design guidelines [PDF] from Microsoft. It's a set of advice -- dos and don'ts -- for your F# code.
Related
I get an error when I put the type and size of an array of classes
I have tried:
fun main(args :Array<String>) {
class modul() {
var nommodul: String? = null
var coeff: Int? = null
var note: Int? = null
}
var releve
class notes() {
var releve: array<modul>(10){""} here the erreur
}
}
First of all, your code has several errors. This might be an MCVE and/or copy-paste issue, but I need to address these before I get started on the arrays.
var releve before the notes class isn't allowed. You don't assign it, you don't declare a type, and the compiler will complain if you copy-paste the code from your question.
Second, the array var itself: Array is upper-case, and initialization is separate. This would be more valid (note that this still does not work - the solution for that comes later in this answer):
var releve: Array<modul> = Array(10) {...}
// or
var releve = Array<modul>(10) {...}
And the last thing before I start on the array itself: please read the language conventions, especially the naming ones. Your classes should all start with an upper-case letter.
Kotlin arrays are quite different from Java arrays in many ways, but the most notable one being that direct array initialization also requires an initializer.
The brackets are expected to create a new instance, which you don't. You create a String, which isn't, in your case, a modul.
There are several ways to fix this depending on how you want to do this.
If you have instances you want to add to the array, you can use arrayOf:
arrayOf(modulInstance, modulInstance2, ...)
If you want to create them directly, you can use your approach:
var releve = Array(10) { modul() }
A note about both of these: because of the initialization, you get automatic type inference and don't need to explicitly declare <modul>
If you want Java-style arrays, you need an array of nulls.
There's two ways to do this:
var releve = arrayOfNulls<modul>(10)
// or
var releve = Array<modul?>(10) { null }
I highly recommend the first one, because it's cleaner. I'm not sure if there's a difference performance-wise though.
Note that this does infer a nullable type to the array, but it lets you work with arrays in a similar way to Java. Initialization from this point is just like Java: releve[i] = modul(). This approach is mostly useful if you have arguments you want to add to each of the classes and you need to do so manually. Using the manual initializers also provides you with an index (see the documentation) which you can use while initializing.
Note that if you're using a for loop to initialize, you can use Array(10) { YourClass() } as well, and use the supplied index if you need any index-sensitive information, such as function arguments. There's of course nothing wrong with using a for loop, but it can be cleaner.
Further reading
Array
Lambdas
here some example of kotlin array initialization:
array of Library Method
val strings = arrayOf("January", "February", "March")
Primitive Arrays
val numbers: IntArray = intArrayOf(10, 20, 30, 40, 50)
Late Initialization with Indices
val array = arrayOfNulls<Number>(5)
for (i in array.indices) {
array[i] = i * i
}
See Kotlin - Basic Types for details
How to get the list of types that are extended by a extension method in ndepend cqlinq? Using reflection to code this seems a bit of donkey work where ndepend is already there.
NDepend code model doesn't have a straight way to resolve the method parameter type. So we can come up with a satisfying answer with code query relying on string formatting extended type name, extracted from the method name. But this query is overly complex and there are edge cases where it won't work properly (explained below).
Here is the code query, it runs fast even on large code base thanks to the use of a dictionary:
//
// First let build dicoExtensionMethods
let dicoExtensionMethods =
(from m in Application.Methods
where m.IsExtensionMethod
// extract extended type simple name (with generic parameters like "IEnumerable<String>")
let beginIndex = m.Name.IndexOf("(") + 1
let endIndex = m.Name.IndexOf(',', beginIndex) > 0 ? m.Name.IndexOf(',', beginIndex) : m.Name.IndexOf(")", beginIndex)
let extendedTypeSimpleName1 = m.Name.Substring(beginIndex, endIndex - beginIndex)
// Take care of generic type first char, like "IEnumerable<"
let extendedTypeSimpleName2 = extendedTypeSimpleName1.IndexOf('<') == -1 ? extendedTypeSimpleName1 :
extendedTypeSimpleName1.Substring(0, extendedTypeSimpleName1.IndexOf('<') + 1 )
select new { m, extendedTypeSimpleName2 })
.ToLookup(pair => pair.extendedTypeSimpleName2)
.ToDictionary(g => g.Key, g=> g.Select(p =>p.m))
//
// Second for each type get extension methods from dicoExtensionMethods
from t in Types
// Format type name like "IEnumerable<"
let typeName = !t.IsGeneric ? t.SimpleName : t.Name.Substring(0, t.Name.IndexOf('<') + 1 )
where dicoExtensionMethods.ContainsKey(typeName)
let methods = dicoExtensionMethods[typeName]
select new { t, methods }
As written it is a complex query because of type name formatting, and it works fine most of the time.
However when it comes to extending generic types, it says for example that IEnumerable<T> is extended by both methods that extend IEnumerable<String> and IEnumerable<Int32>. This is acceptable, but it is not 100% correct.
Also if you extend several types with same name but various generic arity (like Func<T1,T2> and Func<T1,T2,T3>), then this code query won't work properly.
The same if you extend several types with same name, declared in different assemblies or namespace (which is a code smell anyway).
Hope this helps!
With the nDepend API, would something like the following be possible?
I want to keep a watch out for instances where our object factory has been bypassed and a concrete class is being instantiated directly.
Obviously I'd need to be able to filter out things like:
StringBuilder stringBuilder = new StringBuilder();
perhaps by adding to the Where clause type names to exclude, or namespaces in which to check, but I want to make sure we see:
IMyCustomType item = ObjectFactory.Get<IMyCustomType>();
and not this:
MyCustomType item = new MyCustomType();
Thanks.
Maybe such code rule below could help you, hopefully it is understandable enough to not have to comment it:
warnif count > 0
let ctors = Application.Namespaces.WithNameLike("Namespaces1*").ChildMethods().Where(m => m.IsConstructor)
let codeThatMustNotCallCtors = Application.Namespaces.WithNameLike("Namespaces2*").ChildMethods()
from m in codeThatMustNotCallCtors.UsingAny(ctors)
select new { m, ctorsCalled = m.MethodsCalled.Intersect(ctors ) }
Hi,
I'm developing an Eclipse plugin. I
need to find all the references in the
source using AST's or jdt.core.dom
or something like that. I need this
references like ASTNodes in order to
get the parent node and check several
things in the expression where
references are involved.
Thanks beforehand.
Edited:
I want to concrete a little more, My problem is that I try to catch some references to a constant but... I have not idea how I can do to catch in the matches this references. I need check the expressions which the references to a determined constant are involved. I only get the source of the method where they are used.
I think the problem is the scope or the pattern:
pattern = SearchPattern.createPattern(field, IJavaSearchConstants.REFERENCES);
scope = SearchEngine.createJavaSearchScope(declaringType.getMethods());
Thanks beforehand!
I used something like:
Search for the declaration of an
method, returns an IMethod
Search for references to the
IMethod, record those IMethods
For each IMethod returned, create an
AST from its compilation unit
Searching for declarations or references looks like the following code.
SearchRequestor findMethod = ...; // do something with the search results
SearchEngine engine = new SearchEngine();
IJavaSearchScope workspaceScope = SearchEngine.createWorkspaceScope();
SearchPattern pattern = SearchPattern.createPattern(searchString,
IJavaSearchConstants.METHOD, IJavaSearchConstants.DECLARATIONS,
SearchPattern.R_EXACT_MATCH);
SearchParticipant[] participant = new SearchParticipant[] { SearchEngine
.getDefaultSearchParticipant() };
engine.search(pattern, participant, workspaceScope, findMethod,
monitor);
Once you have your IMethod references, you can get to the AST using:
ASTParser parser = ASTParser.newParser(AST.JLS3);
parser.setResolveBindings(true);
if (methodToSearch.isBinary()) {
parser.setSource(methodToSearch.getClassFile());
} else {
parser.setSource(methodToSearch.getCompilationUnit());
}
CompilationUnit cu = (CompilationUnit) parser.createAST(null);
See http://help.eclipse.org/helios/index.jsp?topic=/org.eclipse.jdt.doc.isv/guide/jdt_int_core.htm for more details on java search, the java model, and the AST.
I am attempting to create two identical objects in VB6 by assignment statements; something like this...
Dim myobj1 As Class1
Dim myobj2 As Class1
Set myobj1 = New Class1
myobj1.myval = 1
Set myobj2 = myobj1
It has become apparent that this doesn't create two objects but rather two references to the same object, which isn't what I am after. Is there any way to create a second object in this sort of way, or do I have to copy the object one member at a time...
Set myobj2 = new Class1
myobj2.mem1 = myobj1.mem1
...
?
Edit 2 Scott Whitlock has updated his excellent answer and I have incorporated his changes into this now-working code snippet.
Private Type MyMemento
Value1 As Integer
Value2 As String
End Type
Private Memento As MyMemento
Public Property Let myval(ByVal newval As Integer)
Memento.Value1 = newval
End Property
Public Property Get myval() As Integer
myval = Memento.Value1
End Property
Friend Property Let SetMemento(new_memento As MyMemento)
Memento = new_memento
End Property
Public Function Copy() As Class1
Dim Result As Class1
Set Result = New Class1
Result.SetMemento = Memento
Set Copy = Result
End Function
One then performs the assignment in the code thus...
Set mysecondobj = myfirstobj.Copy
Like many modern languages, VB6 has value types and reference types. Classes define reference types. On the other hand, your basic types like Integer are value types.
The basic difference is in assignment:
Dim a as Integer
Dim b as Integer
a = 2
b = a
a = 1
The result is that a is 1 and b is 2. That's because assignment in value types makes a copy. That's because each variable has space allocated for the value on the stack (in the case of VB6, an Integer takes up 2 bytes on the stack).
For classes, it works differently:
Dim a as MyClass
Dim b as MyClass
Set a = New MyClass
a.Value1 = 2
Set b = a
a.Value1 = 1
The result is that both a.Value1 and b.Value1 are 1. That's because the state of the object is stored in the heap, not on the stack. Only the reference to the object is stored on the stack, so Set b = a overwrites the reference. Interestingly, VB6 is explicit about this by forcing you to use the Set keyword. Most other modern languages don't require this.
Now, you can create your own value types (in VB6 they're called User Defined Types, but in most other languages they're called structs or structures). Here's a tutorial.
The differences between a class and a user defined type (aside from a class being a reference type and a UDT being a value type) is that a class can contain behaviors (methods and properties) where a UDT cannot. If you're just looking for a record-type class, then a UDT may be your solution.
You can use a mix of these techniques. Let's say you need a Class because you have certain behaviors and calculations that you want to include along with the data. You can use the memento pattern to hold the state of an object inside of a UDT:
Type MyMemento
Value1 As Integer
Value2 As String
End Type
In your class, make sure that all your internal state is stored inside a private member of type MyMemento. Write your properties and methods so they only use data in that one private member variable.
Now making a copy of your object is simple. Just write a new method on your class called Copy() that returns a new instance of your class and initialize it with a copy of its own memento:
Private Memento As MyMemento
Friend Sub SetMemento(NewMemento As MyMemento)
Memento = NewMemento
End Sub
Public Function Copy() as MyClass
Dim Result as MyClass
Set Result = new MyClass
Call Result.SetMemento(Memento)
Set Copy = Result
End Function
The Friend only hides it from stuff outside your project, so it doesn't do much to hide the SetMemento sub, but it's all you can do with VB6.
HTH
#Scott Whitlock, I was not able to make your code work but if it works it would be great.
I've created a regular module where I put the memento type
Type MyMemento
Value1 As Integer
Value2 As String
End Type
Then I create a class module called MyClass with the code
Private Memento As MyMemento
Friend Sub SetMemento(NewMemento As MyMemento)
Memento = NewMemento
End Sub
Public Function Copy() as MyClass
Dim Result as MyClass
Set Result = new MyClass
Result.SetMemento(Memento)
Set Copy = Result
End Function
Finally I try to call the copy function in another regular module like this
Sub Pruebas()
Dim Primero As MyClass, segundo As MyClass
Set Primero = New MyClass
Set segundo = New MyClass
Set segundo = Primero.Copy
End Sub
I get the message (below the picture): Error de compilacion: El tipo de agumento de ByRef no coincide
Here is an image (short of 10 points so here is the link): http://i.stack.imgur.com/KPdBR.gif
I was not able to get the message in English, I live in Spain.
Would you be so kind to provide with an example in VBA Excel?, I have been really trying to make this work.
Thanks for your work
===============================================
EDIT: Problem Solved:
The problem was on line "Result.SetMemento(Memento)", in VBA it needed to be called with "Call"
Public Function Copy() As MyClass
Dim Result As MyClass
Set Result = New MyClass
Call Result.SetMemento(Memento)
Set Copy = Result
End Function
It works great, thanks Scott Whitlock, you are a genius
or do I have to copy the object one member at a time...
Unfortunately yes.
It is possible (but technically very very difficult) to write a COM server in C++ that - using the IDispatch interface - will copy the value of each property, but really this is High Temple programming, if I had to do it, I don't I know if I could do it, but I'd be looking at something like 10 days work ( and I know how COM is implemented in C++, I'd also need to investigate to see if ATL framework has anything to help etc).
I worked with Vb3, 4,5 & 6 for something like 10 years (hands on, 5 days a week) and never found a good way to do this, beyond manually implementing serialisation patterns like Mementos and Save & Store, which really just boiled down to fancy ways of copying each member, one at a time.