have completion for #:op(a.b) - haxe

The #:op(a.b) feature is described here: https://haxe.io/releases/3.3.0/
I have May<T> abstract which is used for null safety. Here is the simplified version of it:
package;
import haxe.macro.Expr;
abstract May<T>(Null<T>) from(Null<T>){
// convert T, Null<T> or May<T> to May<T>
// accepts Null<T because of 'from(Null<T>)'
// accepts T because Null<T> has 'from(T)'
// we need to accept May<T> to avoid returning May<May<T>> from resolve() if field is already a May
public static inline function from<T>(t:May<T>):May<T> return t;
public inline function exist():Bool return this != null;
public inline function unwrap():T return exist() ? unsafeUnwrap() : throw 'unwrap null';
public inline function unwrapOr(defaultValue:T):T return exist() ? unsafeUnwrap() : defaultValue;
public inline function unsafeUnwrap():T return this;
// return same field from underlying type but as May<FieldType>
#:op(a.b) static macro function resolve<T>(ethis:ExprOf<May<T>>, name:String):Expr {
return macro {
var me = $ethis;
var result = May.from(me.exist() ? me.unsafeUnwrap().$name : null);
result;
}
}
}
Note the resolve() function. It's the new feature that I want to add to my actual May abstract. It allows to safely get fields from May and call unwrap() only once. For example:
may.exist() ? may.unwrap().someField : defaultValue
becomes
may.someField.unwrapOr(defaultValue)
That's very handy and works good. But the completion does not work. It only gives fields from May: unwrap(), exist() etc., but no fields from the underlying class.
I've decided to add #:forward metadata for completion:
#if display #:forward #end
This makes the compiler see all fields during completion. It's better than nothing, but fields have an incorrect type: T instead of May<T>, so I do not get completion for May fields.
I understand why the compiler can't know all possible fields when using #:op(a.b), but maybe there is some more clever trick that will help?

Related

Implementing Set<T>

I'm trying to wrap my head around abstract by implementing a Set data-type, like so:
abstract Set<T>(Map<T, Bool>) {
public inline function new() {
this = new Map<T, Bool>();
}
public inline function has(item:T):Bool {
return this.exists(item);
}
public inline function add(item:T):Set<T> {
this.set(item, true);
return null;
}
public inline function remove(item:T):Set<T> {
this.remove(item);
return null;
}
public inline function iterator():Iterator<T> {
return this.keys();
}
}
The compiler doesn't like this, though. It tells me Set.hx:8: characters 11-29 : Abstract Map has no #:to function that accepts IMap<util.Set.T, Bool>
I don't really understand this at all, since if I change the constructor to
public inline function new(val:Map<T, Bool>) {
this = val;
}
and then instantiate with var set = new Set(new Map());, it works.
That's pretty gross, though. I'd like the ability to instantiate Sets without exposing the underlying implementation. Ultimately, I'd prefer a constructor with the signature new(?initial:Iterable<T>). Is this possible? Am I misunderstanding something?
The problem is that currently it's impossible to instantiate Map without they key type being known (and since Set.T is a free type parameter, this doesn't work). However since the constructor is inline, T may well be known at the call site. The problem is that the compiler still tries to generate Set.new. You can avoid this by prefixing it with #:extern. Working example: https://try.haxe.org/#1D06C

In Haxe, how do you pass Enum values in functions, and then convert them to Strings within the function?

I can't seem to get this working, but I'd be surprised if it wasn't possible in Haxe.
I'm trying to pass a couple of Enum values defined in my game to a function, so that it can then concatenate them as String types and pass that to other functions.
Example:
// In a general Entity class:
public override function kill():Void {
messages.dispatchCombined(entityType, ListMessages.KILLED);
super.kill();
}
And in my Messages.hx class:
package common;
import msignal.Signal.Signal1;
/**
* A Message / Event class using Signals bound to String names.
* #author Pierre Chamberlain
*/
class Messages{
var _messages:MessagesDef;
public function new() {
_messages = new MessagesDef();
}
public function add(pType:String, pCallback:FuncDef) {
if (_messages[pType] == null) {
_messages[pType] = new Signal1<Dynamic>();
}
var signals = _messages[pType];
signals.add( pCallback );
}
public function dispatch(pType:String, pArg:Dynamic):Bool {
var signals = _messages[pType];
if (signals == null) return false;
signals.dispatch(pArg);
return true;
}
//Compiler doesn't like passing enums :(
public inline function addCombined(pSource:Enum, pEvent:Enum, pCallback:FuncDef) {
add( combine(pSource, pEvent), pCallback );
}
public inline function dispatchCombined(pSource:Enum, pEvent:Enum, pArg:Dynamic):Bool {
return dispatch( combine(pSource, pEvent), pArg);
}
//How can I just pass the enum "names" as strings?
static inline function combine(a:Enum, b:Enum):String {
return String(a) + ":" + String(b);
}
}
typedef MessagesDef = Map<String, Signal1<Dynamic>>;
typedef FuncDef = Dynamic->Void;
Note how addCombined, dispatchCombined and combine expect an "Enum" type, but in this case I'm not sure if Haxe actually expects the entire Enum "class" to be passed (ie: ListMessages instead of ListMessages.KILLED) or if a value should work. Anyways, compiler doesn't like it - so I'm assuming another special Type has to be used.
Is there another way to go about passing enums and resolving them to strings?
I think you need EnumValue as parameter type (if it is only for enum values), and use Std.String to convert to String values.
static inline function combine(a:EnumValue, b:EnumValue):String {
return Std.string(a) + ":" + Std.string(b);
}
Of course that can be written smaller using String interpolation:
static inline function combine(a:EnumValue, b:EnumValue):String {
return '$a:$b';
}
Of course that can be 'more dynamic' using type parameters:
static inline function combine<A, B>(a:A, b:B):String {
return '$a:$b';
}
There is totally no need to use Dynamic as suggested. If you use Dynamic, you basically turn off the type system.
live example:
http://try.haxe.org/#a8844
Use Dynamic instead of Enum or pass them as Strings right away since you can always convert to enum from String if you need it later.
Anyway pass the enum as enum:Dynamic and then call Std.string(enum);
EDIT: Using EnumValue is definitely better approach than Dynamic, I use Dynamic in these functions because I send more than just Enums there and I am not worried about type safety in that case.

How do I call a method of an attribute derived from a generic interface, where the specific type is not known?

Core Question:
I have a generic interface IValidatingAttribute<T>, which creates the contract bool IsValid(T value); The interface is implemented by a variety of Attributes, which all serve the purpose of determining if the current value of said Field or Property they decorate is valid per the interface spec that I'm dealing with. What I want to do is create a single validation method that will scan every field and property of the given model, and if that field or property has any attributes that implement IValidatingAttribute<T>, it should validate the value against each of those attributes. So, using reflection I have the sets of fields and properties, and within those sets I can get the list of attributes. How can I determine which attributes implement IValidatingAttribute and then call IsValid(T value)?
background:
I am working on a library project that will be used to develop a range of later projects against the interface for a common third party system. (BL Server, for those interested)
BL Server has a wide range of fairly arcane command structures that have varying validation requirements per command and parameter, and then it costs per transaction to call these commands, so one of the library requirements is to easily define the valdiation requirements at the model level to catch invalid commands before they are sent. It is also intended to aid in the development of later projects by allowing developers to catch invalid models without needing to set up the BL server connections.
Current Attempt:
Here's where I've gotten so far (IsValid is an extension method):
public interface IValidatingAttribute<T>
{
bool IsValid(T value);
}
public static bool IsValid<TObject>(this TObject sourceObject) where TObject : class, new()
{
var properties = typeof(TObject).GetProperties();
foreach (var prop in properties)
{
var attributeData = prop.GetCustomAttributesData();
foreach (var attribute in attributeData)
{
var attrType = attribute.AttributeType;
var interfaces = attrType.GetInterfaces().Where(inf => inf.IsGenericType).ToList();
if (interfaces.Any(infc => infc.Equals(typeof(IValidatingAttribute<>))))
{
var value = prop.GetValue(sourceObject);
//At this point, I know that the current attribute implements 'IValidatingAttribute<>', but I don't know what T is in that implementation.
//Also, I don't know what data type 'value' is, as it's currently boxed as an object.
//The underlying type to value will match the expected T in IValidatingAttribute.
//What I need is something like the line below:
if (!(attribute as IValidatingAttribute<T>).IsValid(value as T)) //I know this condition doesn't work, but it's what I'm trying to do.
{
return false;
}
}
}
return true;
}
}
Example usage:
Just to better explain what I am trying to achieve:
public class SomeBLRequestObject
{
/// <summary>
/// Required, only allows exactly 2 alpha characters.
/// </summary>
[MinCharacterCount(2), MaxCharacterCount(2), IsRequired, AllowedCharacterSet(CharSets.Alpha))]
public string StateCode {get; set;}
}
And then, later on in code:
...
var someBLObj = SomeBLRequestObjectFactory.Create();
if(!someBLObj.IsValid())
{
throw new InvalidObjectException("someBLObj is invalid!");
}
Thank you, I'm really looking for a solution to the problem as it stands, but I'm more than willing to listen if somebody has a viable alternative approach.
I'm trying to go generic extension method with this because there are literally hundreds of the BL Server objects, and I'm going with attributes because each of these objects can have upper double digit numbers of properties, and it's going to make things much, much easier if the requirements for each object are backed in and nice and readable for the next developer to have to use this thing.
Edit
Forgot to mention : This Question is the closest I've found, but what I really need are the contents of \\Do Something in TcKs's answer.
Well, after about 6 hours and a goods nights sleep, I realized that I was over-complicating this thing. Solved it with the following (ExtValidationInfo is the class that the below two extensions are in.):
Jon Skeet's answer over here pointed me at a better approach, although it still smells a bit, this one at least works.
public static bool IsValid<TObject>(this TObject sourceObject) where TObject : class, new()
{
var baseValidationMethod = typeof(ExtValidationInfo).GetMethod("ValidateProperty", BindingFlags.Static | BindingFlags.Public);
var properties = TypeDataHandler<TObject>.Properties;
foreach (var prop in properties)
{
var attributes = prop.GetCustomAttributes(typeof(IValidatingAttribute<>)).ToList();
if (!attributes.Any())
{
continue; // No validators, skip.
}
var propType = prop.PropertyType;
var validationMethod = baseValidationMethod.MakeGenericMethod(propType);
var propIsValid = validationMethod.Invoke(null, prop.GetValue(sourceObject), attributes);
if(!propIsValid)
{
return false;
}
}
return true;
}
public static bool ValidateProperty<TPropType>(TPropType value, List<IValidatingAttribute<TPropType>> validators)
{
foreach (var validator in validators)
{
if (!validator.IsValid(value))
{
return false;
}
}
return true;
}

How to get `this' from a passed method

If one pass a method as a funarg, how one can tell if passed function is a method, and get `this' object of a method is?
class A {
public function f():Void{
trace("f");
}
}
class B {
static function withFunarg(f:Void->Void):Void{
//HERE
}
public static function main(){
var a = new A();
withFunarg(a.f);
}
}
You cannot and there is no way to retrieve this. But it seems to me like an anti-pattern trying to do that. If you want the method and the container you can define a typedef:
typedef F = {
f : Void -> Void
}
Now you have the method and the container.
Haxe doesn't offer a cross-platform way to do that and it is generally not recomended.
But if you ultimately need this feature, you can use some platform-specific ways.
For example on js the following will work(at least on current haxe dev version):
static function getThis(f:Dynamic):Dynamic{
return (f.scope && f.method) ? f.scope : null;
}
It will return the object if the function is a method and a null otherwise. Result on calling on non-function is unspecified.
If you want to get the implicit `this' argument of a method, you have to make it explicit, like this
static function withMethodFunarg(o:{}, f:{}->Void):Void{
//HERE you have both object and function on this object
trace(o);
f(o);
}
public static function main(){
var a = new A();
withMethodFunarg(a,function(a){a.f()});
}
Which is, actually, pretty straight-forward: function is a function, no implicits, method caller is a method caller.

Code Contracts in C# 4.0

I made a method like this
class PersonCollection
{
[Contracts.CanReturnNull] //dont know if something like this exists?
IPerson GetPerson(Guid personId)
{
if (this.persons.Contains(personId))
return this.persons[personId];
else
return null;
}
}
Now the calling code needs to handle the null value properly. Is there a way to express a contract for all callers that they need to be able to handle the null value returned by this method?
PersonCollection pc = new PersonCollection();
IPerson p = pc.GetPerson(anyId);
p.Name = "Hugo"; // here I want to have a curly line
What I want is that the p gets marked as potential problematic.
EDIT
I just modified the code and added the calling code and the expcected behaviour. Also I added an attribute that probalbly does not exists on the method GetPerson
Code Contract does not provide such a feature, nor does C#
Code Contracts only requires from the caller to comply to certain constraints at the start of the called method. These are the so-called preconditions.
The postconditions are the responsibility of the callee, and defines what the state of the program will be on exit of the called method.
Design by Contract is a way to define these responsibilities, not to tell callers how they should handle certain conditions caused by the called method.
What you seem to want (after reading the comments) will happen by default:
If you enable Code Contracts in the calling code, the verifier will consider that the return of GetPerson() can be null. So:
IPerson GetPerson(Guid personId)
{
// no pre/post conditions
}
void PrintPerson(IPerson p)
{
Contract.Requires(p != null);
...
}
void Foo()
{
var p = GetPerson(id);
PrintPerson(p); // a warning here: can not verify p != null
}
And, totally irrelevant to the question, this will usually be more efficient if persons is (like) a Dictionary:
IPerson GetPerson(Guid personId)
{
Person p = null;
this.persons.TryGetValue(personId, out p);
return p;
}

Resources