rfc:compact-object-property-assignment

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revisionPrevious revision
Next revision
Previous revision
rfc:compact-object-property-assignment [2020/03/17 02:53] jgivonirfc:compact-object-property-assignment [2020/04/14 06:30] (current) jgivoni
Line 1: Line 1:
 ====== PHP RFC: Compact Object Property Assignment ====== ====== PHP RFC: Compact Object Property Assignment ======
  
-**COPA: A pragmatic approach to object literals**+**COPA: A //pragmatic// approach to object literals**
  
-  * Version: 1.1 +  * Version: 1.3 
-  * Date: 2020-03-16+  * Date: 2020-03-17
   * Author: Jakob Givoni <jakob@givoni.dk>   * Author: Jakob Givoni <jakob@givoni.dk>
-  * Status: Under Discussion +  * Status: Declined 
-  * Discussion: https:%%//%%externals.io/message/109055+  * Discussion: [[https://externals.io/message/109055]]
  
 ===== Introduction ===== ===== Introduction =====
Line 15: Line 15:
 This RFC proposes a new, compact syntax to assign values to multiple properties on an object in a single expression. This RFC proposes a new, compact syntax to assign values to multiple properties on an object in a single expression.
  
-This //**pseudo object literal**// notation, (though not limited to such use) is intended to enable the developer to create an object and populating it inline, similar to what is possible for arrays.+This //**pseudo object literal**// notation, (though not limited to such use) is intended to enable the developer to //**create an object and populating it inline**//f.ex. directly as an argument in a function call. 
 + 
 +As an alternative to writing a data structure as an associative array, COPA gives the data a **//documented signature//** so that you know what parameters are expected and their value types. 
 + 
 +> COPA does not introduce any new concepts or complexities, but merely a new syntax aimed at making millions of PHP developers write their code in a simpler way. The code becomes easier to write and read and thus more maintainable without any lateral limitations or factual downsides.
  
 ==== Example ==== ==== Example ====
Line 37: Line 41:
 ]; ];
 </code> </code>
-And that’s all there is to it - the rest follow from this, as you’ll see in the use cases below.+> COPA is everything that comes after the object expression. You can use COPA on any expression that evaluates to an object. COPA is not tied to object construction, but can be used anytime, anywhere in the objects lifeas many times as you want. 
 + 
 +//See more use cases below.//
  
 ==== Motivation ==== ==== Motivation ====
  
-The purpose of this feature is to lighten the effort of populating data structures, especially medium to large ones.+The purpose of this feature is to lighten the effort of creating, populating and sending data structures, from small to large ones.
  
-Ideally the solution should meet the following criteria:+The goal was to find a solution that meets the following criteria:
  
   * **Brief** - only mention the object once (less repetition)   * **Brief** - only mention the object once (less repetition)
Line 52: Line 58:
   * **Sparcity** - any property can be “skipped” (“skipped” properties may acquire a default value)   * **Sparcity** - any property can be “skipped” (“skipped” properties may acquire a default value)
   * **Simple** - does what you would expect without introducing any new concepts into the language   * **Simple** - does what you would expect without introducing any new concepts into the language
 +
 +> My focus is to find a **pragmatic** solution that is trivial to implement, and won’t impede futher development of the language.
 +
 +//If you have ever wanted to create, populate and send an object inside a function call, COPA is for you!//
  
 ===== Proposal ===== ===== Proposal =====
Line 57: Line 67:
 ==== Syntax ==== ==== Syntax ====
  
-The proposed syntax following an object expression ''%%$myObj%%'' | ''%%(new MyClass())%%'' is the object arrow operator ''%%->%%'' followed by a set of square brackets ''%%[…]%%'' containing a comma-separated list of property name equals ''%%=%%'' expressionA trailing comma ''%%,%%'' is permitted for the same reasons it's permitted in array literals and [[https://wiki.php.net/rfc/trailing-comma-function-calls|function calls (as of PHP 7.3)]]. The whole block is considered an expression that returns the object we started with ''%%eg$myObj%%''.+The proposed syntax consists of the object arrow operator followed by a set of square brackets containing a comma-separated list of assignments in the form of property name equals expression: 
 + 
 +<code> 
 +<object-expression> -> [ 
 +    `<property-name> <expression>`, 
 +    [repeat as needed], 
 +
 +</code> 
 +A trailing comma is permitted for the same reasons it's permitted in array literals and [[https://wiki.php.net/rfc/trailing-comma-function-calls|function calls (as of PHP 7.3)]]. 
 + 
 +The whole block is considered an expression that returns the object before the arrow. 
 + 
 +This syntax was chosen for its availability in the language. If we land on another syntax, I’m not married to this oneThe only criteria are that it doesn’t conflict with anything else, that it is trivial to implement, brief and feels ok.
  
 ==== Interpretation ==== ==== Interpretation ====
Line 63: Line 85:
 Each comma-separated assignment inside the brackets is executed as an assignment of the named property on the object preceding the block. If the property is defined and publicly accessible, it will simply be set, or possible throw a ''%%TypeError%%''. If there's no property with that name, or if it's protected or private, the magic method ''%%__set%%'' will be called just like you would expect. When used in an expression, **COPA** simply returns the object itself. Each comma-separated assignment inside the brackets is executed as an assignment of the named property on the object preceding the block. If the property is defined and publicly accessible, it will simply be set, or possible throw a ''%%TypeError%%''. If there's no property with that name, or if it's protected or private, the magic method ''%%__set%%'' will be called just like you would expect. When used in an expression, **COPA** simply returns the object itself.
  
 +If you replace COPA with single line assignments, you will always get the same result, f.ex.:
 +
 +<code php>
 +$foo->[
 +    a = 1,
 +    b = myfunc(),
 +    c = $foo->bar(),
 +];
 +
 +// The COPA above is identical to
 +$foo->a = 1;
 +$foo->b = myfunc();
 +$foo->c = $foo->bar();
 +</code>
 ==== Use cases ==== ==== Use cases ====
  
-=== DTOs - data transfer objects ===+=== Create and send struct ===
  
-Typical characteristics of DTOs:+<code php> 
 +// Instead of this:
  
-  * many properties +$myObj = new Foo; // 1Create struct-like object without constructor arguments
-  * properties may be optional, with default values +
-  * public visibility on properties, i.e. no desire for boilerplate code to create setters and getters for each one +
-  * desirability to create, populate and send in one go+
  
-== With current syntax ==+$myObj->1; // 2. Populate public properties 
 +$myObj->2; 
 +$myObj->3;
  
-<code php> +doTheFoo($myObj); // 3. Send or process
-class FooDto { +
-    public string $mane; +
-    public int $padme = 1; // Optional, with default +
-    public FooDto $hum; +
-}+
  
-$foo = new FooDto(); // Instantiating the object first +// Use COPA:
-$foo->mane = 'get'; // Setting the properties +
-// Skipping the $padme property which has a default value +
-$foo->hum = new FooDto(); // Creating a nested DTO +
-$foo->hum->mane = 'life'; // Populating the nested DTO+
  
-doTheFoo($foo); // Passing it to function+doTheFoo((new Foo)->[ 
 +    = 1, 
 +    b = 2, 
 +    c = 3, 
 +]);
 </code> </code>
-== With new COPA syntax ==+//No boilerplate needed.// 
 + 
 +=== Stop using arrays ===
  
 <code php> <code php>
-doTheFoo((new FooDto())->// Constructing and populating inline +// Instead of this: 
-    mane = 'get'+ 
-    hum (new FooDto())->// Even nested structures +doSomething([ 
-        mane = 'life', +    'a' => 1, // Anonymous array doesn't provide any help on parameter names 
-    ],+    'b=> 2// or types
 ]); ]);
-</code> 
-//Though the example is not a typical DTO, it represents the characteristics.// 
  
-=== Argument bags ===+// Use COPA:
  
-Argument bags are typically used when:+class Options { // Give the data a signature, a well-defined structure 
 +    public $a; 
 +    public $b; 
 +}
  
-  * many arguments needs to be passed to function +doSomething((new Options)->
-  * some arguments are optional +    = 1, // Parameter name and type checking 
-  * order of arguments is not important+    b = 2, 
 +]); 
 +</code> 
 +//If you often create, populate and send the same families of data structure, declaring those structures and using COPA makes it a breeze.//
  
-With the proposed new syntax we can **avoid using simple arrays** and instead get **autocomplete** and **type-checking** in the IDE with a syntax that smells of **named parameters**:+=== Nested COPA ===
  
-== With current syntax ==+COPA is not limited to a flat structure.
  
 <code php> <code php>
-class Foo +(new Foo)->
-    protected string $mane; +    om = 'get', 
-    protected int $padme; +    mane = 'a', 
-    protected string $hum; +    hum (new Foo)->
- +        mane = 'life', 
-    public function __construct(array $options+    ]
-        $this->mane = $options['mane']; +];
-        $this->padme $options['padme'] ?? 1; +
-        $this->hum $options['hum']; +
-    +
-+
- +
-$myFoo = new Foo(+
-    'mane='get', // Array syntax doesn't provide any help on parameter names +
-    'hum' => 'life'// or types +
-]);+
 </code> </code>
-== With new COPA syntax ==+=== Split options from services === 
 + 
 +Separate concerns and use composition. In this example, once you have instantiated Foo, the options are no longer writeable, even though the options were public properties.
  
 <code php> <code php>
-class FooOptions { // Separate concerns into an options class that handles optional and default values... +class FooOptions { 
-    public string $mane;+    public ?string $mane = null;
     public int $padme = 1; // Optional, with default     public int $padme = 1; // Optional, with default
-    public string $hum;+    public ?string $hum = null;
 } }
  
-class Foo { // And the main class that receives the options and handles some feature+class Foo {
     protected FooOptions $options;     protected FooOptions $options;
  
     public function __construct(FooOptions $options) {     public function __construct(FooOptions $options) {
-        $this->options = $options;+        // Do some validate here if you must, f.ex. checking for mandatory parameters 
 +        $this->options = clone $options;
     }     }
 } }
  
-$myFoo = new Foo((new FooOptions())->// Objects as argument bags (pseudo named parameters?) +$myFoo = new Foo((new FooOptions)->
-    mane = 'get', // Parameter name and type checking+    mane = 'get',
     hum = 'life',     hum = 'life',
 ]); ]);
 </code> </code>
-//The other alternative to an argument bag is usually constructor with many arguments, which is something that has been attempted to solve with RFCs arguing for automatic promotion of arguments to properties (f.ex. [[https://wiki.php.net/rfc/automatic_property_initialization|RFC: Automatic Property Initialization]], but which is probably also better left to the COPA argument bag example above.//+//If you can’t wait for “named parameters” and often resort to “parameter bags” this is a perfectly valid and saner alternative.//
  
 ==== Special cases ==== ==== Special cases ====
Line 173: Line 204:
 var_dump($myObj->bar); // int(30) var_dump($myObj->bar); // int(30)
 </code> </code>
-//As the assignments are carried out in order on the object, you can use the new value of a previous assigment in a following one.// +As the assignments are carried out in order on the object, you can use the new value of a previous assigment in a following one.
- +
-//There may be arguments equally for and against this behavior, but ultimately the simplicity of the design and implementation wins, in my opinion.//+
  
 === Exceptions === === Exceptions ===
Line 194: Line 223:
 } }
 </code> </code>
-Then the following two blocks behave identically:+Then the following examples behave identically:
  
 <code php> <code php>
-// With COPA+// With COPA
 try { try {
     $foo->[     $foo->[
Line 210: Line 240:
 </code> </code>
 <code php> <code php>
-// Without COPA+// Without COPA
 + 
 +try { 
 +    $foo->setA('a'
 +        ->setB(iThrow()) 
 +        ->setC('c'); 
 +} catch (\Throwable $e) { 
 +    var_dump($foo); 
 +
 + 
 +// OR 
 try { try {
     $foo->a = 'a';     $foo->a = 'a';
Line 219: Line 260:
 } }
 </code> </code>
-Result, ''%%a%%'' will be set, ''%%b%%'' and ''%%c%%'' will not:+The result in all cases is that ''%%a%%'' will be set, while ''%%b%%'' and ''%%c%%'' will not:
  
-<code php>+<code>
 object(Foo)#1 (3) { object(Foo)#1 (3) {
   ["a"]=>   ["a"]=>
Line 231: Line 272:
 } }
 </code> </code>
 +> COPA is **not** an atomic operation in the same way that method chaining isn’t.
 +
 ==== Out of scope / future scope ==== ==== Out of scope / future scope ====
  
-This section contains a tentative list of features that may not be implemented.+This section contains features that is not considered for implementation in version 1 of COPA but may be considered later.
  
-=== Can you do that===+=== You can’t do that ===
  
-The following examples show some things that is now possible using regular property accessor, but which will not also be supported with COPA:+The following examples show various things that are currently possible when using regular property accessor, though they won’t work inside a COPA block:
  
 <code php> <code php>
Line 255: Line 298:
 ]; ];
 </code> </code>
-//If anyone can show that any these features would be significantly desirable and simultaneously rather trivial to implement, let’s discuss.//+//These can be implemented in the future if there is a demand.//
  
-=== Nested COPA ===+=== Nested COPA on existing objects ===
  
-It might be nice to be able to populate a nested object in the same block, even if it has already been created:+The following syntax could be supported in the future:
  
 <code php> <code php>
 // This example, using current syntax... // This example, using current syntax...
 +
 $foo->a = 1; $foo->a = 1;
 $foo->b->c = 2; $foo->b->c = 2;
  
 // Could be written with COPA like this: // Could be written with COPA like this:
 +
 $foo->[ $foo->[
     a = 1,     a = 1,
Line 273: Line 318:
     ],     ],
 ]; ];
 +
 +// But for now you'll have to do this:
 +
 +$foo->[
 +    a = 1,
 +    b = $foo->b->[
 +        c = 2,
 +    ],
 +];
 +
 </code> </code>
-===== Why don't you just... =====+===== Backward Incompatible Changes =====
  
-What follows is a handful of alternative ways to populate an object with existing syntax, and some hints as to why they just doesn't cut it:+None.
  
-==== Vanilla style population ====+> **Note!** Array followed by square bracket causes syntax error in PHP 7.4. This new syntax is optional. If you don't use it, your code will continue to run.
  
-<code php> +===== Proposed PHP Version(s) =====
-class Foo { +
-    public int $bar; +
-    public int $baz; +
-}+
  
-$foo = new Foo(); +PHP 8.0
-$foo->bar = 1; +
-$foo->baz = 2;+
  
-doTheFoo($foo); // Cannot be done as an inline expression+===== Open Issues =====
  
-// Oh yeah? What if only need to set single property? +==== Alternative syntaxes ==== 
-doTheFoo((new Foo())->bar = 3); // Oopsfatal error: Can't use temporary expression in write context+ 
 +I’m going to suggest some alternative syntaxes, which we can vote on, provided their feasibility has been vetted by an experienced internals developer: 
 + 
 +=== Syntax A === 
 + 
 +This is the originally proposed one: 
 + 
 +<code php> 
 +$foo->
 +    = 1, 
 +    b = 2, 
 +    c = (new Foo)->
 +        a = 3, 
 +        b = 4, 
 +    ], 
 +];
 </code> </code>
-==== Applying a touch of magic ====+=== Syntax B === 
 + 
 +Since the [[https://wiki.php.net/rfc/deprecate_curly_braces_array_access|deprecation of curly brackets as array access in PHP 7.4]], that notation could be used to assign properties:
  
 <code php> <code php>
-/** +$foo{ 
- * @method self setBar(int $bar) // Use annotations +    a = 1, 
- * @method self setBaz(int $baz) // Duplicate the property signatures +    b = 2, 
- */ +    c = (new Foo){ 
-class Foo { +        a = 3, 
-    protected int $bar+        b = 4, 
-    protected int $baz;+    }, 
 +}
 +</code> 
 +> Going from deprecation in 7.4 to removal of support in 8.0 may is not unprecedented. Old code that has not been mended won’t silently do something spurious.
  
-    // This generic method could be injected using a trait +=== Syntax C ===
-    public function __call(string $method, array $params): self { +
-        if (strpos($method, 'get'=== 0) { +
-            $name substr($method, 3); +
-            $this->$name current($params); +
-        } +
-        return $this; +
-    } +
-}+
  
-doTheFoo((new Foo())  +No wrapper: 
-    ->setBar(1) + 
-    ->setBaz(2) +<code php> 
-);+$foo-> 
 +    a = 1, 
 +    b = 2, 
 +    c = (new Foo)-> 
 +        a = 3, 
 +        b = 4, 
 +    ;, 
 +;
 </code> </code>
-//Works, but requires some boilerplate code//+Nesting becomes awkward - how do we jump out again?
  
-==== Anonymous classes have some tricks up their sleeves! ====+> **Note!** This looks more like a chain of normal assignments, but that can be confusion since those normally return the value assigned, not the object itself. 
 + 
 +=== Syntax D === 
 + 
 +Repeating the arrow for familiarity with regular property assignment:
  
 <code php> <code php>
-class Foo { +$foo 
-    public int $bar; +    ->a = 1, 
-    public int $baz; +    ->b = 2, 
-    public Foo $sub+    ->c = (new Foo
-}+        ->a = 3, 
 +        ->b = 4, 
 +    ;
 +
 +</code> 
 +Same issues as previous.
  
-doTheFoo(new class extends Foo { +=== Syntax E ===
-    public int $bar 1; // Assigning values inline is now possible without creating setters! +
-    public int $baz 2; // But I have to repeat their signature, which is annoying+
  
-    public function __construct() { +Like the original but with normal brackets instead of square ones: 
-        // And if I need expressions, I have to use the constructor + 
-        $this->sub  new class extends Foo { +<code php> 
-            public int $bar = 3; +$foo->
-            public int $baz = 4; +    a 1, 
-        }; +    b = 2, 
-   } +    c = (new Foo)->( 
-});+        = 3, 
 +        = 4, 
 +    ), 
 +);
 </code> </code>
-//Pretty ugly, I’m afraid…//+=== Syntax F ===
  
-==== Lambda expression ====+**WITH** syntax
  
 <code php> <code php>
-class Foo +myObj.with 
-    public int $bar; +     foo = 10 
-    public int $baz;+     bar = foo + 20
 } }
- 
-doTheFoo((function(){ 
-   $foo = new Foo(); 
-   $foo->bar = 1; 
-   $foo->baz = 2; 
-   return $foo; 
-})());  
 </code> </code>
-//Pretty good, if you can get those brackets straight… until you need to use values from the outside scope :-(//+If this is the preferred syntax it will require a new RFC.
  
-===== Anti-proposal =====+===== Rejected Features =====
  
-This proposal is related to previous RFCs and shares motivation with them. However, though **COPA** claims to be in the same family, here are some disclaimers:+Some suggested features have been rejected due to the fact that COPA aims to be pragmatic, with a trivial implementation and without introducing any new concepts to avoid a combinatorial explosion of complexities in the future.
  
-==== COPA is NOT json ====+==== Mandatory properties ====
  
-This is not a way to write object literals using JavaScript Object Notation [[https://wiki.php.net/rfc/objectarrayliterals|(RFC: First-Class Object and Array Literals)]]. It's similar to an array literal, but with each key actually corresponding to a defined property of the object. We don't want to quote the property names as there's no advantage, only added overhead. The equals sign is used straightforwardly to denote assignment. Square brackets have been chosen instead of curly ones because the latter already has an interpretation when following the object arrow, namely to create an expression which will return a property or method name.+Some have voiced criticism that COPA is of little use without also enforcing mandatory properties to be set.
  
-==== COPA is NOT object initializer ====+**Rowan Tommins:**
  
-You could call it **//pseudo// object literal** notation because we're not dictating the //actual// inner state of the object, we're merely populating properties after construction. But this //does// allow for a ''%%"literal syntax for creating an object and initializing properties"%%'' [[https://wiki.php.net/rfc/object-initializer|(RFC: Object Initializer)]]giving you benefits very similar to object literals in simplepragmatic way.+> It seems pretty rare that an object would have no mandatory properties, so saying “if you have mandatory propertyCOPA is not for you” is ruling out a lot of uses.
  
-==== COPA is NOT named parameters ====+**Michał Brzuchalski:**
  
-Though on the wish list since 2013, named parameters [[https://wiki.php.net/rfc/named_params|(RFC: Named Parameters)]] have proven to be tough nut to crack. But with this RFC you will be able to create parameter objects that may give you benefits very similar to named parameters [[https://wiki.php.net/rfc/simplified_named_params|(RFC: Simplified Named Arguments)]] when you pass it to function that expects it.+> This helps to avoid bugs where property is added to the class but forgot to be assigned it a value in all cases where the class is instantiated and initialized
  
-===== Backward Incompatible Changes =====+Mandatory properties are typed properties without a default value. They are in the uninitialized state until they are assigned a value. It has been suggested that an exception should be thrown at the end of the constructor if any property is still uninitialized, but this idea has not yet caught on. COPA doesn’t have any obvious way of enforcing mandatory properties.
  
-NoneArray followed by square bracket causes syntax error in PHP 7.4. This new syntax is optional. If you don't use it, your code will continue to run.+> COPA won’t support this since COPA doesn’t introduce any new concepts or complexitiesThe lack of this feature is not a limitation of COPA when compared to current functionality.
  
-===== Proposed PHP Version(s) =====+//For now you must continue to write your own validation code to be carried out at the appropriate “point of no return”.//
  
-PHP 8.0+==== Atomic operations ====
  
-===== Open Issues =====+It’s also been suggested that assigning multiple values using COPA should be an atomic operation that either succeeds or fails in its entirety (i.e. like a “transaction”).
  
-==== Alternative syntaxes ====+Though that sounds cool, this is an edge case that won’t have any significant impact. If you were planning to resume gracefully with an incomplete object you should probably reconsider your goals in life.
  
-Some alternative syntaxes for COPA has been suggested, but I’m not convinced they can be implemented without hassle:+> **Note!** Chaining method calls is not an atomic operation either. The cost/benefit of implementing “transaction” and “rollback” behavior is negative.
  
-<code php> +===== Vote =====
-$foo new Foo()[ +
-   property1 "hello", +
-   property2 5, +
- ]; +
-</code> +
-For some reason it’s not possible to do this:+
  
-<code php> +Voting starts 2020-03-31 and ends 2020-04-13.
-new Foo()->doSomething(); // syntax error, unexpected '->' +
-</code> +
-It’s necessary to wrap the instantiation in brackets:+
  
-<code php> +The primary vote of whether or not to accept this RFC requires a 2/3 majority.
-(new Foo())->doSomething(); // Ok +
-</code> +
-Which is why I think it will be necessary in my proposal as well.+
  
-Furthermore, a variable or object directly followed by square brackets usually imply array access on it. That syntax would conflict with COPA.+<doodle title="Would you like to add support for COPA?" auth="jgivoni" voteType="single" closed="true"> 
 +   * Yes 
 +   * No 
 +</doodle>
  
 +\\
  
-----+<doodle title="If you voted no, what was the main reason?" auth="jgivoni" voteType="single" closed="true"> 
 +   * I voted yes! 
 +   * I don’t find the feature useful 
 +   * I don’t like the syntax 
 +   * I prefer a more comprehensive solution to this problem 
 +   * I prefer a narrower solution to this problem 
 +   * This breaks backwards compatibility 
 +   * This will have negative implications for future language evolution 
 +   * This will be a nightmare to implement and maintain 
 +   * I prefer not to say 
 +</doodle>
  
-//Unless someone can convince me that it’s trivial to implement another syntax that looks even better, my stance is that the people who are going to vote no on this because they don’t find the feature useful are not gonna change their mind if the syntax changes, and the people who find this feature useful will prefer rapid adaptation over solving implementation issues.// +\\
- +
-===== Proposed Voting Choices ===== +
- +
-The primary vote of whether or not to accept this RFC requires a 2/3 majority.+
  
-There may be a secondary “vote” directed at no-voters, where you’ll be asked the primary reason for voting “No”. This will help understand what the obstacles arewhen studying this RFC in the future, should anyone be tempted to have another shot at object literals et. al.+<doodle title="If you did not like the proposed syntaxwhich alternative would you prefer?" auth="jgivoni" voteType="single" closed="true"> 
 +   * A (the proposed one) 
 +   * B 
 +   * C 
 +   * D 
 +   * E 
 +   * F 
 +   * Irrelevant 
 +</doodle>
  
 ===== Patches and Tests ===== ===== Patches and Tests =====
rfc/compact-object-property-assignment.1584413600.txt.gz · Last modified: 2020/03/17 02:53 by jgivoni