rfc:chaining_comparison

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:chaining_comparison [2016/12/12 23:35] – bp1222rfc:chaining_comparison [2021/03/27 14:58] (current) – Move to inactive ilutov
Line 1: Line 1:
 ====== PHP RFC: Chaining Comparison ====== ====== PHP RFC: Chaining Comparison ======
-  * Version: 0.1 +  * Version: 0.2 
-  * Date: 2016-12-08 +  * Date: 2016-12-13 
-  * Author: David Walker (dave@mudsite.com), Richard Fussenegger (php@fleshgrinder.com) +  * Author: David Walker (dave@mudsite.com) 
-  * Status: Draft+  * Author: Richard Fussenegger (php@fleshgrinder.com) 
 +  * Status: Inactive
   * First Published at: http://wiki.php.net/rfc/chaining_comparison   * First Published at: http://wiki.php.net/rfc/chaining_comparison
  
 ===== Introduction ===== ===== Introduction =====
-The point of this RFC is to allow the chaining together of comparison and equality operations [''==, !=, !==, ===, <, <=, >, >=''] to allow arbitrary comparisons.  The initial request that spawned this RFC was for interval checking.+This RFC proposes a syntax change to allow arbitrary chaining together of comparison and equality operations ''[==, !=, !==, ===, <, <=, >, >=]''.  The initial request that spawned this RFC[1] was initially only for interval checking.  Discussion on the thread expanded the scope of the request to go from strictly interval checking to allowing more arbitrary number of comparisons.  It evolved from there to expand to a majority of the comparison operations.  The primary benefit to this proposal would be to make for more readable code when doing numerous comparisons between a single variable.
  
-Today such comparisons must be written as such: 
 <file php> <file php>
 <?php <?php
 $a = 10; $a = 10;
  
 +/*
 + * The initial request of this proposal was to change the following syntax
 + */
 if (0 < $a && $a < 100) { if (0 < $a && $a < 100) {
     echo "Value is between 0 and 100\n";     echo "Value is between 0 and 100\n";
 } }
-</file> 
- 
-The proposal of this RFC is to allow new syntax as follows: 
-<file php> 
-<?php 
-$a = 10; 
  
 +/*
 + * To be functionally equivalent to this syntax
 + */
 if (0 < $a < 100) { if (0 < $a < 100) {
     echo "Value is between 0 and 100\n";     echo "Value is between 0 and 100\n";
Line 30: Line 30:
  
 ===== Proposal ===== ===== Proposal =====
-==== Comparison Chaining ==== +Proposals herein will contain a dump of relevant AST (php-ast) nodes and OPCodes (vld) to better visualize the compilation, and execution.
-The proposal creates a new AST operation type ''ZEND_AST_COMPARE_OP'' which will be compiled in a left-precedence required manor.  In doing this compilation we introduce a new means of emitting an operation, by noting where a ''JMPZ_EX'' may need to exist, depending if we are continuing the compare chain.  This will shift operations that may have been emitted by compiling the right side of this AST compare to allow jumping over them if the left side of the operation is evaluated to false.+
  
-The proposal also changes the precedence of the equality, and comparison, operations to being left recursive.  This is required, since if the left node of this operation is true, and it itself is a comparison operation, it should return the right node (for less than) to be used in the next comparison op.  Example: +==== Comparison Chaining ==== 
 +The proposal creates a new AST operation type ''ZEND_AST_COMPARE_OP'' which will be compiled in a left-recursive manor.  
  
 <file php> <file php>
Line 67: Line 67:
  *    2        INIT_FCALL                                               'var_dump'  *    2        INIT_FCALL                                               'var_dump'
  *    3        IS_SMALLER                                       ~4      !0, 5  *    3        IS_SMALLER                                       ~4      !0, 5
- *    4      > JMPZ_EX                                          ~6      ~4, ->7+ *    4      > JMPZ_EX                                          ~4      ~4, ->7
  *    5    >   POST_INC                                         ~5      !1  *    5    >   POST_INC                                         ~5      !1
- *    6    >   IS_SMALLER                                       ~6      ~4, ~5 + *    6    >   IS_SMALLER                                       ~4      ~4, ~5 
- *    7    >   SEND_VAL                                                 ~6+ *    7    >   SEND_VAL                                                 ~4
  *    8        DO_ICALL                                                   *    8        DO_ICALL                                                 
  */  */
Line 112: Line 112:
  *  *
  *   2        INIT_FCALL                                               'var_dump'  *   2        INIT_FCALL                                               'var_dump'
- *   3        IS_EQUAL                                         ~4      !0, 1 + *   3        IS_EQUAL                                         ~2      !0, 1 
- *   4      > JMPZ_EX                                          ~5      ~4, ->6 + *   4      > JMPZ_EX                                          ~2      ~2, ->6 
- *   5    >   IS_IDENTICAL                                     ~5      ~4, <true> + *   5    >   IS_IDENTICAL                                     ~2      ~2, <true> 
- *   6    >   SEND_VAL                                                 ~5+ *   6    >   SEND_VAL                                                 ~2
  */  */
 </file> </file>
  
-==== Combined Uses (Open Discussion Topic)==== +==== False Short Circuiting ==== 
-One of the concerns raised in the implementation of this feature would be how to handle expressions such as ''1 < 2 == 3 < 4''.  This is valid syntax in PHP 7.1 and before that essentially checks if true == true.  The current implementation respects the current syntax, by evaluating the comparison operators before equality operators.  Combining the two chaining methods we can write a new expression like this example: +In doing this compilation we can ensure short cutting of righter operations if the left sides have evaluated to false.  To accomplish this we introduce a new means of emitting an operation, by noting where a ''JMPZ_EX'' may need to exist (see implementations for ''zend_emit_op_at'').  This will shift operations that may have been emitted by compiling the right side of this AST compare to allow jumping over them if the left side of the operation is evaluated to false.  I believe this means is necessary because we can't just shortcut if the left operation is false, ''false < $a++'' should still evaluate the right part of the expression.  We should only inject the JMPZ_EX ops, IF, the left child is a chained ''ZEND_AST_COMPARE_OP''.  The proposal also changes the associativity of the equality, and comparison, operations to being left associative.
- +
-<file php> +
-<?php +
-$a = 1; +
-$b = 4; +
-$c = 10; +
- +
-var_dump($a < 2 == 3 < $b < 5 == 20 > $c); // bool(true) +
- +
-/* +
- * AST Dump +
- * +
- *  3: AST_CALL +
- *       expr: AST_NAME +
- *           flags: NAME_NOT_FQ (1) +
- *           name: "var_dump" +
- *       args: AST_ARG_LIST +
- *           0: AST_COMPARE_OP +
- *               flags: COMPARE_IS_EQUAL (17) +
- *               left: AST_COMPARE_OP +
- *                   flags: COMPARE_IS_EQUAL (17) +
- *                   left: AST_COMPARE_OP +
- *                       flags: COMPARE_IS_SMALLER (19) +
- *                       left: AST_VAR +
- *                           name: "a" +
- *                       right: 2 +
- *                   right: AST_COMPARE_OP +
- *                       flags: COMPARE_IS_SMALLER (19) +
- *                       left: AST_COMPARE_OP +
- *                           flags: COMPARE_IS_SMALLER (19) +
- *                           left: 3 +
- *                           right: AST_VAR +
- *                               name: "b" +
- *                       right: 5 +
- *               right: AST_COMPARE_OP +
- *                   flags: COMPARE_IS_SMALLER (19) +
- *                   left: 20 +
- *                   right: AST_VAR +
- *                       name: "c" +
- */ +
- +
-/* +
- * OPCodes +
- * +
- *   3        INIT_FCALL                                               'var_dump' +
- *   4        IS_SMALLER                                       ~6      !0, 2 +
- *   5      > JMPZ_EX                                          ~9      ~6, ->10 +
- *   6    >   IS_SMALLER                                       ~7      3, !1 +
- *   7      > JMPZ_EX                                          ~8      ~7, ->9 +
- *   8    >   IS_SMALLER                                       ~8      ~7, 5 +
- *   9    >   IS_EQUAL                                         ~9      ~6, ~8 +
- *  10    > > JMPZ_EX                                          ~11     ~9, ->13 +
- *  11    >   IS_SMALLER                                       ~10     !2, 20 +
- *  12        IS_EQUAL                                         ~11     ~9, ~10 +
- *  13    >   SEND_VAL                                                 ~11 +
- *  14        DO_ICALL                                                  +
- */ +
-</file>+
  
 ===== Backward Incompatible Changes ===== ===== Backward Incompatible Changes =====
-No BC Breaking changes expected (see: Open Issues)+BC Breaking changes expected depending on open-issue answers
  
 ===== Proposed PHP Version(s) ===== ===== Proposed PHP Version(s) =====
Line 188: Line 130:
 ===== RFC Impact ===== ===== RFC Impact =====
 ==== To Opcache ==== ==== To Opcache ====
-Yes, we're adding new JMPZ_EX codes when chaining to ensure false values correctly jump over any pre/post inc/dev ops from eval.+I'm unsure; we're adding new op-codes and/or order of opcodes, but are not introducing any new codes
  
 ===== Open Issues ===== ===== Open Issues =====
 ====Should equality and comparison expressions be treated as same precedence?==== ====Should equality and comparison expressions be treated as same precedence?====
-This is a tough question.  It's essentially asking if an expression like ''1 < 2 == 3 < 4'' should be evaluated as ''(1 < 2) == (3 < 4)'' OR ''(1 < 2) && (2 == 3) && (3 < 4)''.+This is harder of a question that it seems.  What we are asking is how should we parse a seemingly simple expression: 
 +''1 < 2 == 3 < 4''
  
-Why is this even a question?  A majority of languages would evaluate the expression with the former interpretation.  However, Python, and WolframAlpha, appear to do the latter.  The latter does enforce more of a "chaining" of these comparisons.  During discussion, this question should be addressed, as the underlying code to support either means is very minor.+Why is this even a question, much less a challenging one?  Well, a seemingly majority of languages ''[C[2], C++[3], Java[4], Ruby[5], Perl[6]]'' all would tell you that the expression would evaluate to true.  However some, like ''Python[7]'', would evaluate that expression to false.  Some, like ''[Numbers, LibreOffice]'' will raise a syntax error, or give awkward answers.  The question we have is which way should PHP go with the evaluation of this expression?  Clearly we can ascertain that the true-evaluating languages have the precedence of the less-than operator more imporatant than that of the equality, so they check if true == true.  Whereas the false-evaluating languages treat comparisons and equality with the same precedence.  As such they compare 1 less than 2, 2 is-equal 3.  The latter group are apparently more strictly typed and won't compare bools to numbers, but even there we can see the precedence is equal, as it's comparing the result of the first expression into the next ''(1 < 2) == 3''
  
-If we do decide to proceed with the Python/Wolfram interpretation of a very chained left-to-right this would introduce a BC break as PHP 7.1 and prior do support the syntax ''expr1 < expr2 == expr3 < expr4'' to be interpreted as ''(expr1 < expr2) == (expr3 < expr4)'' where boolean == boolean.  If implemented as a strict left-to-right comparison, the new interpretation would be ''(expr1 < expr2) && (expr2 == expr3) && (expr3 < expr4)'' where boolean && boolean && boolean.  You can be explicit with parenthesis using strict-chaining to get the same result.+It is important to point out that the example syntax is currently valid in PHP 7.1.  PHP 7.1 currently has a C-like precedence where ''[<, <=, >, >=]'' are a higher precedence than ''[==, !=, ===, !==]''[8].  Below are expressions and their return values in PHP 7.1, and with the two potential methods of evaluating that expression.
  
 <file php> <file php>
Line 220: Line 163:
 var_dump((1 < 2) == (3 < 4) == (5 < 6)) // bool(true) var_dump((1 < 2) == (3 < 4) == (5 < 6)) // bool(true)
 </file> </file>
 +
 +==== Right Recursion ====
 +Another syntax difference that could be BC problematic is with right-recursion of the chained expression.  Currently PHP will evaluate right recursive single expression comparisons.  The proposed feature would raise a compile time error doing this.  The question is should it, or should we permit right-recursive chaining? The test case we can look at:
 +
 +<file php>
 +<?php
 +var_dump(1 < (2 < 3));
 +var_dump(1 < 2 == 3);
 +var_dump(1 < 2 == 3 == 4);
 +var_dump(1 < 2 == (3 == 4));
 +</file>
 +
 +We will go over how PHP 7.1 currently would evaluate each, and then how a right-recursive chain would pan out.
 +<file php>
 +<?php
 +var_dump(1 < (2 < 3));
 +/*
 + * 1 < (2 < 3) := 1 < true := false
 + */
 +
 +var_dump(1 < 2 == 3);
 +/*
 + * (1 < 2) == 3 := true == 3 := true
 + */
 +
 +var_dump(1 < 2 == 3 == 4);
 +/*
 + * Parse Error, unexpected ==
 + */
 +
 +var_dump(1 < 2 == (3 == 4));
 +/*
 + * (1 < 2) == (3 == 4) := true == false := false
 + */
 +</file>
 +
 +The current proposal (implemented) evaluation method.  You'll notice that we do permit right-recursion for equality operations.  This is due to the fact that equality operations will evaluate against boolean, or boolean-converted values.  Since you don't really care what the left-node of the right-recursive side is, you only care if the right side evaluates to true or not.
 +<file php>
 +<?php
 +var_dump(1 < (2 < 3));
 +/*
 + * Parse Error: No right recursion
 + */
 +
 +var_dump(1 < 2 == 3);
 +/*
 + * (1 < 2) == 3 := true == 3 := true
 + */
 +
 +var_dump(1 < 2 == 3 == 4);
 +/*
 + * ((1 < 2) == 3) == 4 := (true == 3) == 4 := true == 4 := true
 + */
 +
 +var_dump(1 < 2 == (3 == 4));
 +/*
 + * (1 < 2) == (3 == 4) := true == false := false
 + */
 +</file>
 +
 +If however we permitted right recursive comparison operations we would evaluate as such:
 +<file php>
 +<?php
 +var_dump(1 < (2 < 3));
 +/*
 + * 1 < (2 < 3) := 1 && (2 < 3) && (1 < 2) := true && true && true := true
 + */
 +
 +var_dump(1 < 2 == 3);
 +/*
 + * (1 < 2) == 3 := true == 3 := true
 + */
 +
 +var_dump(1 < 2 == 3 == 4);
 +/*
 + * ((1 < 2) == 3) == 4 := (true == 3) == 4 := true == 4 := true
 + */
 +
 +var_dump(1 < 2 == (3 == 4));
 +/*
 + * (1 < 2) == (3 == 4) := true == false := false
 + */
 +</file>
 +
 +If the first example in this last one looks a little odd, it's because it is.  We do design for short-cutting of a long expression when a fault is found to prevent further execution much like you have in ''if()'' statements.  However, we do process in a left-to-right manor.  So the first thing would require us to ensure the left most side evaluates to true, and if it wasn't ''1'' but rather ''$a++'', we'd want to ensure to get that left nodes potential opcodes to execute before comparing the right hand side.  Since we are chaining, we'd want to evaluate the right, then return the left node of it to be evaluated against the top's left node.  This, odd syntax is why I didn't implement a right-recursive chaining of comparison operations.
 +
 +Although allow right-recursion of equality operations does itself introduce some slightly odd syntax like:
 +<file php>
 +<?php
 +/*
 + * Right chained comparison syntax
 + */
 +var_dump(1 < (2 == 2)); // bool(false)
 +
 +/*
 + * Is Functionally identical to PHP 7.1's allowed syntax
 + */
 +var_dump(1 < (2 <= 2)); // bool(false)
 +</file>
 +Since we don't chain together the right/left node of an equality operator, this is functionally identical to PHP 7.1's allowed syntax.  We could, for equality operations denote if they were in-fact a right node-continuation of a chain, thus would allow them to evaluate to either the left node, or false.
 +
 +
 +As we can see right-recursive comparison operations do have numerous caveats and oddities.  For these reasons we didn't implement it, and generally are on the side of forbidding right-recursive comparison operations.
  
 ===== Unaffected PHP Functionality ===== ===== Unaffected PHP Functionality =====
 Does not alter the operation of the comparison Spaceship [<=>] operator. Does not alter the operation of the comparison Spaceship [<=>] operator.
- 
-===== Future Scope ===== 
  
 ===== Proposed Voting Choices ===== ===== Proposed Voting Choices =====
Line 230: Line 274:
  
 ===== Patches and Tests ===== ===== Patches and Tests =====
-Implementation #1: comparisons evaluated before equality: https://github.com/php/php-src/compare/master...bp1222:multi-compare +Working Implementation: comparisons evaluated before equality: https://github.com/php/php-src/compare/master...bp1222:multi-compare
- +
-Implementation #2: comparisons and equality evaluated together: https://github.com/php/php-src/compare/master...bp1222:multi-compare-equal-prec+
  
 Will need eyes of those more familiar with AST/VM to review. Will need eyes of those more familiar with AST/VM to review.
- 
-For changes affecting the core language, you should also provide a patch for the language specification. 
  
 ===== Implementation ===== ===== Implementation =====
Line 242: Line 282:
  
 ===== References ===== ===== References =====
-Initial idea on Internals: http://marc.info/?l=php-internals&m=147846422102802&w=2+  * [1] - [[http://marc.info/?l=php-internals&m=147846422102802&w=2|Initial idea on Internals]] 
 +  * [2] - [[https://www.gnu.org/software/gnu-c-manual/gnu-c-manual.html#Operator-Precedence|Precedence in C]] 
 +  * [3] - [[https://msdn.microsoft.com/en-us/library/126fe14k.aspx|Precedence in C++]] 
 +  * [4] - [[http://introcs.cs.princeton.edu/java/11precedence/|Precedence in Java]] 
 +  * [5] - [[https://ruby-doc.org/core-2.2.0/doc/syntax/precedence_rdoc.html|Precedence in Ruby]] 
 +  * [6] - [[http://perldoc.perl.org/perlop.html#Operator-Precedence-and-Associativity|Precedence in Perl]] 
 +  * [7] - [[https://docs.python.org/2/reference/expressions.html#operator-precedence|Precedence in Python]] 
 +  * [8] - [[http://php.net/manual/fa/language.operators.precedence.php|Precedence in PHP]] 
  
 ===== Rejected Features ===== ===== Rejected Features =====
 Keep this updated with features that were discussed on the mail lists. Keep this updated with features that were discussed on the mail lists.
rfc/chaining_comparison.1481585737.txt.gz · Last modified: 2017/09/22 13:28 (external edit)

Page Tools