Differences

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

--- rfc:random_extension_improvement [2022/06/20 16:06] – remove SerializableEngine zeriyoshi
+++ rfc:random_extension_improvement [2022/07/22 09:55] (current) – reword two issues as the problem (VS the proposed solution) guilliamxavier
@@ Line 1: / Line 1: @@
 ====== PHP RFC: Random Extension Improvement ======
-  * Version: 1.4
+  * Version: 1.5.4
   * Date: 2022-06-16
   * Author: Go Kudo <g-kudo@colopl.co.jp> <zeriyoshi@gmail.com>
-  * Status: Under Discussion
+  * Status: Implemented
   * First Published at: http://wiki.php.net/rfc/rng_extension
 ===== Introduction =====
-There are several issues with RFC: Random Extension 5.x that are already in the voting phase due to lack of proper discussion.
+Several issues with [[rfc:rng_extension|the "Random Extension 5.x" RFC]] were only raised after voting started (due to lack of proper discussion):
-Key issues include:
 ==== Engine implementations are not final ====
-The Random Extension has classes that are natively implemented as RNG engines, but they are not marked as final.
+The Random Extension has classes that are natively implemented as RNG engines, but they are not marked as <php>final</php>.
 This allows classes to be created that inherit from native classes, but as stated in the previous RFC, user-implemented engines are inferior to native classes in terms of execution efficiency. This is true even for inheritance without method overrides, which often leads to confusion.
-The extension already provides a Random\Engine interface with a single generate(): string method.
+The extension already provides a ''Random\Engine'' interface with a single ''generate(): string'' method.
 Even if the native classes are made final, it is easy for the user to create an alternative class using delegates.
 This is clearly an API design error, and the native implementations of the interface should be marked as final.
-==== Randomizer lacks array_rand() replacement method ====
+==== Random\SerializableEngine is not useful ====
-array_rand() uses RNG internally, but there is no alternative method in Randomizer.
+This interface is a remnant from when the <php>Serializable</php> interface was still useful. This interface is no longer needed in PHP, as serializability is now determined by the existence of a magic method.
-As per the previous RFC, this was the intent, but upon further investigation, array_rand() is used by many packages and should probably be drop-in replaceable.
-However, the name of the method may need to be reconsidered. How about Randomizer::pickArrayKey(array $array, int $num = 1): int|string|array or something similar?
+For this reason, remove ''Random\SerializableEngine''.
-==== "string" means a binary ====
+This means that an Engine that implements SerializableEngine will no longer implement it.
+However, serializability is determined by the implementation of the magic method in current PHP, so it has no effect.
-In PHP, "string" means a binary. This is often a problem when using multibyte characters. For example, str_shuffle on a Japanese (UTF-8) string will give messed up results.
+==== Random\Engine\CombinedLCG is low quality ====
-Therefore, it may be better to change the alternative method of str_shuffle(), Randomizer::shuffleString(), to Randomizer::shuffleBytes(). This is a more appropriate name.
+The newly added ''Random\Engine\CombinedLCG'' is only for use with PHP's <php>lcg_value()</php> function.
+However, this algorithm is very classical and the quality of the output random numbers is at the lowest level.
-==== PCG64 is ambiguous ====
+In order to preserve the implementation of the lcg_value() function, the internal implementation is retained, but the implementation as a class is being dropped to prevent users from unintentionally using it.
-The Random Extension has implemented PCG64 as a new PRNG, but it is officially one variant of PCG64, oneseq-128-xsl-rr-64, and may need to be made more easily understood.
+==== There is no equivalent of array_rand() ====
-Therefore, change the class name from Random\Engine\PCG64 to Random\Engine\PCGOneseq128XslRr64.
+<php>array_rand()</php> uses RNG internally, but there is no alternative method in ''Randomizer''.
+As per the previous RFC, this was the intent, but upon further investigation, array_rand() is used by many packages and should probably be drop-in replaceable.
-==== Mersenne Twister is ambiguous ====
+So add a method <php>Randomizer::pickArrayKeys(array $array, int $num): array</php>.
-When PCG makes the choice to reflect a more detailed implementation in the name, Random\Engine\MersenneTwister is also an issue at the same time.
+It may look incompatible with <php>array_rand(array $array, int $num = 1): int|string|array</php>, but you can get completely consistent results by doing the following:
-Random\Engine\MersenneTwister is precisely an implementation of the MT19937 algorithm.
+<code php>
+$array = ['foo', 'bar', 'baz'];
-MT19937 also has an algorithm called MT19937_64 that generates 64-bit values. If for some reason MT19937_64 were to be implemented, the fact that the name Mersenne Twister is already in use could be problematic.
+// Before:
+mt_srand(1234, MT_RAND_PHP);
+$single = array_rand($array); // (int) 0
+$multiple = array_rand($array, 2); // (array) [1, 2]
-==== PCG is not so famous ====
+// After:
+$engine = new Random\Engine\Mt19937(1234, MT_RAND_PHP);
+$randomizer = new Random\Randomizer($engine);
+$single = $randomizer->pickArrayKeys($array, 1)[0]; // (int) 0
+ // or: [$single] = $randomizer->pickArrayKeys($array, 1);
+$multiple = $randomizer->pickArrayKeys($array, 2); // (array) [1, 2]
+</code>
-PCG is a very good algorithm, boasting great randomness and performance.
+==== "string" means a binary ====
-However, I think its name recognition is in some ways inferior to Vigna's RNG, which started with Xorshift.
+In PHP, "string" means a binary. This is often a problem when using multibyte characters. For example, <php>str_shuffle()</php> on a Japanese (UTF-8) string will give messed up results.
-Therefore, I reimplement Xoshiro256**, which was previously mentioned as a candidate, and create the class Random\Engine\Xoshiro256StarStar. This will avoid the problem of using the old MT19937 since it is not familiar with PCG.
+Therefore, it may be better to change the alternative method of str_shuffle(), ''Randomizer::shuffle**String**()'', to ''Randomizer::shuffle**Bytes**()''. This is a more appropriate name.
-==== CombinedLCG is outdated ====
+==== Engine classnames are not precise ====
-The newly added Random\Engine\CombinedLCG is only for use with PHP's lcg_value() function.
+To clearly identify the implemented algorithm, the PCG64 and MersenneTwister engines should be renamed to their canonical upstream name:
-However, this algorithm is very classical and the quality of the output random numbers is at the lowest level.
-In order to preserve the implementation of the lcg_value() function, the internal implementation is retained, but the implementation as a class is being dropped to prevent users from unintentionally using it.
+  * ''Random\Engine\**PCG64**'' → ''Random\Engine\**PcgOneseq128XslRr64**''
+  * ''Random\Engine\**MersenneTwister**'' → ''Random\Engine\**Mt19937**''
-==== Random\SerializableEngine is outdated ====
+==== PCG is not so famous ====
-This interface is a remnant from when the Serializable interface was still useful. This interface is no longer needed in PHP, as serializability is now determined by the existence of a magic method.
+PCG is a very good algorithm, boasting great randomness and performance.
+However, I think its name recognition is in some ways inferior to Vigna's RNG, which started with Xorshift.
-For this reason, remove RandomEngine\SerializableEngine.
+Therefore, I reimplement %%Xoshiro256**%%, which was previously mentioned as a candidate, and create the class ''Random\Engine\Xoshiro256StarStar''. This will avoid the problem of using the old MT19937 since it is not familiar with PCG.
 ===== Proposal =====
@@ Line 76: / Line 87: @@
 ==== Engine implementations are not final ====
-<doodle title="Engine implementations to final" auth="zeriyoshi" voteType="single" closed="true">
+<doodle title="Make all implemented engines final?" auth="zeriyoshi" voteType="single" closed="true">
    * Yes
    * No
 </doodle>
-==== Randomizer lacks array_rand() replacement method ====
+==== Random\SerializableEngine is not useful ====
-<doodle title="Add Random\Randomizer::pickArrayKey(array $array, int $num = 1): int|string|array" auth="zeriyoshi" voteType="single" closed="true">
+<doodle title="Remove the SerializableEngine interface?" auth="zeriyoshi" voteType="single" closed="true">
    * Yes
    * No
 </doodle>
-==== "string" means a binary ====
+==== Random\Engine\CombinedLCG is low quality ====
-<doodle title="Rename Random\Randomizer::shuffleString() to Random\Randomizer::shuffleBytes()" auth="zeriyoshi" voteType="single" closed="true">
+<doodle title="Remove the CombinedLCG class?" auth="zeriyoshi" voteType="single" closed="true">
    * Yes
    * No
 </doodle>
-==== PCG64 is ambiguous ====
+==== There is no equivalent of array_rand() ====
-<doodle title="Rename Random\Engine\PCG64 to Random\Engine\PCGOneseq128XslRr64" auth="zeriyoshi" voteType="single" closed="true">
+<doodle title="Add the pickArrayKeys() method to the Randomizer?" auth="zeriyoshi" voteType="single" closed="true">
    * Yes
    * No
 </doodle>
-==== Mersenne Twister is ambiguous ====
+==== "string" means a binary ====
-<doodle title="Rename Random\Engine\MersenneTwister to Random\Engine\MT19937" auth="zeriyoshi" voteType="single" closed="true">
+<doodle title="Rename Randomizer::shuffleString() to Randomizer::shuffleBytes()?" auth="zeriyoshi" voteType="single" closed="true">
    * Yes
    * No
 </doodle>
-==== PCG is not so famous ====
+==== Engine classnames are not precise ====
-<doodle title="Implement Random\Engine\Xoshiro256StarStar" auth="zeriyoshi" voteType="single" closed="true">
+<doodle title="Rename PCG64 and MersenneTwister?" auth="zeriyoshi" voteType="single" closed="true">
    * Yes
    * No
 </doodle>
-==== CombinedLCG is outdated ====
+==== PCG is not so famous ====
-<doodle title="Drop Random\Engine\CombinedLCG" auth="zeriyoshi" voteType="single" closed="true">
-   * Yes
-   * No
-</doodle>
-==== Random\SerializableEngine is outdated ====
-<doodle title="Remove Random\SerializableEngine" auth="zeriyoshi" voteType="single" closed="true">
+<doodle title="Add the Xoshiro256StarStar engine and class?" auth="zeriyoshi" voteType="single" closed="true">
    * Yes
    * No
@@ Line 134: / Line 138: @@
 The following names have been reserved and will no longer be available:
-  * Random\Engine\PCGOneseq128XslRr64
+  * Random\Engine\Mt19937
+  * Random\Engine\PcgOneseq128XslRr64
   * Random\Engine\Xoshiro256StarStar
-  * Random\Engine\MT19937
 The following class names will be made available again:
   * Random\Engine\CombinedLCG
+  * Random\Engine\MersenneTwister
+  * Random\Engine\PCG64
   * Random\SerializableEngine
@@ Line 163: / Line 169: @@
 ===== Patches and Tests =====
-currently none
+  * https://github.com/php/php-src/pull/8094
 ===== References =====
   * https://externals.io/message/117939

Differences

Page Tools