From c108d690de4b7a46327697f4c7c0458f4bc64d9e Mon Sep 17 00:00:00 2001
From: Niels Dossche <7771979+nielsdos@users.noreply.github.com>
Date: Sat, 24 May 2025 12:33:45 +0200
Subject: [PATCH 1/2] Fix GH-18566: [intl] Weird numeric sort in Collator

This aligns the behaviour with normal (non-intl) asort() by making the following changes:
  - Use the same trailing whitespace logic as Zend's is_numeric_ex()
  - Don't allow errors on trailing data

Targeting master because of the BC break.
---
 UPGRADING                               |  3 ++
 ext/intl/collator/collator_convert.c    |  2 +-
 ext/intl/collator/collator_is_numeric.c | 10 +++++
 ext/intl/tests/gh18566.phpt             | 51 +++++++++++++++++++++++++
 4 files changed, 65 insertions(+), 1 deletion(-)
 create mode 100644 ext/intl/tests/gh18566.phpt

diff --git a/UPGRADING b/UPGRADING
index 400187a40f2c6..ecefb0eca867a 100644
--- a/UPGRADING
+++ b/UPGRADING
@@ -53,6 +53,9 @@ PHP 8.5 UPGRADE NOTES
 
 - Intl:
   . The extension now requires at least ICU 57.1.
+  . The behaviour of Collator::SORT_REGULAR with respect to handling numeric
+    strings is now aligned with the behaviour of SORT_REGULAR in ext/standard.
+    This is a consequence of fixing bug GH-18566.
 
 - LDAP:
   . ldap_get_option() and ldap_set_option() now throw a ValueError when
diff --git a/ext/intl/collator/collator_convert.c b/ext/intl/collator/collator_convert.c
index 463348f2a196b..84eccb7904564 100644
--- a/ext/intl/collator/collator_convert.c
+++ b/ext/intl/collator/collator_convert.c
@@ -317,7 +317,7 @@ zval* collator_convert_string_to_number_if_possible( zval* str, zval *rv )
 		COLLATOR_CONVERT_RETURN_FAILED( str );
 	}
 
-	if ( ( is_numeric = collator_is_numeric( (UChar*) Z_STRVAL_P(str), UCHARS( Z_STRLEN_P(str) ), &lval, &dval, /* allow_errors */ 1 ) ) )
+	if ( ( is_numeric = collator_is_numeric( (UChar*) Z_STRVAL_P(str), UCHARS( Z_STRLEN_P(str) ), &lval, &dval, /* allow_errors */ false ) ) )
 	{
 		if( is_numeric == IS_LONG ) {
 			ZVAL_LONG(rv, lval);
diff --git a/ext/intl/collator/collator_is_numeric.c b/ext/intl/collator/collator_is_numeric.c
index cf18c8a672223..4c473d9477116 100644
--- a/ext/intl/collator/collator_is_numeric.c
+++ b/ext/intl/collator/collator_is_numeric.c
@@ -200,6 +200,14 @@ static zend_long collator_u_strtol(const UChar *nptr, UChar **endptr, int base)
 }
 /* }}} */
 
+/* Consume (trailing) whitespace just like collator_u_strtol() consumes leading whitespace */
+static zend_always_inline UChar *collator_skip_ws(UChar *end_ptr)
+{
+	while (u_isspace(*end_ptr)) {
+		end_ptr++;
+	}
+	return end_ptr;
+}
 
 /* {{{ collator_is_numeric]
  * Taken from PHP6:is_numeric_unicode()
@@ -217,6 +225,7 @@ uint8_t collator_is_numeric( UChar *str, int32_t length, zend_long *lval, double
 	errno=0;
 	local_lval = collator_u_strtol(str, &end_ptr_long, 10);
 	if (errno != ERANGE) {
+		end_ptr_long = collator_skip_ws(end_ptr_long);
 		if (end_ptr_long == str+length) { /* integer string */
 			if (lval) {
 				*lval = local_lval;
@@ -233,6 +242,7 @@ uint8_t collator_is_numeric( UChar *str, int32_t length, zend_long *lval, double
 	if (local_dval == 0 && end_ptr_double == str) {
 		end_ptr_double = NULL;
 	} else {
+		end_ptr_double = collator_skip_ws(end_ptr_double);
 		if (end_ptr_double == str+length) { /* floating point string */
 			if (!zend_finite(local_dval)) {
 				/* "inf","nan" and maybe other weird ones */
diff --git a/ext/intl/tests/gh18566.phpt b/ext/intl/tests/gh18566.phpt
new file mode 100644
index 0000000000000..2319c78f1e779
--- /dev/null
+++ b/ext/intl/tests/gh18566.phpt
@@ -0,0 +1,51 @@
+--TEST--
+GH-18566 ([intl] Weird numeric sort in Collator)
+--FILE--
+<?php
+
+$arr = [
+    '2023-02-04 14:00:00',
+    '2023-01-08 12:00:00',
+    '2023-01-03 12:00:00',
+    '2023-01-03 12:00:00',
+    '2021-01-03 12:00:00',
+    '2023-01-05 14:00:00',
+    '2024-01-03 12:00:00',
+    '2023-01-03 12:00:00',
+    '   ',
+];
+
+$coll = Collator::create('en');
+$coll->asort($arr, Collator::SORT_REGULAR);
+print_r($arr);
+
+$arr = [
+    '   100000',
+    '  10',
+    '  -100  ',
+];
+
+$coll = Collator::create('en');
+$coll->asort($arr, Collator::SORT_REGULAR);
+print_r($arr);
+
+?>
+--EXPECT--
+Array
+(
+    [8] =>    
+    [4] => 2021-01-03 12:00:00
+    [2] => 2023-01-03 12:00:00
+    [3] => 2023-01-03 12:00:00
+    [7] => 2023-01-03 12:00:00
+    [5] => 2023-01-05 14:00:00
+    [1] => 2023-01-08 12:00:00
+    [0] => 2023-02-04 14:00:00
+    [6] => 2024-01-03 12:00:00
+)
+Array
+(
+    [2] =>   -100  
+    [1] =>   10
+    [0] =>    100000
+)

From bd9ca44468ecaa93c07ca4c1404dfce708fbef4b Mon Sep 17 00:00:00 2001
From: Niels Dossche <7771979+nielsdos@users.noreply.github.com>
Date: Sat, 24 May 2025 13:14:19 +0200
Subject: [PATCH 2/2] Forgot extensions section

---
 ext/intl/tests/gh18566.phpt | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/ext/intl/tests/gh18566.phpt b/ext/intl/tests/gh18566.phpt
index 2319c78f1e779..2c689de7ab186 100644
--- a/ext/intl/tests/gh18566.phpt
+++ b/ext/intl/tests/gh18566.phpt
@@ -1,5 +1,7 @@
 --TEST--
 GH-18566 ([intl] Weird numeric sort in Collator)
+--EXTENSIONS--
+intl
 --FILE--
 <?php