Native Markup Expressions (JSX in PHP)

Zend/tests/markup_operand_position_bc.phpt

@@ -0,0 +1,131 @@
+--TEST--
+Native markup: bare "<" after operand-ending tokens stays a comparison (no-space BC)
+--FILE--
+<?php
+// Postfix ++/-- end an operand: "<" is a comparison, not markup.
+$i = 0; $n = 5;
+$a = [1, 2, 3];
+var_dump($i++<PHP_INT_MAX);
+var_dump($i--<$n);
+$count = 0;
+while ($count++<count($a)) {}
+var_dump($count);
+
+// The `class` keyword in ::class ends an operand too.
+const FOO = "zzz";
+var_dump(\stdClass::class<FOO);
+var_dump(self_name()<FOO);
+function self_name(): string { return "aaa"; }
+
+// Plain constants / qualified names end an operand.
+var_dump(FOO<"zzzz");
+var_dump(PHP_INT_MAX<FOO);
+
+// `]` (array dereference) and `}` end an operand.
+var_dump($a[0]<FOO);
+var_dump(match(true) { default => 1 }<FOO);
+
+// A closing heredoc label ends an operand.
+var_dump(<<<EOT
+aaa
+EOT<FOO);
+
+// Interpolated string closing quote ends an operand.
+$s = "aa";
+var_dump("a$s"<FOO);
+
+// `static` is a complete operand (instanceof static), never markup.
+class Base { public function check(object $o): bool { return $o instanceof static<FOO; } }
+var_dump((new Base)->check(new Base));
+
+// Bare `exit` is an expression operand; "<" after it is a comparison.
+// (Only reached when the left side is falsy-compared, so guard it.)
+function never_exit(): bool { return false && exit<FOO; }
+var_dump(never_exit());
+
+// The legacy `<>` operator is untouched in infix (operator) position.
+var_dump($a[0]<>2);
+
+// Plain variables, literals, and `)` end an operand too, with or without spaces.
+$lhs = 1; $rhs = 2;
+var_dump($lhs < $rhs);
+var_dump($lhs<$rhs);
+var_dump(3 < 4);
+var_dump(($lhs) < $rhs);
+
+// Other "<"-led operators are unchanged: spaceship, less-or-equal, shifts.
+var_dump($lhs <=> $rhs);
+var_dump($lhs <= $rhs);
+var_dump(1 << 4);
+var_dump(64 >> 2);
+
+// A heredoc body containing "<" stays literal text, not markup.
+echo <<<EOT
+literal <not-markup> text
+EOT, "\n";
+
+// Error suppression with "@" still works before an expression.
+$sup = []; var_dump(@$sup['missing']);
+
+// Bare `yield` is a complete operand: `<` and `<>` after it stay comparisons
+// ((yield) != 5, (yield) < FOO), with or without spaces. Yielding markup
+// therefore needs parentheses: `yield (<div/>)`.
+function gen_ne() { var_dump(yield <> 5); }
+$g = gen_ne(); $g->current(); $g->send(7);
+function gen_lt() { var_dump(yield < FOO); }
+$g = gen_lt(); $g->current(); $g->send("aaa");
+function gen_lt_nospace() { $r = yield<FOO; var_dump($r); }
+$g = gen_lt_nospace(); $g->current(); $g->send("aaa");
+
+// Inverse direction: in operand position (after `=`, `(`, `,`, `=>`, `return`,
+// and parenthesized after `yield`) a "<tag" IS markup. Compile-time
+// classification only - never executed, so this parses even in builds where
+// the Html\ runtime classes are unavailable.
+function never_runs() {
+    $x = <div/>;
+    take(<p>hi</p>, <br/>);
+    $f = fn() => <span>a</span>;
+    yield (<li>item</li>);
+    return <em>done</em>;
+}
+
+// Dynamic tags: "<$tag" and "<{expr}" in operand position are markup
+// ("$i--<$n" above is the operator-position inverse, still a comparison).
+function never_runs_dynamic() {
+    $tag = 'div';
+    $x = <$tag class="a">hi</$tag>;
+    $y = <{$tag . 'x'}>hi</>;
+    take(<{pick()}/>);
+    return <$tag/>;
+}
+echo "parsed\n";
+?>
+--EXPECT--
+bool(true)
+bool(true)
+int(4)
+bool(true)
+bool(true)
+bool(true)
+bool(true)
+bool(true)
+bool(true)
+bool(true)
+bool(true)
+bool(false)
+bool(false)
+bool(true)
+bool(true)
+bool(true)
+bool(true)
+bool(true)
+int(-1)
+bool(true)
+int(16)
+int(16)
+literal <not-markup> text
+NULL
+bool(true)
+bool(true)
+bool(true)
+parsed

Zend/zend_globals.h

@@ -402,6 +402,11 @@ struct _zend_php_scanner_globals {
 	/* initial string length after scanning to first variable */
 	int scanned_string_len;
 
+	/* Last significant token returned, used to decide whether a bare "<" in
+	 * ST_IN_SCRIPTING is in operand position (begins native markup) or operator
+	 * position (a comparison). See zend_language_scanner.l. */
+	int last_token;
+
 	/* hooks */
 	void (*on_event)(
 		zend_php_scanner_event event, int token, int line,

Zend/zend_language_parser.y

@@ -26,6 +26,7 @@
 #include "zend_constants.h"
 #include "zend_language_scanner.h"
 #include "zend_exceptions.h"
+#include "zend_markup.h"
 
 #define YYSIZE_T size_t
 #define yytnamerr zend_yytnamerr
@@ -250,6 +251,20 @@ static YYSIZE_T zend_yytnamerr(char*, const char*);
 /* Token used to force a parse error from the lexer */
 %token T_ERROR
 
+/* Native markup tokens. Declared last so they append to the token enum:
+ * inserting them mid-list would renumber every later token (T_INLINE_HTML
+ * onward), breaking code that hardcodes token integers. */
+%token <ast> T_MARKUP_NAME "markup tag name"
+%token <ast> T_MARKUP_TEXT "markup text"
+%token <ast> T_MARKUP_ATTR_VALUE "markup attribute value"
+%token <ast> T_MARKUP_COMMENT "markup comment"
+%token T_MARKUP_OPEN "markup tag start"
+%token T_MARKUP_CLOSE_OPEN "markup closing tag start"
+%token T_MARKUP_TAG_END "markup tag end"
+%token T_MARKUP_SELF_CLOSE "markup self-close"
+%token T_MARKUP_INTERP_START "markup interpolation start"
+%token <ast> T_MARKUP_DOCTYPE "markup doctype"
+
 %type <ast> top_statement namespace_name name statement function_declaration_statement
 %type <ast> class_declaration_statement trait_declaration_statement legacy_namespace_name
 %type <ast> interface_declaration_statement interface_extends_list
@@ -283,6 +298,8 @@ static YYSIZE_T zend_yytnamerr(char*, const char*);
 %type <ast> attributed_statement attributed_top_statement attributed_class_statement attributed_parameter
 %type <ast> attribute_decl attribute attributes attribute_group namespace_declaration_name
 %type <ast> match match_arm_list non_empty_match_arm_list match_arm match_arm_cond_list
+%type <ast> markup_element markup_children markup_child
+%type <ast> markup_attributes markup_attribute
 %type <ast> enum_declaration_statement enum_backing_type enum_case enum_case_expr
 %type <ast> function_name non_empty_member_modifiers
 %type <ast> property_hook property_hook_list optional_property_hook_list hooked_property property_hook_body
@@ -1249,9 +1266,58 @@ new_non_dereferenceable:
 			{ $$ = zend_ast_create(ZEND_AST_NEW, $2, zend_ast_create_list(0, ZEND_AST_ARG_LIST)); }
 ;
 
+markup_element:
+		T_MARKUP_OPEN T_MARKUP_NAME markup_attributes T_MARKUP_TAG_END markup_children T_MARKUP_CLOSE_OPEN T_MARKUP_NAME T_MARKUP_TAG_END
+			{ $$ = zend_ast_create_markup_checked($2, $3, $5, $7, false); if (!$$) { YYERROR; } }
+	|	T_MARKUP_OPEN T_MARKUP_NAME markup_attributes T_MARKUP_SELF_CLOSE
+			{ $$ = zend_ast_create_markup_element($2, $3, NULL); if (!$$) { YYERROR; } }
+	|	T_MARKUP_OPEN T_MARKUP_TAG_END markup_children T_MARKUP_CLOSE_OPEN T_MARKUP_TAG_END
+			{ $$ = zend_ast_create_markup_element(NULL, NULL, $3); if (!$$) { YYERROR; } }
+	|	T_MARKUP_OPEN T_VARIABLE markup_attributes T_MARKUP_TAG_END markup_children T_MARKUP_CLOSE_OPEN T_VARIABLE T_MARKUP_TAG_END
+			{ $$ = zend_ast_create_markup_checked($2, $3, $5, $7, true); if (!$$) { YYERROR; } }
+	|	T_MARKUP_OPEN T_VARIABLE markup_attributes T_MARKUP_SELF_CLOSE
+			{ $$ = zend_ast_create_markup_dynamic($2, $3, NULL); }
+	|	T_MARKUP_OPEN T_MARKUP_INTERP_START expr '}' markup_attributes T_MARKUP_TAG_END markup_children T_MARKUP_CLOSE_OPEN T_MARKUP_TAG_END
+			{ $$ = zend_ast_create_markup_dynamic_expr($3, $5, $7); }
+	|	T_MARKUP_OPEN T_MARKUP_INTERP_START expr '}' markup_attributes T_MARKUP_SELF_CLOSE
+			{ $$ = zend_ast_create_markup_dynamic_expr($3, $5, NULL); }
+;
+
+markup_attributes:
+		%empty								{ $$ = zend_ast_create_list(0, ZEND_AST_ARRAY); }
+	|	markup_attributes markup_attribute	{ $$ = zend_ast_list_add($1, $2); }
+;
+
+markup_attribute:
+		T_MARKUP_NAME '=' T_MARKUP_ATTR_VALUE
+			{ $$ = zend_ast_create(ZEND_AST_ARRAY_ELEM, $3, $1); }
+	|	T_MARKUP_NAME '=' T_MARKUP_INTERP_START expr '}'
+			{ $$ = zend_ast_create(ZEND_AST_ARRAY_ELEM, $4, $1); }
+	|	T_MARKUP_NAME
+			{ zval t; ZVAL_TRUE(&t); $$ = zend_ast_create(ZEND_AST_ARRAY_ELEM, zend_ast_create_zval(&t), $1); }
+	|	T_MARKUP_INTERP_START T_ELLIPSIS expr '}'
+			{ $$ = zend_ast_create(ZEND_AST_UNPACK, $3); }
+;
+
+markup_children:
+		%empty							{ $$ = zend_ast_create_list(0, ZEND_AST_ARRAY); }
+	|	markup_children markup_child	{ $$ = zend_ast_list_add($1, $2); }
+;
+
+markup_child:
+		T_MARKUP_TEXT					{ $$ = zend_ast_create(ZEND_AST_ARRAY_ELEM, $1, NULL); }
+	|	T_MARKUP_COMMENT				{ $$ = zend_ast_create(ZEND_AST_ARRAY_ELEM, zend_ast_create_markup_raw($1), NULL); }
+	|	T_MARKUP_DOCTYPE				{ $$ = zend_ast_create(ZEND_AST_ARRAY_ELEM, zend_ast_create_markup_raw($1), NULL); }
+	|	T_MARKUP_INTERP_START expr '}'	{ $$ = zend_ast_create(ZEND_AST_ARRAY_ELEM, $2, NULL); }
+	|	T_MARKUP_INTERP_START '}'		{ zval n; ZVAL_NULL(&n); $$ = zend_ast_create(ZEND_AST_ARRAY_ELEM, zend_ast_create_zval(&n), NULL); }
+	|	markup_element					{ $$ = zend_ast_create(ZEND_AST_ARRAY_ELEM, $1, NULL); }
+;
+
 expr:
 		variable
 			{ $$ = $1; }
+	|	markup_element
+			{ $$ = $1; }
 	|	T_LIST '(' array_pair_list ')' '=' expr
 			{ $3->attr = ZEND_ARRAY_SYNTAX_LIST; $$ = zend_ast_create(ZEND_AST_ASSIGN, $3, $6); }
 	|	'[' array_pair_list ']' '=' expr

Zend/zend_language_scanner.h

@@ -30,6 +30,7 @@ typedef struct _zend_lex_state {
 	unsigned char *yy_marker;
 	unsigned char *yy_limit;
 	int yy_state;
+	int last_token; /* operand-vs-operator position for native markup */
 	zend_stack state_stack;
 	zend_ptr_stack heredoc_label_stack;
 	zend_stack nest_location_stack; /* for syntax error reporting */