Merge pull request #192 from zephir-lang/fix-glued-arithmetic-operators

Fix support of glued arithmetic operators

Author: Jeckerson

CHANGELOG.md

@@ -4,6 +4,11 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](http://keepachangelog.com/)
 and this project adheres to [Semantic Versioning](http://semver.org/).
 
+## [2.0.4] - 2026-06-04
+### Fixed
+- A `-` glued to a digit right after a value (e.g. `len-1`, `5-1`, `arr[0]-1`, `len -1`) is now scanned as the binary subtraction operator instead of the sign of a negative literal, which previously produced a `Syntax error`. Unary negative literals (after `=`, `(`, `,`, `return`, etc.) are unchanged, so parameter defaults, constants and array keys keep their negative-literal AST
+  ([zephir-lang/zephir#2011](https://github.com/zephir-lang/zephir/issues/2011)).
+
 ## [2.0.3] - 2026-06-01
 ### Fixed
 - `instanceof` now binds tighter than logical not, matching PHP — `!a instanceof b` parses as `!(a instanceof b)` instead of `(!a) instanceof b`.

VERSION

@@ -1 +1 @@
-2.0.3
+2.0.4

package.xml

@@ -12,25 +12,28 @@
         <email>anton@phalcon.io</email>
         <active>yes</active>
     </lead>
-    <date>2026-06-01</date>
+    <date>2026-06-04</date>
     <time>10:00:00</time>
     <version>
-        <release>2.0.3</release>
-        <api>2.0.3</api>
+        <release>2.0.4</release>
+        <api>2.0.4</api>
     </version>
     <stability>
         <release>stable</release>
         <api>stable</api>
     </stability>
     <license uri="https://github.com/zephir-lang/php-zephir-parser/blob/development/LICENSE">MIT</license>
     <notes>
-        Mon, Jun 01, 2026 - Zephir Parser 2.0.3
+        Thu, Jun 04, 2026 - Zephir Parser 2.0.4
 
         = Fixed:
 
-        - `instanceof` now binds tighter than logical not, matching PHP:
-          `!a instanceof b` parses as `!(a instanceof b)` instead of
-          `(!a) instanceof b`.
+        - A `-` glued to a digit right after a value (e.g. `len-1`, `5-1`,
+          `arr[0]-1`, `len -1`) is now scanned as the binary subtraction
+          operator instead of the sign of a negative literal, which previously
+          produced a `Syntax error`. Unary negative literals (after `=`, `(`,
+          `,`, `return`, etc.) are unchanged, so parameter defaults, constants
+          and array keys keep their negative-literal AST.
     </notes>
     <contents>
         <dir name="/">
@@ -169,6 +172,7 @@
                     </dir>
 
                     <file name="and.phpt" role="test"/>
+                    <file name="arithmetic-unspaced.phpt" role="test"/>
                     <file name="bitwise.phpt" role="test"/>
                     <file name="bug89.phpt" role="test"/>
                     <file name="cast-numeric-types.phpt" role="test"/>
@@ -209,6 +213,7 @@
                     <file name="special.phpt" role="test"/>
                     <file name="static-constant-access.phpt" role="test"/>
                     <file name="static-property-access.phpt" role="test"/>
+                    <file name="sub-unspaced.phpt" role="test"/>
                     <file name="sub.phpt" role="test"/>
                     <file name="ternary.phpt" role="test"/>
                 </dir>

parser/base.c

@@ -118,7 +118,16 @@ void xx_parse_program(zval *return_value, char *program, size_t program_length,
 
 	while (0 <= (scanner_status = xx_get_token(state, &token))) {
 
-		state->active_token = token.opcode;
+		/*
+		 * Keep track of the last *significant* token so the scanner can tell
+		 * whether a '-' glued to a digit is a binary subtraction (it follows a
+		 * value) or part of a negative literal. Whitespace tokens must not
+		 * reset it, otherwise `len -1` would lose the preceding identifier.
+		 * See https://github.com/zephir-lang/zephir/issues/2011
+		 */
+		if (token.opcode != XX_T_IGNORE) {
+			state->active_token = token.opcode;
+		}
 		state->start_length = (program + program_length - state->cursor);
 
 		switch (token.opcode) {

parser/scanner.re

@@ -23,6 +23,33 @@
 #define YYLIMIT (s->limit)
 #define YYMARKER (s->marker)
 
+/*
+ * Returns 1 when the given (previous, significant) token ends a value, in
+ * which case a following '-' glued to a digit is a binary subtraction operator
+ * rather than the sign of a negative literal.
+ *
+ * See https://github.com/zephir-lang/zephir/issues/2011
+ */
+static int xx_token_ends_value(int opcode) {
+	switch (opcode) {
+		case XX_T_INTEGER:
+		case XX_T_DOUBLE:
+		case XX_T_STRING:
+		case XX_T_ISTRING:
+		case XX_T_CHAR:
+		case XX_T_IDENTIFIER:
+		case XX_T_CONSTANT:
+		case XX_T_TRUE:
+		case XX_T_FALSE:
+		case XX_T_NULL:
+		case XX_T_PARENTHESES_CLOSE:
+		case XX_T_SBRACKET_CLOSE:
+			return 1;
+		default:
+			return 0;
+	}
+}
+
 int xx_get_token(xx_scanner_state *s, xx_scanner_token *token) {
 	char *start = YYCURSOR;
 	int status = XX_SCANNER_RETCODE_IMPOSSIBLE;
@@ -35,6 +62,17 @@ int xx_get_token(xx_scanner_state *s, xx_scanner_token *token) {
 
 		INTEGER = ("-"?[0-9]+)|("-"?[0][x][0-9A-Fa-f]+);
 		INTEGER {
+			/*
+			 * A '-' glued to a digit right after a value (e.g. `len-1`) is a
+			 * binary subtraction, not a negative literal: emit SUB and consume
+			 * only the '-'. See zephir issue #2011.
+			 */
+			if (start[0] == '-' && xx_token_ends_value(s->active_token)) {
+				YYCURSOR = start + 1;
+				s->active_char++;
+				token->opcode = XX_T_SUB;
+				return 0;
+			}
 			token->opcode = XX_T_INTEGER;
 			token->value = estrndup(start, YYCURSOR - start);
 			token->len = YYCURSOR - start;
@@ -44,6 +82,12 @@ int xx_get_token(xx_scanner_state *s, xx_scanner_token *token) {
 
 		DOUBLE = ("-"?[0-9]+"."[0-9]+);
 		DOUBLE {
+			if (start[0] == '-' && xx_token_ends_value(s->active_token)) {
+				YYCURSOR = start + 1;
+				s->active_char++;
+				token->opcode = XX_T_SUB;
+				return 0;
+			}
 			token->opcode = XX_T_DOUBLE;
 			token->value = estrndup(start, YYCURSOR - start);
 			token->len = YYCURSOR - start;

tests/operators/arithmetic-unspaced.phpt

@@ -0,0 +1,51 @@
+--TEST--
+Arithmetic operators with operands glued to digits, no spaces (zephir #2011)
+--SKIPIF--
+<?php include(__DIR__ . '/../skipif.inc'); ?>
+--FILE--
+<?php
+$code =<<<ZEP
+function test() {
+	let a = b+1;
+	let a = b-1;
+	let a = b*1;
+	let a = b/1;
+	let a = b%1;
+	let a = 1+2;
+	let a = 5-1;
+	let a = arr[0]*1;
+	let a = b*-1;
+	let a = b+-1;
+	let a = b/-2;
+	let a = b%-1;
+}
+ZEP;
+
+function describe(array $e): string
+{
+	if ($e["type"] === "int") {
+		return "int(" . $e["value"] . ")";
+	}
+	return sprintf("%s(%s, %s)", $e["type"], $e["left"]["type"], describe($e["right"]));
+}
+
+$ir = zephir_parse_file($code, '(eval code)');
+foreach ($ir[0]["statements"] as $statement) {
+	echo describe($statement["assignments"][0]["expr"]), "\n";
+}
+?>
+--EXPECT--
+add(variable, int(1))
+sub(variable, int(1))
+mul(variable, int(1))
+div(variable, int(1))
+mod(variable, int(1))
+add(int, int(2))
+sub(int, int(1))
+mul(array-access, int(1))
+mul(variable, int(-1))
+add(variable, int(-1))
+div(variable, int(-2))
+mod(variable, int(-1))
+--CREDITS--
+Zephir Team

tests/operators/sub-unspaced.phpt

@@ -0,0 +1,36 @@
+--TEST--
+Subtraction with a "-" glued to a digit after a value (zephir #2011)
+--SKIPIF--
+<?php include(__DIR__ . '/../skipif.inc'); ?>
+--FILE--
+<?php
+$code =<<<ZEP
+function test() {
+	let a = b-1;
+	let a = b -1;
+	let a = b- 1;
+	let a = 5-1;
+	let a = arr[0]-1;
+	let a = -1;
+}
+ZEP;
+
+$ir = zephir_parse_file($code, '(eval code)');
+foreach ($ir[0]["statements"] as $statement) {
+	$expr = $statement["assignments"][0]["expr"];
+	if ($expr["type"] === "int") {
+		printf("int %s\n", $expr["value"]);
+	} else {
+		printf("%s(%s, %s)\n", $expr["type"], $expr["left"]["type"], $expr["right"]["type"]);
+	}
+}
+?>
+--EXPECT--
+sub(variable, int)
+sub(variable, int)
+sub(variable, int)
+sub(int, int)
+sub(array-access, int)
+int -1
+--CREDITS--
+Zephir Team

zephir_parser.h

@@ -15,7 +15,7 @@ extern zend_module_entry zephir_parser_module_entry;
 #define phpext_zephir_parser_ptr &zephir_parser_module_entry
 
 #define PHP_ZEPHIR_PARSER_NAME "zephir_parser"
-#define PHP_ZEPHIR_PARSER_VERSION "2.0.3"
+#define PHP_ZEPHIR_PARSER_VERSION "2.0.4"
 #define PHP_ZEPHIR_PARSER_AUTHOR "Zephir Team and contributors"
 #define PHP_ZEPHIR_PARSER_DESCRIPTION "The Zephir Parser delivered as a C extension for the PHP language."