Merge pull request #911 from boriel-basic/fix/fix_peephole_opt_recipe_parser

boriel · web-flow · commit a5ac45480f20 · 2024-11-24T23:27:07.000Z
refact: improve typing and improve parsing in opt parser
diff --git a/src/arch/z80/peephole/evaluator.py b/src/arch/z80/peephole/evaluator.py
@@ -75,8 +75,8 @@ class FN(StrEnum):
     FN.CTEST: lambda x: memcell.MemCell(x, 1).condition_flag,  # condition test, if any. E.g. retz returns 'z'
     FN.NEEDS: lambda x: memcell.MemCell(x[0], 1).needs(x[1]),
     FN.FLAGVAL: lambda x: helpers.new_tmp_val(),
-    FN.OP1: lambda x: (x.strip().replace(",", " ", 1).split() + [""])[1],
-    FN.OP2: lambda x: (x.strip().replace(",", " ", 1).split() + ["", ""])[2],
+    FN.OP1: lambda x: (x.strip().replace(",", " ", 1).split() + [""])[1],  # 1st Operand of an instruction or ""
+    FN.OP2: lambda x: (x.strip().replace(",", " ", 1).split() + ["", ""])[2],  # 2nd Operand of an instruction or ""
 }
 
 # Binary operators
@@ -228,14 +228,22 @@ def eval(self, vars_: dict[str, Any] | None = None) -> str | Evaluator | list[An
             return vars_[val]
 
         if len(self.expression) == 2:
-            oper = FN(self.expression[0])
-            assert oper in UNARY
+            try:
+                oper = FN(self.expression[0])
+                assert oper in UNARY
+            except (AssertionError, ValueError):
+                raise ValueError(f"Invalid unary operator '{self.expression[0]}'")
+
             operand = self.expression[1].eval(vars_)
             return self.normalize(UNARY[oper](operand))
 
         if len(self.expression) == 3 and self.expression[1] != FN.OP_COMMA:
-            oper = FN(self.expression[1])
-            assert oper in BINARY
+            try:
+                oper = FN(self.expression[1])
+                assert oper in BINARY
+            except (AssertionError, ValueError):
+                raise ValueError(f"Invalid binary operator '{self.expression[1]}'")
+
             # Do lazy evaluation
             left_ = lambda: self.expression[0].eval(vars_)
             right_ = lambda: self.expression[2].eval(vars_)
diff --git a/src/arch/z80/peephole/parser.py b/src/arch/z80/peephole/parser.py
@@ -25,7 +25,7 @@
 O_LEVEL = "OLEVEL"
 O_FLAG = "OFLAG"
 
-# Operators : priority (lower number -> highest priority)
+# Operators : precedence (lower number -> highest priority)
 IF_OPERATORS: Final[MappingProxyType[FN, int]] = MappingProxyType(
     {
         FN.OP_NMUL: 3,
@@ -54,6 +54,10 @@
 REQUIRED = (REG_REPLACE, REG_WITH, O_LEVEL, O_FLAG)
 
 
+class PeepholeParserSyntaxError(SyntaxError):
+    pass
+
+
 def simplify_expr(expr: list[Any]) -> list[Any]:
     """Simplifies ("unnest") a list, removing redundant brackets.
     i.e. [[x, [[y]]] becomes [x, [y]]
@@ -79,23 +83,35 @@ class DefineLine(NamedTuple):
     expr: Evaluator
 
 
-def parse_ifline(if_line: str, lineno: int) -> TreeType | None:
-    """Given a line from within a IF region (i.e. $1 == "af'")
-    returns it as a list of tokens ['$1', '==', "af'"]
-    """
-    stack: list[TreeType] = []
-    expr: TreeType = []
-    paren = 0
-    error_ = False
+class Tokenizer:
+    def __init__(self, source: str, lineno: int) -> None:
+        self.source = source
+        self.lineno = lineno
 
-    while not error_ and if_line:
-        if_line = if_line.strip()
-        if not if_line:
-            break
-        qq = RE_IFPARSE.match(if_line)
+    def get_token(self) -> str:
+        """Returns next token, or "" as EOL"""
+        tok = self.lookahead()
+        self.source = self.source[len(tok) :]
+        return tok
+
+    def get_next_token(self) -> str:
+        if self.has_finished():
+            raise PeepholeParserSyntaxError("Unexpected EOL")
+
+        return self.get_token()
+
+    def has_finished(self) -> bool:
+        return self.source == ""
+
+    def lookahead(self) -> str:
+        """Returns next token, or "" as EOL"""
+        self.source = self.source.strip()
+        if self.has_finished():
+            return ""
+
+        qq = RE_IFPARSE.match(self.source)
         if not qq:
-            error_ = True
-            break
+            raise PeepholeParserSyntaxError(f"Syntax error in line {self.lineno}: {self.source}")
 
         tok = qq.group()
         if not RE_ID.match(tok):
@@ -104,7 +120,25 @@ def parse_ifline(if_line: str, lineno: int) -> TreeType | None:
                     tok = tok[: len(oper)]
                     break
 
-        if_line = if_line[len(tok) :]
+        return tok
+
+
+def parse_ifline(if_line: str, lineno: int) -> TreeType | None:
+    """Given a line from within a IF region (i.e. $1 == "af'")
+    returns it as a list of tokens ['$1', '==', "af'"]
+    """
+    stack: list[TreeType] = []
+    expr: TreeType = []
+    paren = 0
+    error_ = False
+    tokenizer = Tokenizer(if_line, lineno)
+
+    while not tokenizer.has_finished():
+        try:
+            tok = tokenizer.get_token()
+        except PeepholeParserSyntaxError as e:
+            errmsg.warning(lineno, str(e))
+            return None
 
         if tok == "(":
             paren += 1
@@ -139,8 +173,8 @@ def parse_ifline(if_line: str, lineno: int) -> TreeType | None:
                 errmsg.warning(lineno, f"Unexpected {tok} in list")
                 return None
 
-        while len(expr) == 2 and isinstance(expr[-2], str):
-            op: str | TreeType = expr[-2]
+        while len(expr) == 2 and isinstance(expr[0], str):
+            op: str = expr[0]
             if op in UNARY:
                 stack[-1].append(expr)
                 expr = stack.pop()
@@ -162,7 +196,11 @@ def parse_ifline(if_line: str, lineno: int) -> TreeType | None:
 
             expr = [expr]
 
-    if not error_ and paren:
+    if error_:
+        errmsg.warning(lineno, "syntax error in IF section")
+        return None
+
+    if paren:
         errmsg.warning(lineno, "unclosed parenthesis in IF section")
         return None
 
@@ -181,11 +219,20 @@ def parse_ifline(if_line: str, lineno: int) -> TreeType | None:
                 errmsg.warning(lineno, f"unexpected binary operator '{op}'")
                 return None
 
-    if error_:
-        errmsg.warning(lineno, "syntax error in IF section")
+    expr = simplify_expr(expr)
+    if len(expr) == 2 and isinstance(expr[-1], str) and expr[-1] in BINARY:
+        errmsg.warning(lineno, f"Unexpected binary operator '{expr[-1]}'")
+        return None
+
+    if len(expr) == 3 and (expr[1] not in BINARY or expr[1] == FN.OP_COMMA):
+        errmsg.warning(lineno, f"Unexpected binary operator '{expr[1]}'")
+        return None
+
+    if len(expr) > 3:
+        errmsg.warning(lineno, "Lists not allowed in IF section condition. Missing operator")
         return None
 
-    return simplify_expr(expr)
+    return expr
 
 
 def parse_define_line(sourceline: SourceLine) -> tuple[str | None, TreeType | None]:
@@ -329,12 +376,12 @@ def check_entry(key: str) -> bool:
     return result
 
 
-def parse_file(fname: str):
+def parse_file(fname: str) -> dict[str, str | int | list[str | list]] | None:
     """Opens and parse a file given by filename"""
     tmp = global_.FILENAME
     global_.FILENAME = fname  # set filename so it shows up in error/warning msgs
 
-    with open(fname, "rt") as f:
+    with open(fname, "rt", encoding="utf-8") as f:
         result = parse_str(f.read())
 
     global_.FILENAME = tmp  # restores original filename
diff --git a/tests/arch/zx48k/peephole/test_parser.py b/tests/arch/zx48k/peephole/test_parser.py
@@ -32,25 +32,22 @@ def test_parse_string(self):
 
         self.maxDiff = None
         self.assertIsInstance(result, dict)
-        self.assertDictEqual(
-            result,
-            {
-                "DEFINE": [],
-                "IF": [
-                    [
-                        [["$1", "==", "af'"], "&&", ["$1", "==", 'Hello ""World""']],
-                        "&&",
-                        [["$1", "==", "(hl)"], "||", ["IS_INDIR", ["$1"]]],
-                    ],
-                    "||",
-                    ["$1", "==", "aa"],
+        assert result == {
+            "DEFINE": [],
+            "IF": [
+                [
+                    [["$1", "==", "af'"], "&&", ["$1", "==", 'Hello ""World""']],
+                    "&&",
+                    [["$1", "==", "(hl)"], "||", ["IS_INDIR", ["$1"]]],
                 ],
-                "OFLAG": 15,
-                "OLEVEL": 1,
-                "REPLACE": ["push $1", "pop $1"],
-                "WITH": [],
-            },
-        )
+                "||",
+                ["$1", "==", "aa"],
+            ],
+            "OFLAG": 15,
+            "OLEVEL": 1,
+            "REPLACE": ["push $1", "pop $1"],
+            "WITH": [],
+        }
 
     def test_parse_call(self):
         result = parser.parse_str(
@@ -398,3 +395,92 @@ def test_parse_if_must_start_in_a_new_line(self):
             """
         )
         assert result is None
+
+    def test_parse_with_ending_binary_error(self):
+        result = parser.parse_str(
+            """
+            ;; Remove the boolean normalization if it's done after calling
+            ;; certain routines that return the bool result already normalized.
+
+            ;; The sequence
+            ;;   sub 1
+            ;;   sbc a, a
+            ;;   inc a
+            ;; can be removed
+
+            OLEVEL: 1
+            OFLAG: 20
+
+            REPLACE {{
+              $1
+            }}
+
+            WITH {{
+            }}
+
+            IF {{
+              $1 == "ld a, 0" ||
+            }}
+            """
+        )
+        assert result is None
+
+    def test_parse_with_comma_error(self):
+        result = parser.parse_str(
+            """
+            ;; Remove the boolean normalization if it's done after calling
+            ;; certain routines that return the bool result already normalized.
+
+            ;; The sequence
+            ;;   sub 1
+            ;;   sbc a, a
+            ;;   inc a
+            ;; can be removed
+
+            OLEVEL: 1
+            OFLAG: 20
+
+            REPLACE {{
+              $1
+            }}
+
+            WITH {{
+            }}
+
+            IF {{
+              $1 == "ld a, 0" ,
+              $1 == "ld a, 0"
+            }}
+            """
+        )
+        assert result is None
+
+    def test_parse_with_nested_comma_error(self):
+        result = parser.parse_str(
+            """
+            ;; Remove the boolean normalization if it's done after calling
+            ;; certain routines that return the bool result already normalized.
+
+            ;; The sequence
+            ;;   sub 1
+            ;;   sbc a, a
+            ;;   inc a
+            ;; can be removed
+
+            OLEVEL: 1
+            OFLAG: 20
+
+            REPLACE {{
+              $1
+            }}
+
+            WITH {{
+            }}
+
+            IF {{
+              ($1 == "ld a, 0" ,
+              $1 == "ld a, 0")
+            }}
+            """
+        )
+        assert result is None
