From fc9b6b30c688b3fb72aa7792ac3302822ae5c822 Mon Sep 17 00:00:00 2001
From: Antoine Viallon <antoine@lesviallon.fr>
Date: Mon, 8 May 2023 00:50:42 +0200
Subject: [PATCH] parser: fully rewrite parsing Use a simpler and more direct
 recursive descent method.

---
 compiler/__main__.py  |  74 +----------------
 compiler/nodes.py     |  34 +++++++-
 compiler/parser.py    | 106 +++++++++++++++++++++++++
 compiler/rules.py     | 180 ------------------------------------------
 compiler/tokenizer.py |   9 ++-
 5 files changed, 151 insertions(+), 252 deletions(-)
 create mode 100644 compiler/parser.py
 delete mode 100644 compiler/rules.py

diff --git a/compiler/__main__.py b/compiler/__main__.py
index ec8675a..7b8dac9 100644
--- a/compiler/__main__.py
+++ b/compiler/__main__.py
@@ -1,73 +1,10 @@
 from __future__ import annotations
 
-from .nodes import Node
-from .rules import Rule, And, Or, Terminal, EvalResult
-from .tokenizer import Tokenizer, Tokens, Token
+from .tokenizer import Tokenizer, Tokens
+from .parser import Parser
 
 data = "2 * (32.9 + 1)"
 
-grammar = r"""
-number = r"[0-9]+(\.[0-9]*)?"
-operator = "+" | "-" | "*" | "/"
-value = "(" expression ")" | number | expression
-expression = value operator value
-root = expression
-"""
-
-number = Rule.make_rule("number",
-                        Terminal(token_type=Tokens.Number)
-                        )
-operator = Rule.make_rule("operator", Or(
-    Terminal(Tokens.Op_Plus), Terminal(Tokens.Op_Minus),
-    Terminal(Tokens.Op_Multiply), Terminal(Tokens.Op_Divide)
-))
-
-value = Rule.make_rule("value", Or(
-    And(
-        Terminal(Tokens.Parens_Left), operator, Terminal(Tokens.Parens_Right)
-    ),
-    number,
-    "expression"
-))
-
-expression = Rule.make_rule("expression", And(
-    number,
-    operator,
-    number
-))
-
-root = expression
-root.prepare()
-
-
-def print_results(result: EvalResult) -> str:
-    prefix = "\t" * print_results._depth
-    message = ""
-    if type(result) != EvalResult:
-        message = f"{prefix}{str(result)}\n"
-        return message
-
-    print_results._depth += 1
-
-    if result.name is not None:
-        message += f"{prefix}{result.name}\n"
-
-    if type(result.result) == list:
-        message += prefix + "{\n"
-        mylist = []
-        for r in result.result:
-            mylist += [print_results(r)]
-        message += " ".join(mylist)
-        message += prefix + "}\n"
-    else:
-        message += print_results(result.result) + "\n"
-    print_results._depth -= 1
-    return message
-
-
-print_results._depth = 0
-
-
 def main():
     tokenizer = Tokenizer()
     tokens = tokenizer.tokenize("2 + 3")
@@ -75,11 +12,8 @@ def main():
     tokens = [token for token in tokens if token.kind != Tokens.Blank]
     print(tokens)
 
-    result = root.evaluate(tokens)
-    print(result)
-    print(print_results(result))
-    if result.errors is not None:
-        raise result.errors
+    parser = Parser(tokens)
+    parser.root()
 
 
 if __name__ == "__main__":
diff --git a/compiler/nodes.py b/compiler/nodes.py
index 35d00a0..ef70583 100644
--- a/compiler/nodes.py
+++ b/compiler/nodes.py
@@ -1,4 +1,36 @@
 from __future__ import annotations
 
+
 class Node:
-    pass
\ No newline at end of file
+    pass
+
+
+class Operator(Node):
+    op: str
+
+
+class Sum(Node):
+    def __init__(self, *values: Expression):
+        self.values = values
+
+    def __repr__(self):
+        return f"{self.__class__.__name__}({', '.join(repr(v) for v in self.values)})"
+
+
+class Product(Node):
+    def __init__(self, *values: Expression):
+        self.values = values
+
+    def __repr__(self):
+        return f"{self.__class__.__name__}({', '.join(repr(v) for v in self.values)})"
+
+
+class Number(Node):
+    def __init__(self, value: int):
+        self.value = value
+
+    def __repr__(self):
+        return f"{self.__class__.__name__}({self.value})"
+
+
+Expression = Sum | Product | Number
diff --git a/compiler/parser.py b/compiler/parser.py
new file mode 100644
index 0000000..1cf732a
--- /dev/null
+++ b/compiler/parser.py
@@ -0,0 +1,106 @@
+from __future__ import annotations
+
+from beartype.typing import List
+
+from .logger import make_logger
+from .nodes import Number, Sum, Expression, Product
+from .source import SourceLocation
+from .tokenizer import Tokens, Token
+
+logger = make_logger(__name__)
+
+
+class ParsingError(Exception):
+    def __init__(self, location: SourceLocation, message: str = "Unknown error"):
+        super().__init__(f"{message} at {str(location)}")
+        self.location = location
+
+
+class Parser:
+    def __init__(self, tokens: List[Token]):
+        self.tokens = tokens
+        self.pos = 0
+
+    @property
+    def token(self) -> Token:
+        if self.pos >= len(self.tokens):
+            return Token(kind=Tokens.EOF)
+        return self.tokens[self.pos]
+
+    @property
+    def prev_token(self) -> Token:
+        return self.tokens[self.pos - 1]
+
+    def next_symbol(self):
+        self.pos += 1
+        logger.debug("%s", f"Advancing to token {self.pos} {self.token}")
+
+    def accept(self, *token_types: Tokens) -> False | Token:
+        tok = self.token
+        if self.token.kind in token_types:
+            self.next_symbol()
+            return tok
+        return False
+
+    def peek(self, token_type) -> False | Token:
+        tok = self.token
+        if self.token.kind == token_type:
+            return tok
+        return False
+
+    def expect(self, token_type: Tokens) -> Token:
+        r = self.accept(token_type)
+        logger.debug("%s", f"Expecting {token_type}, got {r}")
+        if r is False:
+            raise ParsingError(self.token.loc, f"Unexpected token '{self.token}', wanted {token_type}")
+        return r
+
+    def factor(self) -> Expression:
+        if self.accept(Tokens.Parens_Left):
+            v = self.expression()
+            self.expect(Tokens.Parens_Right)
+            return v
+        elif tok := self.accept(Tokens.Number):
+            logger.debug("%s", f"Found number {self.prev_token}")
+            return Number(value=int(tok.value))
+        else:
+            raise ParsingError(self.token.loc, f"Unexpected token '{self.token}', wanted parenthesized expression or "
+                                               f"number")
+
+    def term(self) -> Expression:
+        operations = []
+        operand = self.factor()
+        operations += [operand]
+
+        while operator := self.accept(Tokens.Op_Multiply, Tokens.Op_Divide):
+            operand = self.factor()
+            operations += [operand]
+
+        if len(operations) == 1:
+            return operations[0]
+
+        logger.debug("%s", f"Product of the following terms: {operations}")
+        return Product(*operations)
+
+    def summation(self) -> Sum:
+        operations = []
+        operand = self.term()
+        operations += [operand]
+
+        while operator := self.accept(Tokens.Op_Plus, Tokens.Op_Minus):
+            operand = self.term()
+            operations += [operand]
+
+        if len(operations) == 1:
+            return operations[0]
+
+        logger.debug("%s", f"Sum of the following terms: {operations}")
+        return Sum(*operations)
+
+    def expression(self) -> Expression:
+        summation = self.summation()
+        return summation
+
+    def root(self):
+        self.expression()
+        self.expect(Tokens.EOF)
diff --git a/compiler/rules.py b/compiler/rules.py
deleted file mode 100644
index 883b232..0000000
--- a/compiler/rules.py
+++ /dev/null
@@ -1,180 +0,0 @@
-from __future__ import annotations
-
-from dataclasses import dataclass
-
-from beartype.typing import Union, Dict, Optional, List, Any, Tuple
-import abc
-
-from .nodes import Node
-from .source import SourceLocation
-from .tokenizer import Token, Tokens
-from .logger import logger
-
-class ParsingError(Exception):
-    def __init__(self, location: SourceLocation, message: str = "Unknown error"):
-        super().__init__(f"{message} at {str(location)}")
-        self.location = location
-
-@dataclass
-class EvalResult:
-    result: EvalResult | List[EvalResult] | Token | None = None
-    errors: Optional[ParsingError] = None
-    name: Optional[str] = None
-
-    def __str__(self):
-        #if isinstance(self.result, EvalResult):
-        #    return self.result
-
-        name = self.name if self.name is not None else self.__class__.__name__
-        value = str(self.result)
-        if type(self.result) == list:
-            value = ", ".join(str(r) for r in self.result)
-        return f"{name}({value})"
-    
-    def __repr__(self):
-        return super(EvalResult, self).__repr__()
-
-RuleLike = Union[str, 'Rule']
-
-class Rule(abc.ABC):
-    _named_rules: Dict[str, Rule] = dict()
-
-    def __init__(self, *sub_rules: RuleLike):
-        self._prepared: bool = False
-        self._rules: List[RuleLike] = list(sub_rules)
-        self.rules: List[Rule] = []
-        self.name: Optional[str] = None
-        self.node: Optional[Node] = None
-
-    @staticmethod
-    def make_rule(name: str, rule: Rule, node: Optional[Node] = None) -> Rule:
-        rule.name = name
-        rule.node = node
-        Rule._named_rules[name] = rule
-        return rule
-
-    @abc.abstractmethod
-    def evaluate(self, tokens: List[Token], *, depth: int = 0, parent: Optional[Rule] = None) -> EvalResult:
-        raise NotImplementedError()
-
-    def prepare(self):
-        if self._prepared:
-            return
-
-        for key, rule in enumerate(self._rules):
-            if type(rule) == str:
-                self._rules[key] = Rule._named_rules[rule]
-
-        self.rules = self._rules
-
-        self._prepared = True
-        for rule in self._rules:
-            if not rule._prepared:
-                rule.prepare()
-
-    def __repr__(self):
-        return "{}:{}({})".format(self.__class__.__name__,
-                                  self.name if self.name is not None else "",
-                                  ", ".join(rule.__class__.__name__ for rule in self.rules))
-
-
-class Terminal(Rule):
-    def __init__(self, token_type: Tokens):
-        super().__init__()
-        self.token_type = token_type
-
-    def evaluate(self, tokens: List[Token], *, depth: int = 0, parent: Optional[Rule] = None) -> EvalResult:
-        assert len(tokens) > 0
-        result = EvalResult(name=self.name)
-        logger.debug("%s", f"{depth}: Terminal: Evaluating terminal token with tokens: '{tokens}'")
-        if len(tokens) != 1:
-            result.errors = ParsingError(tokens[0].loc, message=f"Terminal rule must have exactly one token")
-            return result
-
-        if tokens[0].kind != self.token_type:
-            result.errors = ParsingError(
-                tokens[0].loc,
-                message=f"Unexpected token. Wanted {self.token_type.name}, got {tokens[0].kind.name}"
-            )
-            return result
-
-        result.result = tokens[0]
-
-        logger.debug(f"{depth}: Terminal: Found terminal node: {result}")
-
-        return result
-
-    def __repr__(self):
-        return "{}({})".format(self.__class__.__name__, self.token_type)
-
-
-class Or(Rule):
-    def evaluate(self, tokens: List[Token], *, depth: int = 0, parent: Optional[Rule] = None) -> EvalResult:
-        result = EvalResult(errors=ParsingError(location=tokens[0].loc), name=self.name)
-        rule: Rule
-        for i, rule in enumerate(self.rules):
-            logger.debug(f"{depth}: Or: Rule {i + 1}/{len(self.rules)} : trying rule: {rule}")
-            result = rule.evaluate(tokens, depth=depth + 1)
-            if result.errors is None:
-                logger.debug(f"{depth}: Or: Rule {i + 1}/{len(self.rules)} '{rule}' matched, result: {result}")
-                break
-
-        logger.debug(f"{depth}: Or: Finished with errors: {result.errors}")
-
-        result.name = self.name
-
-        return result
-
-
-class And(Rule):
-    def evaluate(self, tokens: List[Token], *, depth: int = 0, parent: Optional[Rule] = None) -> EvalResult:
-        result = EvalResult(errors=ParsingError(tokens[0].loc), name=self.name)
-        result.result = []
-        begin = 0
-        end = 0
-        for i, rule in enumerate(self.rules):
-            logger.debug("%s", f"{depth}: And: Trying rule '{rule}'")
-
-            if end == len(tokens):
-                logger.error("%s", f"{depth}: And: Oops, reached the end of the tokens")
-
-            best_r: Optional[EvalResult] = None
-            while end < len(tokens):
-                tokens_ = tokens[begin:end + 1]
-                r = rule.evaluate(tokens_, depth=depth + 1)
-
-                # No previous match, but we found one
-                if best_r is None and r.errors is None:
-                    logger.debug(f"{depth}: And: Rule {i + 1}/{len(self.rules)} '{rule}' matched, result: {r}")
-                    best_r = r
-
-                # No result at all. That's an error.
-                elif best_r is None and r.errors is not None:
-                    result.errors = r.errors
-                    return result
-
-                # We had a result, but we still matched with more tokens
-                elif best_r is not None and r.errors is None:
-                    logger.debug(f"{depth}: And: Rule {i + 1}/{len(self.rules)} '{rule}' **improved**, result: {r}")
-                    best_r = r
-
-                # We already have a match, and we can't improve it. Finish this rule.
-                elif best_r is not None and r.errors is not None:
-                    logger.debug(
-                        f"{depth}: And: Rule {i + 1}/{len(self.rules)} '{rule}' has FINAL match, result: {best_r}")
-                    result.result += [best_r]
-                    # Matching rule ended at 'end - 1', meaning next rule will begin at end
-                    begin = end
-                    break
-
-                end += 1
-
-            else:
-                logger.debug(
-                    f"{depth}: And: Rule {i + 1}/{len(self.rules)} '{rule}' has FINAL match, finishing rule, result: {best_r}")
-                result.result += [best_r]
-                result.errors = None
-
-        if end != len(tokens):
-            logger.debug(f"{depth}: And: Didn't consume all tokens")
-        return result
diff --git a/compiler/tokenizer.py b/compiler/tokenizer.py
index 504b9c4..6145765 100644
--- a/compiler/tokenizer.py
+++ b/compiler/tokenizer.py
@@ -1,7 +1,7 @@
 from __future__ import annotations
 from dataclasses import dataclass, field
 from beartype import beartype
-from beartype.typing import Optional
+from beartype.typing import Optional, List
 
 import enum
 import re
@@ -31,8 +31,12 @@ class Tokens(enum.Enum):
     Parens_Left = re.compile(r"\(")
     Parens_Right = re.compile(r"\)")
     Blank = re.compile(r"\s+")
+    EOF = re.compile(r"\Z")
     Unknown = re.compile(r".*")
 
+    def __bool__(self):
+        return True
+
 
 class Tokenizer:
     def __init__(self):
@@ -67,4 +71,7 @@ class Tokenizer:
 
             results += [best_result]
             begin += len(best_result.value)
+        results += [Token(Tokens.EOF, value=None, loc=SourceLocation(
+            Location(line=0, character=len(data)), source=data
+        ))]
         return results
\ No newline at end of file