def test_non_empty_handler_is_true(self):
#arrange
err = ErrorHandler()
#act
err.push(1, "foo")
#assert
self.assertTrue(bool(err))
def test_reset_results_in_false_handler(self):
#arrange
err = ErrorHandler(1)
#act
err.push(1, "foo")
err.reset()
#assert
self.assertFalse(bool(err))
def test_attempt_push_on_non_full_queue_is_true(self):
#arrange
err = ErrorHandler(2)
#act
err.push(1, "foo")
status = err.push(2, "SUT")
#assert
self.assertTrue(status)
def test_pushed_error_is_recovered_on_iteration(self):
#arrange
msg = "line 1: test error"
err = ErrorHandler()
#act
err.push(1, "test error")
#assert
for i in err:
self.assertEqual(msg, i)
def test_reset_then_push_results_in_true_handler(self):
#arrange
err = ErrorHandler(1)
#act
err.push(1, "foo")
err.reset()
err.push(1, "bar")
#assert
self.assertTrue(bool(err))
def test_pushed_mulitple_errors_are_recovered_on_iteration(self):
#arrange
msgs = ["line 1: test error 1", "line 2: test error 2"]
err = ErrorHandler()
#act
err.push(1, "test error 1")
err.push(2, "test error 2")
#assert
for i, val in enumerate(err):
self.assertEqual(msgs[i], val)
class Parser():
def __init__(self):
"""\
convert a list of tokens to an abstract syntax tree
@i: tokens index
@ast: abstract syntax tree generated by self.expression()
"""
self.err = ErrorHandler()
self.tokens = []
self.i = 0
self.ast = None
def parse(self, tokens, limit=10):
"""\
recursive descent entry point
@tokens: list of tokens provided by lexical analysis, tokens[-1] == EOF
@limit: internal ErrorHandler limit
Returns: abstract syntax tree, top level is a list of statements
"""
self.err = ErrorHandler(limit)
self.tokens = tokens
self.i = 0
try:
return (self.program(), self.err)
except ParseError:
return (None, self.err)
def curr_type(self):
"""\
helper function: returns token type of token at current list index
"""
token = self.tokens[self.i]
return token.type
def curr_token(self):
"""\
helper function: syntactic sugar to fetch current token
"""
return self.tokens[self.i]
def advance(self):
"""\
helper function: syntactic sugar for iteration over the tokens list
"""
assert (self.i + 1 < len(self.tokens))
self.i += 1
def prev_token(self):
"""\
helper function: syntactic sugar to fetch previous token
"""
assert (self.i - 1 >= 0)
return self.tokens[self.i - 1]
def check_semicolon(self):
if self.curr_type() == TokenType.SEMICOLON:
self.advance()
return True
else:
tok = self.curr_token()
if tok.type == TokenType.EOF:
self.trap("expected ';' at end of file")
else:
self.trap("expected ';' before {}".format(tok.lexeme))
return False
def program(self):
"""\
<program> := <declaration>* EOF
"""
tree = []
while self.curr_type() != TokenType.EOF:
tree.append(self.declaration())
assert (self.curr_type() == TokenType.EOF)
return tree
def declaration(self):
"""\
<declaration> := <variable declaration> | <statement>
"""
if self.curr_type() == TokenType.VAR:
self.advance()
return self.var_declaration()
return self.statement()
def var_declaration(self):
"""\
<var_declaration> := "var" IDENTIFIER ("=" <expression>)? ";"
"""
name = None
#if no initializer is present, assume that there was actually
#an intializer to nil, i.e., var x = nil; instead of var x;
initializer = Literal(Token(TokenType.NIL, -1, "nil", None))
if self.curr_type() == TokenType.IDENTIFIER:
name = self.curr_token()
self.advance()
if self.curr_type() == TokenType.EQUAL:
self.advance()
initializer = self.expression()
self.check_semicolon()
else:
self.trap("missing variable identifier")
return VariableDeclaration(name, initializer)
def statement(self):
"""\
<statement> := <expression statement> | <print statement> |
<block statement> | <if statement> | <while statement> |
<for statement>
"""
# this isn't the world's fastest code, a jump table or dictionary-based
# switch would be better, but hey we're writing an interpreter in
# python! This is hardly the bottleneck!
if self.curr_type() == TokenType.PRINT:
self.advance()
stmt = self.print_stmt()
elif self.curr_type() == TokenType.LEFT_BRACE:
self.advance()
stmt_list = self.block_stmt()
stmt = Block(stmt_list)
elif self.curr_type() == TokenType.IF:
self.advance()
stmt = self.branch_stmt()
elif self.curr_type() == TokenType.WHILE:
self.advance()
stmt = self.while_stmt()
elif self.curr_type() == TokenType.FOR:
self.advance()
stmt = self.for_stmt()
else:
stmt = self.generic_stmt()
return stmt
def print_stmt(self):
"""\
<print statement> := "print" <expression> ";"
"""
stmt = Printer(self.expression())
self.check_semicolon()
return stmt
def block_stmt(self):
"""\
<block statement> := "{" <declaration>* "}"
this method returns a list of statements rather than a block node, b/c
it is used for both generic block statements and function blocks. The
caller must wrap the list into the appropriate node class.
"""
stmt_list = []
while self.curr_type() != TokenType.RIGHT_BRACE:
expr = self.declaration()
stmt_list.append(expr)
if self.curr_type() == TokenType.EOF:
self.trap("expected '}' at end of file")
elif self.curr_type() != TokenType.RIGHT_BRACE:
tok = self.curr_token()
self.trap("expected '}' at {}".format(tok.lexeme))
else:
self.advance()
return stmt_list
def branch_stmt(self):
"""\
<branch> := "if" "(" <expr> ")" <stmt> ("else" <stmt>)?
"""
if self.curr_type() != TokenType.LEFT_PAREN:
self.trap("expected open parenthesis after 'if'")
return Branch(None, None, None)
self.advance()
condition = self.expression()
if self.curr_type() != TokenType.RIGHT_PAREN:
self.trap("expected close parenthesis after condition")
return Branch(None, None, None)
self.advance()
then_branch = self.statement()
else_branch = None
if self.curr_type() == TokenType.ELSE:
self.advance()
else_branch = self.statement()
return Branch(condition, then_branch, else_branch)
def while_stmt(self):
"""
<while> := "while" "(" <expression> ")" <statement>
"""
if self.curr_type() != TokenType.LEFT_PAREN:
self.trap("expected open parenthesis after 'if'")
return Loop(None, None)
self.advance()
condition = self.expression()
if self.curr_type() != TokenType.RIGHT_PAREN:
self.trap("expected close parenthesis after condition")
return Loop(None, None)
self.advance()
body = self.statement()
return Loop(condition, body)
def for_stmt(self):
"""\
<for statement> := "for" "(" (<var decl> | <expr stmt> | ";")
<expression>? ";" <expression>? ")" <statement>
for statements are desugared into an equivalent while statement.
"""
if self.curr_type() != TokenType.LEFT_PAREN:
self.trap("expected '(' after 'for'")
return Loop(None, None)
self.advance()
initializer = None
if self.curr_type() == TokenType.SEMICOLON:
self.advance()
elif self.curr_type() == TokenType.VAR:
self.advance()
initializer = self.var_declaration()
else:
initializer = self.generic_stmt()
condition = None
if self.curr_type() == TokenType.SEMICOLON:
self.advance()
else:
condition = self.expression()
if not self.check_semicolon():
return Loop(None, None)
increment = None
if self.curr_type() != TokenType.RIGHT_PAREN:
increment = self.expression()
if self.curr_type() == TokenType.RIGHT_PAREN:
self.advance()
else:
self.trap("expected ')' after a for loop clause")
return Loop(None, None)
body = self.statement()
#desugar the for loop by nesting the disjoint nodes into blocks
if increment is not None:
body = Block([body, increment])
if condition is not None:
body = Loop(condition, body)
else:
#a little different from standard lox
#if you want an infinite loop, make it clear with a while(true)
#so in reality, this isn't a by-the-book lox implementation
self.trap("infinite loop detected, use while(true) instead")
return Loop(None, None)
if initializer is not None:
body = Block([initializer, body])
return body
def generic_stmt(self):
"""\
<expression statement> := <expression> ";"
"""
stmt = Generic(self.expression())
self.check_semicolon()
return stmt
def expression(self):
"""\
dummy method used to encode the lox grammar explicity in the source.
<expression> := <assignment>
"""
return self.assignment()
def assignment(self):
"""\
assign rvalue to lvalue
<assignment> := (IDENTIFIER "=" <assignment>) | <logical or>
"""
lval = self.logical_or()
if self.curr_type() == TokenType.EQUAL:
self.advance()
rval = self.assignment()
if isinstance(lval, Variable):
#extract token from variable node as a valid lvalue
return Assignment(lval.name, rval)
self.trap("assignment target is not an l-value")
#if trap was initiated, this return node is just a dummy.
#trap synchronized to the next statement anyways so its no risk.
#on the other hand, if the branch was skipped entirely, then lval
#is just some expression.
return lval
def logical_or(self):
"""\
<logical or> := <logical and> ("or" <logical and>)*
"""
expr = self.logical_and()
if self.curr_type() == TokenType.OR:
self.advance()
left = expr
tok = self.prev_token()
right = self.logical_and()
return Logical(left, tok, right)
return expr
def logical_and(self):
"""\
<logical and> := <equality> ("and" <equality>)*
"""
expr = self.equality()
if self.curr_type() == TokenType.AND:
self.advance()
left = expr
tok = self.prev_token()
right = self.logical_and()
return Logical(left, tok, right)
return expr
def equality(self):
"""\
<equality> := <comparison> (("==" | "!=") <comparison>)*
"""
expr = self.comparison()
types = set([TokenType.EQUAL_EQUAL, TokenType.BANG_EQUAL])
while self.curr_type() in types:
self.advance()
left = expr
operator = self.prev_token()
right = self.comparison()
expr = Binary(left, operator, right)
return expr
def comparison(self):
"""\
<comparison> := <term> ((">" | "<" | "<=" | ">=") <term>)*
"""
expr = self.term()
types = set([TokenType.GREATER, TokenType.GREATER_EQUAL, \
TokenType.LESS, TokenType.LESS_EQUAL])
while self.curr_type() in types:
self.advance()
left = expr
operator = self.prev_token()
right = self.term()
expr = Binary(left, operator, right)
return expr
def term(self):
"""\
<term> := <factor> (("+" | "-") <factor>)*
"""
expr = self.factor()
while self.curr_type() in set([TokenType.PLUS, TokenType.MINUS]):
self.advance()
left = expr
operator = self.prev_token()
right = self.factor()
expr = Binary(left, operator, right)
return expr
def factor(self):
"""\
<factor> := <unary> (("*" | "/") <unary>)*
"""
expr = self.unary()
while self.curr_type() in set([TokenType.STAR, TokenType.SLASH]):
self.advance()
left = expr
operator = self.prev_token()
right = self.unary()
expr = Binary(left, operator, right)
return expr
def unary(self):
"""\
<unary> := ("!" | "-") <unary> | <primary>
"""
if self.curr_type() in set([TokenType.BANG, TokenType.MINUS]):
self.advance()
return Unary(self.prev_token(), self.unary())
return self.primary()
def primary(self):
"""\
<primary> := NUMBER | STRING | "true" | "false" | "nil"
<primary> := "(" <expression> ")"
"""
types = set([TokenType.NUMBER, TokenType.STRING, TokenType.NIL, \
TokenType.TRUE, TokenType.FALSE])
if self.curr_type() in types:
expr = Literal(self.curr_token())
self.advance()
elif self.curr_type() == TokenType.LEFT_PAREN:
self.advance()
expr = Grouping(self.expression())
if self.curr_type() == TokenType.RIGHT_PAREN:
self.advance()
else:
self.trap("missing right parenthesis for grouped expression")
elif self.curr_type() == TokenType.IDENTIFIER:
expr = Variable(self.curr_token())
self.advance()
elif self.curr_type() == TokenType.EOF:
#this situation occurs when the user has a grammar error at the
#end of file such as "3-". In this situation, the parser has been
#passing the EOF token along the call stack. The else branch can
#handle this issue, but its not user friendly because it presents
#a EOF:"None" lexeme to the user.
#
#trap is at EOF so no need to create a dummy expr for return
tok = self.prev_token()
self.trap("misplaced symbol '{}' at end of file".format(
tok.lexeme))
else:
lexeme = (self.tokens[self.i]).lexeme
self.trap("misplaced symbol '{}'".format(lexeme))
#dummy statement will be added to program tree
expr = None
return expr
def trap(self, msg):
"""\
push parameters to error handler then enter panic mode to reset at the
next sequence point.
"""
tok = self.tokens[self.i]
line = tok.line
if not self.err.push(line, msg):
self.err.grow(1)
self.error.push(line, "additional errors found (hidden)")
#synchronize parser to continue at next program statement
types = set([TokenType.CLASS, TokenType.FUN, TokenType.VAR, \
TokenType.FOR, TokenType.IF, TokenType.WHILE, \
TokenType.PRINT, TokenType.RETURN])
while self.curr_type() not in types:
if self.curr_type() == TokenType.EOF:
#no statements left in program so no need to continue parsing
#unwind call stack back to self.program and let it handle return
raise ParseError
else:
self.advance()