| # Copyright 2014-2017 The Meson development team |
| |
| # Licensed under the Apache License, Version 2.0 (the "License"); |
| # you may not use this file except in compliance with the License. |
| # You may obtain a copy of the License at |
| |
| # http://www.apache.org/licenses/LICENSE-2.0 |
| |
| # Unless required by applicable law or agreed to in writing, software |
| # distributed under the License is distributed on an "AS IS" BASIS, |
| # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| # See the License for the specific language governing permissions and |
| # limitations under the License. |
| |
| import re |
| import codecs |
| import textwrap |
| import types |
| import typing as T |
| from .mesonlib import MesonException |
| from . import mlog |
| |
| if T.TYPE_CHECKING: |
| from .ast import AstVisitor |
| |
| # This is the regex for the supported escape sequences of a regular string |
| # literal, like 'abc\x00' |
| ESCAPE_SEQUENCE_SINGLE_RE = re.compile(r''' |
| ( \\U[A-Fa-f0-9]{8} # 8-digit hex escapes |
| | \\u[A-Fa-f0-9]{4} # 4-digit hex escapes |
| | \\x[A-Fa-f0-9]{2} # 2-digit hex escapes |
| | \\[0-7]{1,3} # Octal escapes |
| | \\N\{[^}]+\} # Unicode characters by name |
| | \\[\\'abfnrtv] # Single-character escapes |
| )''', re.UNICODE | re.VERBOSE) |
| |
| class MesonUnicodeDecodeError(MesonException): |
| def __init__(self, match: str) -> None: |
| super().__init__(match) |
| self.match = match |
| |
| def decode_match(match: T.Match[str]) -> str: |
| try: |
| return codecs.decode(match.group(0).encode(), 'unicode_escape') |
| except UnicodeDecodeError: |
| raise MesonUnicodeDecodeError(match.group(0)) |
| |
| class ParseException(MesonException): |
| def __init__(self, text: str, line: str, lineno: int, colno: int) -> None: |
| # Format as error message, followed by the line with the error, followed by a caret to show the error column. |
| super().__init__("{}\n{}\n{}".format(text, line, '{}^'.format(' ' * colno))) |
| self.lineno = lineno |
| self.colno = colno |
| |
| class BlockParseException(MesonException): |
| def __init__( |
| self, |
| text: str, |
| line: str, |
| lineno: int, |
| colno: int, |
| start_line: str, |
| start_lineno: int, |
| start_colno: int, |
| ) -> None: |
| # This can be formatted in two ways - one if the block start and end are on the same line, and a different way if they are on different lines. |
| |
| if lineno == start_lineno: |
| # If block start and end are on the same line, it is formatted as: |
| # Error message |
| # Followed by the line with the error |
| # Followed by a caret to show the block start |
| # Followed by underscores |
| # Followed by a caret to show the block end. |
| super().__init__("{}\n{}\n{}".format(text, line, '{}^{}^'.format(' ' * start_colno, '_' * (colno - start_colno - 1)))) |
| else: |
| # If block start and end are on different lines, it is formatted as: |
| # Error message |
| # Followed by the line with the error |
| # Followed by a caret to show the error column. |
| # Followed by a message saying where the block started. |
| # Followed by the line of the block start. |
| # Followed by a caret for the block start. |
| super().__init__("%s\n%s\n%s\nFor a block that started at %d,%d\n%s\n%s" % (text, line, '%s^' % (' ' * colno), start_lineno, start_colno, start_line, "%s^" % (' ' * start_colno))) |
| self.lineno = lineno |
| self.colno = colno |
| |
| TV_TokenTypes = T.TypeVar('TV_TokenTypes', int, str, bool) |
| |
| class Token(T.Generic[TV_TokenTypes]): |
| def __init__(self, tid: str, filename: str, line_start: int, lineno: int, colno: int, bytespan: T.Tuple[int, int], value: TV_TokenTypes): |
| self.tid = tid # type: str |
| self.filename = filename # type: str |
| self.line_start = line_start # type: int |
| self.lineno = lineno # type: int |
| self.colno = colno # type: int |
| self.bytespan = bytespan # type: T.Tuple[int, int] |
| self.value = value # type: TV_TokenTypes |
| |
| def __eq__(self, other: object) -> bool: |
| if isinstance(other, str): |
| return self.tid == other |
| elif isinstance(other, Token): |
| return self.tid == other.tid |
| return NotImplemented |
| |
| class Lexer: |
| def __init__(self, code: str): |
| self.code = code |
| self.keywords = {'true', 'false', 'if', 'else', 'elif', |
| 'endif', 'and', 'or', 'not', 'foreach', 'endforeach', |
| 'in', 'continue', 'break'} |
| self.future_keywords = {'return'} |
| self.token_specification = [ |
| # Need to be sorted longest to shortest. |
| ('ignore', re.compile(r'[ \t]')), |
| ('fstring', re.compile(r"f'([^'\\]|(\\.))*'")), |
| ('id', re.compile('[_a-zA-Z][_0-9a-zA-Z]*')), |
| ('number', re.compile(r'0[bB][01]+|0[oO][0-7]+|0[xX][0-9a-fA-F]+|0|[1-9]\d*')), |
| ('eol_cont', re.compile(r'\\\n')), |
| ('eol', re.compile(r'\n')), |
| ('multiline_string', re.compile(r"'''(.|\n)*?'''", re.M)), |
| ('comment', re.compile(r'#.*')), |
| ('lparen', re.compile(r'\(')), |
| ('rparen', re.compile(r'\)')), |
| ('lbracket', re.compile(r'\[')), |
| ('rbracket', re.compile(r'\]')), |
| ('lcurl', re.compile(r'\{')), |
| ('rcurl', re.compile(r'\}')), |
| ('dblquote', re.compile(r'"')), |
| ('string', re.compile(r"'([^'\\]|(\\.))*'")), |
| ('comma', re.compile(r',')), |
| ('plusassign', re.compile(r'\+=')), |
| ('dot', re.compile(r'\.')), |
| ('plus', re.compile(r'\+')), |
| ('dash', re.compile(r'-')), |
| ('star', re.compile(r'\*')), |
| ('percent', re.compile(r'%')), |
| ('fslash', re.compile(r'/')), |
| ('colon', re.compile(r':')), |
| ('equal', re.compile(r'==')), |
| ('nequal', re.compile(r'!=')), |
| ('assign', re.compile(r'=')), |
| ('le', re.compile(r'<=')), |
| ('lt', re.compile(r'<')), |
| ('ge', re.compile(r'>=')), |
| ('gt', re.compile(r'>')), |
| ('questionmark', re.compile(r'\?')), |
| ] |
| |
| def getline(self, line_start: int) -> str: |
| return self.code[line_start:self.code.find('\n', line_start)] |
| |
| def lex(self, filename: str) -> T.Generator[Token, None, None]: |
| line_start = 0 |
| lineno = 1 |
| loc = 0 |
| par_count = 0 |
| bracket_count = 0 |
| curl_count = 0 |
| col = 0 |
| while loc < len(self.code): |
| matched = False |
| value = None # type: T.Union[str, bool, int] |
| for (tid, reg) in self.token_specification: |
| mo = reg.match(self.code, loc) |
| if mo: |
| curline = lineno |
| curline_start = line_start |
| col = mo.start() - line_start |
| matched = True |
| span_start = loc |
| loc = mo.end() |
| span_end = loc |
| bytespan = (span_start, span_end) |
| match_text = mo.group() |
| if tid == 'ignore' or tid == 'comment': |
| break |
| elif tid == 'lparen': |
| par_count += 1 |
| elif tid == 'rparen': |
| par_count -= 1 |
| elif tid == 'lbracket': |
| bracket_count += 1 |
| elif tid == 'rbracket': |
| bracket_count -= 1 |
| elif tid == 'lcurl': |
| curl_count += 1 |
| elif tid == 'rcurl': |
| curl_count -= 1 |
| elif tid == 'dblquote': |
| raise ParseException('Double quotes are not supported. Use single quotes.', self.getline(line_start), lineno, col) |
| elif tid in {'string', 'fstring'}: |
| # Handle here and not on the regexp to give a better error message. |
| if match_text.find("\n") != -1: |
| mlog.warning(textwrap.dedent("""\ |
| Newline character in a string detected, use ''' (three single quotes) for multiline strings instead. |
| This will become a hard error in a future Meson release.\ |
| """), |
| self.getline(line_start), |
| str(lineno), |
| str(col) |
| ) |
| value = match_text[2 if tid == 'fstring' else 1:-1] |
| try: |
| value = ESCAPE_SEQUENCE_SINGLE_RE.sub(decode_match, value) |
| except MesonUnicodeDecodeError as err: |
| raise MesonException(f"Failed to parse escape sequence: '{err.match}' in string:\n {match_text}") |
| elif tid == 'multiline_string': |
| tid = 'string' |
| value = match_text[3:-3] |
| lines = match_text.split('\n') |
| if len(lines) > 1: |
| lineno += len(lines) - 1 |
| line_start = mo.end() - len(lines[-1]) |
| elif tid == 'number': |
| value = int(match_text, base=0) |
| elif tid == 'eol_cont': |
| lineno += 1 |
| line_start = loc |
| break |
| elif tid == 'eol': |
| lineno += 1 |
| line_start = loc |
| if par_count > 0 or bracket_count > 0 or curl_count > 0: |
| break |
| elif tid == 'id': |
| if match_text in self.keywords: |
| tid = match_text |
| else: |
| if match_text in self.future_keywords: |
| mlog.warning(f"Identifier '{match_text}' will become a reserved keyword in a future release. Please rename it.", |
| location=types.SimpleNamespace(filename=filename, lineno=lineno)) |
| value = match_text |
| yield Token(tid, filename, curline_start, curline, col, bytespan, value) |
| break |
| if not matched: |
| raise ParseException('lexer', self.getline(line_start), lineno, col) |
| |
| class BaseNode: |
| def __init__(self, lineno: int, colno: int, filename: str, end_lineno: T.Optional[int] = None, end_colno: T.Optional[int] = None): |
| self.lineno = lineno # type: int |
| self.colno = colno # type: int |
| self.filename = filename # type: str |
| self.end_lineno = end_lineno if end_lineno is not None else self.lineno |
| self.end_colno = end_colno if end_colno is not None else self.colno |
| |
| # Attributes for the visitors |
| self.level = 0 # type: int |
| self.ast_id = '' # type: str |
| self.condition_level = 0 # type: int |
| |
| def accept(self, visitor: 'AstVisitor') -> None: |
| fname = 'visit_{}'.format(type(self).__name__) |
| if hasattr(visitor, fname): |
| func = getattr(visitor, fname) |
| if callable(func): |
| func(self) |
| |
| class ElementaryNode(T.Generic[TV_TokenTypes], BaseNode): |
| def __init__(self, token: Token[TV_TokenTypes]): |
| super().__init__(token.lineno, token.colno, token.filename) |
| self.value = token.value # type: TV_TokenTypes |
| self.bytespan = token.bytespan # type: T.Tuple[int, int] |
| |
| class BooleanNode(ElementaryNode[bool]): |
| def __init__(self, token: Token[bool]): |
| super().__init__(token) |
| assert isinstance(self.value, bool) |
| |
| class IdNode(ElementaryNode[str]): |
| def __init__(self, token: Token[str]): |
| super().__init__(token) |
| assert isinstance(self.value, str) |
| |
| def __str__(self) -> str: |
| return "Id node: '%s' (%d, %d)." % (self.value, self.lineno, self.colno) |
| |
| class NumberNode(ElementaryNode[int]): |
| def __init__(self, token: Token[int]): |
| super().__init__(token) |
| assert isinstance(self.value, int) |
| |
| class StringNode(ElementaryNode[str]): |
| def __init__(self, token: Token[str]): |
| super().__init__(token) |
| assert isinstance(self.value, str) |
| |
| def __str__(self) -> str: |
| return "String node: '%s' (%d, %d)." % (self.value, self.lineno, self.colno) |
| |
| class FormatStringNode(ElementaryNode[str]): |
| def __init__(self, token: Token[str]): |
| super().__init__(token) |
| assert isinstance(self.value, str) |
| |
| def __str__(self) -> str: |
| return "Format string node: '{self.value}' ({self.lineno}, {self.colno})." |
| |
| class ContinueNode(ElementaryNode): |
| pass |
| |
| class BreakNode(ElementaryNode): |
| pass |
| |
| class ArgumentNode(BaseNode): |
| def __init__(self, token: Token[TV_TokenTypes]): |
| super().__init__(token.lineno, token.colno, token.filename) |
| self.arguments = [] # type: T.List[BaseNode] |
| self.commas = [] # type: T.List[Token[TV_TokenTypes]] |
| self.kwargs = {} # type: T.Dict[BaseNode, BaseNode] |
| self.order_error = False |
| |
| def prepend(self, statement: BaseNode) -> None: |
| if self.num_kwargs() > 0: |
| self.order_error = True |
| if not isinstance(statement, EmptyNode): |
| self.arguments = [statement] + self.arguments |
| |
| def append(self, statement: BaseNode) -> None: |
| if self.num_kwargs() > 0: |
| self.order_error = True |
| if not isinstance(statement, EmptyNode): |
| self.arguments += [statement] |
| |
| def set_kwarg(self, name: IdNode, value: BaseNode) -> None: |
| if name.value in [x.value for x in self.kwargs.keys() if isinstance(x, IdNode)]: |
| mlog.warning(f'Keyword argument "{name.value}" defined multiple times.', location=self) |
| mlog.warning('This will be an error in future Meson releases.') |
| self.kwargs[name] = value |
| |
| def set_kwarg_no_check(self, name: BaseNode, value: BaseNode) -> None: |
| self.kwargs[name] = value |
| |
| def num_args(self) -> int: |
| return len(self.arguments) |
| |
| def num_kwargs(self) -> int: |
| return len(self.kwargs) |
| |
| def incorrect_order(self) -> bool: |
| return self.order_error |
| |
| def __len__(self) -> int: |
| return self.num_args() # Fixme |
| |
| class ArrayNode(BaseNode): |
| def __init__(self, args: ArgumentNode, lineno: int, colno: int, end_lineno: int, end_colno: int): |
| super().__init__(lineno, colno, args.filename, end_lineno=end_lineno, end_colno=end_colno) |
| self.args = args # type: ArgumentNode |
| |
| class DictNode(BaseNode): |
| def __init__(self, args: ArgumentNode, lineno: int, colno: int, end_lineno: int, end_colno: int): |
| super().__init__(lineno, colno, args.filename, end_lineno=end_lineno, end_colno=end_colno) |
| self.args = args |
| |
| class EmptyNode(BaseNode): |
| def __init__(self, lineno: int, colno: int, filename: str): |
| super().__init__(lineno, colno, filename) |
| self.value = None |
| |
| class OrNode(BaseNode): |
| def __init__(self, left: BaseNode, right: BaseNode): |
| super().__init__(left.lineno, left.colno, left.filename) |
| self.left = left # type: BaseNode |
| self.right = right # type: BaseNode |
| |
| class AndNode(BaseNode): |
| def __init__(self, left: BaseNode, right: BaseNode): |
| super().__init__(left.lineno, left.colno, left.filename) |
| self.left = left # type: BaseNode |
| self.right = right # type: BaseNode |
| |
| class ComparisonNode(BaseNode): |
| def __init__(self, ctype: str, left: BaseNode, right: BaseNode): |
| super().__init__(left.lineno, left.colno, left.filename) |
| self.left = left # type: BaseNode |
| self.right = right # type: BaseNode |
| self.ctype = ctype # type: str |
| |
| class ArithmeticNode(BaseNode): |
| def __init__(self, operation: str, left: BaseNode, right: BaseNode): |
| super().__init__(left.lineno, left.colno, left.filename) |
| self.left = left # type: BaseNode |
| self.right = right # type: BaseNode |
| self.operation = operation # type: str |
| |
| class NotNode(BaseNode): |
| def __init__(self, token: Token[TV_TokenTypes], value: BaseNode): |
| super().__init__(token.lineno, token.colno, token.filename) |
| self.value = value # type: BaseNode |
| |
| class CodeBlockNode(BaseNode): |
| def __init__(self, token: Token[TV_TokenTypes]): |
| super().__init__(token.lineno, token.colno, token.filename) |
| self.lines = [] # type: T.List[BaseNode] |
| |
| class IndexNode(BaseNode): |
| def __init__(self, iobject: BaseNode, index: BaseNode): |
| super().__init__(iobject.lineno, iobject.colno, iobject.filename) |
| self.iobject = iobject # type: BaseNode |
| self.index = index # type: BaseNode |
| |
| class MethodNode(BaseNode): |
| def __init__(self, filename: str, lineno: int, colno: int, source_object: BaseNode, name: str, args: ArgumentNode): |
| super().__init__(lineno, colno, filename) |
| self.source_object = source_object # type: BaseNode |
| self.name = name # type: str |
| assert(isinstance(self.name, str)) |
| self.args = args # type: ArgumentNode |
| |
| class FunctionNode(BaseNode): |
| def __init__(self, filename: str, lineno: int, colno: int, end_lineno: int, end_colno: int, func_name: str, args: ArgumentNode): |
| super().__init__(lineno, colno, filename, end_lineno=end_lineno, end_colno=end_colno) |
| self.func_name = func_name # type: str |
| assert(isinstance(func_name, str)) |
| self.args = args # type: ArgumentNode |
| |
| class AssignmentNode(BaseNode): |
| def __init__(self, filename: str, lineno: int, colno: int, var_name: str, value: BaseNode): |
| super().__init__(lineno, colno, filename) |
| self.var_name = var_name # type: str |
| assert(isinstance(var_name, str)) |
| self.value = value # type: BaseNode |
| |
| class PlusAssignmentNode(BaseNode): |
| def __init__(self, filename: str, lineno: int, colno: int, var_name: str, value: BaseNode): |
| super().__init__(lineno, colno, filename) |
| self.var_name = var_name # type: str |
| assert(isinstance(var_name, str)) |
| self.value = value # type: BaseNode |
| |
| class ForeachClauseNode(BaseNode): |
| def __init__(self, token: Token, varnames: T.List[str], items: BaseNode, block: CodeBlockNode): |
| super().__init__(token.lineno, token.colno, token.filename) |
| self.varnames = varnames # type: T.List[str] |
| self.items = items # type: BaseNode |
| self.block = block # type: CodeBlockNode |
| |
| class IfNode(BaseNode): |
| def __init__(self, linenode: BaseNode, condition: BaseNode, block: CodeBlockNode): |
| super().__init__(linenode.lineno, linenode.colno, linenode.filename) |
| self.condition = condition # type: BaseNode |
| self.block = block # type: CodeBlockNode |
| |
| class IfClauseNode(BaseNode): |
| def __init__(self, linenode: BaseNode): |
| super().__init__(linenode.lineno, linenode.colno, linenode.filename) |
| self.ifs = [] # type: T.List[IfNode] |
| self.elseblock = None # type: T.Union[EmptyNode, CodeBlockNode] |
| |
| class UMinusNode(BaseNode): |
| def __init__(self, current_location: Token, value: BaseNode): |
| super().__init__(current_location.lineno, current_location.colno, current_location.filename) |
| self.value = value # type: BaseNode |
| |
| class TernaryNode(BaseNode): |
| def __init__(self, condition: BaseNode, trueblock: BaseNode, falseblock: BaseNode): |
| super().__init__(condition.lineno, condition.colno, condition.filename) |
| self.condition = condition # type: BaseNode |
| self.trueblock = trueblock # type: BaseNode |
| self.falseblock = falseblock # type: BaseNode |
| |
| comparison_map = {'equal': '==', |
| 'nequal': '!=', |
| 'lt': '<', |
| 'le': '<=', |
| 'gt': '>', |
| 'ge': '>=', |
| 'in': 'in', |
| 'notin': 'not in', |
| } |
| |
| # Recursive descent parser for Meson's definition language. |
| # Very basic apart from the fact that we have many precedence |
| # levels so there are not enough words to describe them all. |
| # Enter numbering: |
| # |
| # 1 assignment |
| # 2 or |
| # 3 and |
| # 4 comparison |
| # 5 arithmetic |
| # 6 negation |
| # 7 funcall, method call |
| # 8 parentheses |
| # 9 plain token |
| |
| class Parser: |
| def __init__(self, code: str, filename: str): |
| self.lexer = Lexer(code) |
| self.stream = self.lexer.lex(filename) |
| self.current = Token('eof', '', 0, 0, 0, (0, 0), None) # type: Token |
| self.getsym() |
| self.in_ternary = False |
| |
| def getsym(self) -> None: |
| try: |
| self.current = next(self.stream) |
| except StopIteration: |
| self.current = Token('eof', '', self.current.line_start, self.current.lineno, self.current.colno + self.current.bytespan[1] - self.current.bytespan[0], (0, 0), None) |
| |
| def getline(self) -> str: |
| return self.lexer.getline(self.current.line_start) |
| |
| def accept(self, s: str) -> bool: |
| if self.current.tid == s: |
| self.getsym() |
| return True |
| return False |
| |
| def accept_any(self, tids: T.Sequence[str]) -> str: |
| tid = self.current.tid |
| if tid in tids: |
| self.getsym() |
| return tid |
| return '' |
| |
| def expect(self, s: str) -> bool: |
| if self.accept(s): |
| return True |
| raise ParseException(f'Expecting {s} got {self.current.tid}.', self.getline(), self.current.lineno, self.current.colno) |
| |
| def block_expect(self, s: str, block_start: Token) -> bool: |
| if self.accept(s): |
| return True |
| raise BlockParseException(f'Expecting {s} got {self.current.tid}.', self.getline(), self.current.lineno, self.current.colno, self.lexer.getline(block_start.line_start), block_start.lineno, block_start.colno) |
| |
| def parse(self) -> CodeBlockNode: |
| block = self.codeblock() |
| self.expect('eof') |
| return block |
| |
| def statement(self) -> BaseNode: |
| return self.e1() |
| |
| def e1(self) -> BaseNode: |
| left = self.e2() |
| if self.accept('plusassign'): |
| value = self.e1() |
| if not isinstance(left, IdNode): |
| raise ParseException('Plusassignment target must be an id.', self.getline(), left.lineno, left.colno) |
| assert isinstance(left.value, str) |
| return PlusAssignmentNode(left.filename, left.lineno, left.colno, left.value, value) |
| elif self.accept('assign'): |
| value = self.e1() |
| if not isinstance(left, IdNode): |
| raise ParseException('Assignment target must be an id.', |
| self.getline(), left.lineno, left.colno) |
| assert isinstance(left.value, str) |
| return AssignmentNode(left.filename, left.lineno, left.colno, left.value, value) |
| elif self.accept('questionmark'): |
| if self.in_ternary: |
| raise ParseException('Nested ternary operators are not allowed.', |
| self.getline(), left.lineno, left.colno) |
| self.in_ternary = True |
| trueblock = self.e1() |
| self.expect('colon') |
| falseblock = self.e1() |
| self.in_ternary = False |
| return TernaryNode(left, trueblock, falseblock) |
| return left |
| |
| def e2(self) -> BaseNode: |
| left = self.e3() |
| while self.accept('or'): |
| if isinstance(left, EmptyNode): |
| raise ParseException('Invalid or clause.', |
| self.getline(), left.lineno, left.colno) |
| left = OrNode(left, self.e3()) |
| return left |
| |
| def e3(self) -> BaseNode: |
| left = self.e4() |
| while self.accept('and'): |
| if isinstance(left, EmptyNode): |
| raise ParseException('Invalid and clause.', |
| self.getline(), left.lineno, left.colno) |
| left = AndNode(left, self.e4()) |
| return left |
| |
| def e4(self) -> BaseNode: |
| left = self.e5() |
| for nodename, operator_type in comparison_map.items(): |
| if self.accept(nodename): |
| return ComparisonNode(operator_type, left, self.e5()) |
| if self.accept('not') and self.accept('in'): |
| return ComparisonNode('notin', left, self.e5()) |
| return left |
| |
| def e5(self) -> BaseNode: |
| return self.e5addsub() |
| |
| def e5addsub(self) -> BaseNode: |
| op_map = { |
| 'plus': 'add', |
| 'dash': 'sub', |
| } |
| left = self.e5muldiv() |
| while True: |
| op = self.accept_any(tuple(op_map.keys())) |
| if op: |
| left = ArithmeticNode(op_map[op], left, self.e5muldiv()) |
| else: |
| break |
| return left |
| |
| def e5muldiv(self) -> BaseNode: |
| op_map = { |
| 'percent': 'mod', |
| 'star': 'mul', |
| 'fslash': 'div', |
| } |
| left = self.e6() |
| while True: |
| op = self.accept_any(tuple(op_map.keys())) |
| if op: |
| left = ArithmeticNode(op_map[op], left, self.e6()) |
| else: |
| break |
| return left |
| |
| def e6(self) -> BaseNode: |
| if self.accept('not'): |
| return NotNode(self.current, self.e7()) |
| if self.accept('dash'): |
| return UMinusNode(self.current, self.e7()) |
| return self.e7() |
| |
| def e7(self) -> BaseNode: |
| left = self.e8() |
| block_start = self.current |
| if self.accept('lparen'): |
| args = self.args() |
| self.block_expect('rparen', block_start) |
| if not isinstance(left, IdNode): |
| raise ParseException('Function call must be applied to plain id', |
| self.getline(), left.lineno, left.colno) |
| assert isinstance(left.value, str) |
| left = FunctionNode(left.filename, left.lineno, left.colno, self.current.lineno, self.current.colno, left.value, args) |
| go_again = True |
| while go_again: |
| go_again = False |
| if self.accept('dot'): |
| go_again = True |
| left = self.method_call(left) |
| if self.accept('lbracket'): |
| go_again = True |
| left = self.index_call(left) |
| return left |
| |
| def e8(self) -> BaseNode: |
| block_start = self.current |
| if self.accept('lparen'): |
| e = self.statement() |
| self.block_expect('rparen', block_start) |
| return e |
| elif self.accept('lbracket'): |
| args = self.args() |
| self.block_expect('rbracket', block_start) |
| return ArrayNode(args, block_start.lineno, block_start.colno, self.current.lineno, self.current.colno) |
| elif self.accept('lcurl'): |
| key_values = self.key_values() |
| self.block_expect('rcurl', block_start) |
| return DictNode(key_values, block_start.lineno, block_start.colno, self.current.lineno, self.current.colno) |
| else: |
| return self.e9() |
| |
| def e9(self) -> BaseNode: |
| t = self.current |
| if self.accept('true'): |
| t.value = True |
| return BooleanNode(t) |
| if self.accept('false'): |
| t.value = False |
| return BooleanNode(t) |
| if self.accept('id'): |
| return IdNode(t) |
| if self.accept('number'): |
| return NumberNode(t) |
| if self.accept('string'): |
| return StringNode(t) |
| if self.accept('fstring'): |
| return FormatStringNode(t) |
| return EmptyNode(self.current.lineno, self.current.colno, self.current.filename) |
| |
| def key_values(self) -> ArgumentNode: |
| s = self.statement() # type: BaseNode |
| a = ArgumentNode(self.current) |
| |
| while not isinstance(s, EmptyNode): |
| if self.accept('colon'): |
| a.set_kwarg_no_check(s, self.statement()) |
| potential = self.current |
| if not self.accept('comma'): |
| return a |
| a.commas.append(potential) |
| else: |
| raise ParseException('Only key:value pairs are valid in dict construction.', |
| self.getline(), s.lineno, s.colno) |
| s = self.statement() |
| return a |
| |
| def args(self) -> ArgumentNode: |
| s = self.statement() # type: BaseNode |
| a = ArgumentNode(self.current) |
| |
| while not isinstance(s, EmptyNode): |
| potential = self.current |
| if self.accept('comma'): |
| a.commas.append(potential) |
| a.append(s) |
| elif self.accept('colon'): |
| if not isinstance(s, IdNode): |
| raise ParseException('Dictionary key must be a plain identifier.', |
| self.getline(), s.lineno, s.colno) |
| a.set_kwarg(s, self.statement()) |
| potential = self.current |
| if not self.accept('comma'): |
| return a |
| a.commas.append(potential) |
| else: |
| a.append(s) |
| return a |
| s = self.statement() |
| return a |
| |
| def method_call(self, source_object: BaseNode) -> MethodNode: |
| methodname = self.e9() |
| if not(isinstance(methodname, IdNode)): |
| raise ParseException('Method name must be plain id', |
| self.getline(), self.current.lineno, self.current.colno) |
| assert isinstance(methodname.value, str) |
| self.expect('lparen') |
| args = self.args() |
| self.expect('rparen') |
| method = MethodNode(methodname.filename, methodname.lineno, methodname.colno, source_object, methodname.value, args) |
| if self.accept('dot'): |
| return self.method_call(method) |
| return method |
| |
| def index_call(self, source_object: BaseNode) -> IndexNode: |
| index_statement = self.statement() |
| self.expect('rbracket') |
| return IndexNode(source_object, index_statement) |
| |
| def foreachblock(self) -> ForeachClauseNode: |
| t = self.current |
| self.expect('id') |
| assert isinstance(t.value, str) |
| varname = t |
| varnames = [t.value] # type: T.List[str] |
| |
| if self.accept('comma'): |
| t = self.current |
| self.expect('id') |
| assert isinstance(t.value, str) |
| varnames.append(t.value) |
| |
| self.expect('colon') |
| items = self.statement() |
| block = self.codeblock() |
| return ForeachClauseNode(varname, varnames, items, block) |
| |
| def ifblock(self) -> IfClauseNode: |
| condition = self.statement() |
| clause = IfClauseNode(condition) |
| self.expect('eol') |
| block = self.codeblock() |
| clause.ifs.append(IfNode(clause, condition, block)) |
| self.elseifblock(clause) |
| clause.elseblock = self.elseblock() |
| return clause |
| |
| def elseifblock(self, clause: IfClauseNode) -> None: |
| while self.accept('elif'): |
| s = self.statement() |
| self.expect('eol') |
| b = self.codeblock() |
| clause.ifs.append(IfNode(s, s, b)) |
| |
| def elseblock(self) -> T.Union[CodeBlockNode, EmptyNode]: |
| if self.accept('else'): |
| self.expect('eol') |
| return self.codeblock() |
| return EmptyNode(self.current.lineno, self.current.colno, self.current.filename) |
| |
| def line(self) -> BaseNode: |
| block_start = self.current |
| if self.current == 'eol': |
| return EmptyNode(self.current.lineno, self.current.colno, self.current.filename) |
| if self.accept('if'): |
| ifblock = self.ifblock() |
| self.block_expect('endif', block_start) |
| return ifblock |
| if self.accept('foreach'): |
| forblock = self.foreachblock() |
| self.block_expect('endforeach', block_start) |
| return forblock |
| if self.accept('continue'): |
| return ContinueNode(self.current) |
| if self.accept('break'): |
| return BreakNode(self.current) |
| return self.statement() |
| |
| def codeblock(self) -> CodeBlockNode: |
| block = CodeBlockNode(self.current) |
| cond = True |
| while cond: |
| curline = self.line() |
| if not isinstance(curline, EmptyNode): |
| block.lines.append(curline) |
| cond = self.accept('eol') |
| return block |