#!/usr/bin/python3 -tt # Copyright 2012 Jussi Pakkanen # 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 ply.lex as lex import ply.yacc as yacc import nodes from coredata import MesonException class ParserException(MesonException): def __init__(self, text, lineno): MesonException.__init__(self, text) self.lineno = lineno reserved = {'true' : 'TRUE', 'false' : 'FALSE', 'if' : 'IF', 'endif' : 'ENDIF', 'else' : 'ELSE', } tokens = ['LPAREN', 'RPAREN', 'LBRACKET', 'RBRACKET', 'ATOM', 'COMMENT', 'ASSIGN', 'EQUALS', 'NEQUALS', 'COMMA', 'DOT', 'STRING', 'MULTILINE_STRING', 'INT', 'EOL_CONTINUE', 'EOL', 'COLON', ] + list(reserved.values()) t_ASSIGN = '=' t_EQUALS = '==' t_NEQUALS = '\!=' t_LPAREN = '\(' t_RPAREN = '\)' t_LBRACKET = '\[' t_RBRACKET = '\]' t_ignore_COMMENT = '\\#.*?(?=\\n)' t_COMMA = ',' t_DOT = '\.' t_COLON = ':' t_ignore = ' \t' def t_ATOM(t): '[a-zA-Z][_0-9a-zA-Z]*' t.type = reserved.get(t.value, 'ATOM') return t def t_MULTILINE_STRING(t): r"'''(.|\n)*?'''" t.value = t.value[3:-3] t.lexer.lineno += t.value.count('\n') return t def t_STRING(t): "'[^']*'" t.value = t.value[1:-1] return t def t_INT(t): '[0-9]+' t.value = int(t.value) return t def t_EOL(t): r'\n' t.lexer.lineno += 1 return t def t_EOL_CONTINUE(t): r'\\[ \t]*\n' t.lexer.lineno += 1 def t_error(t): raise ParserException("Illegal character '%s'." % t.value[0], t.lineno) # Yacc part def p_codeblock(t): 'codeblock : statement EOL codeblock' cb = t[3] cb.prepend(t[1]) t[0] = cb def p_codeblock_emptyline(t): 'codeblock : EOL codeblock' t[0] = t[2] def p_codeblock_last(t): 'codeblock : statement EOL' cb = nodes.CodeBlock(t[1].lineno()) cb.prepend(t[1]) t[0] = cb def p_expression_atom(t): 'expression : ATOM' t[0] = nodes.AtomExpression(t[1], t.lineno(1)) def p_expression_int(t): 'expression : INT' t[0] = nodes.IntExpression(t[1], t.lineno(1)) def p_expression_bool(t): '''expression : TRUE | FALSE''' if t[1] == 'true': t[0] = nodes.BoolExpression(True, t.lineno(1)) else: t[0] = nodes.BoolExpression(False, t.lineno(1)) def p_expression_string(t): 'expression : STRING' t[0] = nodes.StringExpression(t[1], t.lineno(1)) def p_expression_multiline_string(t): 'expression : MULTILINE_STRING' t[0] = nodes.StringExpression(t[1], t.lineno(1)) def p_statement_assign(t): 'statement : expression ASSIGN statement' t[0] = nodes.Assignment(t[1], t[3], t[1].lineno()) def p_statement_comparison(t): '''statement : statement EQUALS statement | statement NEQUALS statement''' t[0] = nodes.Comparison(t[1], t[2], t[3], t[1].lineno()) def p_statement_array(t): '''statement : LBRACKET args RBRACKET | LBRACKET EOL args RBRACKET''' if len(t) == 4: t[0] = nodes.ArrayStatement(t[2], t.lineno(1)) else: t[0] = nodes.ArrayStatement(t[3], t.lineno(1)) def p_statement_array2(t): '''statement : LBRACKET args EOL RBRACKET | LBRACKET EOL args EOL RBRACKET''' if len(t) == 5: t[0] = nodes.ArrayStatement(t[2], t.lineno(1)) else: t[0] = nodes.ArrayStatement(t[3], t.lineno(1)) def p_statement_func_call(t): 'statement : expression LPAREN args RPAREN' t[0] = nodes.FunctionCall(t[1], t[3], t[1].lineno()) def p_statement_method_call(t): 'statement : expression DOT expression LPAREN args RPAREN' t[0] = nodes.MethodCall(t[1], t[3], t[5], t[1].lineno()) def p_statement_if(t): 'statement : IF statement EOL codeblock elseblock ENDIF' t[0] = nodes.IfStatement(t[2], t[4], t[5], t.lineno(1)) def p_empty_else(t): 'elseblock : ' return None def p_else(t): 'elseblock : ELSE EOL codeblock' t[0] = t[3] def p_statement_expression(t): 'statement : expression' t[0] = nodes.statement_from_expression(t[1]) def p_args_multiple(t): '''args : statement COMMA args | statement COMMA EOL args | statement EOL COMMA args''' if len(t) == 5: args = t[4] else: args = t[3] args.prepend(t[1]) t[0] = args def p_kwargs_multiple(t): '''args : expression COLON statement COMMA args | expression COLON statement COMMA EOL args | expression COLON statement EOL COMMA args''' if len(t) == 7: args = t[6] else: args = t[5] args.set_kwarg(t[1], t[3]) t[0] = args def p_args_single_pos(t): 'args : statement' args = nodes.Arguments(t[1].lineno()) args.prepend(t[1]) t[0] = args def p_args_single_kw(t): 'args : expression COLON statement' a = nodes.Arguments(t[1].lineno()) a.set_kwarg(t[1], t[3]) t[0] = a def p_args_none(t): 'args :' t[0] = nodes.Arguments(t.lineno(0)) def p_error(t): if t is None: txt = 'NONE' else: txt = t.value raise ParserException('Parser errored out at: %s.' % txt, t.lineno) def test_lexer(): s = """hello = (something) # this = (that) two = ['file1', 'file2'] function(h) { stuff } obj.method(lll, \\ 'string') """ lexer = lex.lex() lexer.input(s) while True: tok = lexer.token() if not tok: break print(tok) def generate_parser_files(outputdir): code = """project('empty', 'c') """ build_ast(code, outputdir=outputdir) def build_ast(code, outputdir=None): code = code.rstrip() + '\n' lex.lex() if outputdir: parser = yacc.yacc(outputdir=outputdir) else: parser = yacc.yacc() result = parser.parse(code) return result