J'ai un langage simple pour lequel j'essaie d'écrire un compilateur (oui c'est un devoir) pour compiler un langage simple que je décrirai si nécessaire en code java vm.
Il fonctionne actuellement assez bien, mais je viens de rencontrer un problème avec les ET et les OU logiques.
Les deux fonctionnent bien dans une seule condition if/while, mais si j'essaie de les enchaîner, les choses tournent mal. Corrigez-moi si je me trompe, mais je crois que AND a la priorité, mais je me demandais s'il y avait des façons logiques de les arranger ? Je pense que c'est ce que j'essaie de demander, la sortie du code java vm a juste les déclarations de comparaison et de saut l'une après l'autre (ce qui semble erroné). Je me rends compte que c'est assez abstrait, alors peut-être que ce que je cherche est un pseudo-code/algorithme pour structurer des ET et des OU enchaînés.
EDITAR : Actuellement, toute combinaison de AND et OR est traitée comme des AND. En comparant la connexion facteur/terme/expression (par rapport à booleanfactor, etc.), je pense que AND a la priorité ? C'est juste une idée.
Toutes mes excuses si je n'ai pas bien compris :/
Je me suis donc dit qu'il fallait inclure des informations pertinentes, au cas où.
compilateur
import re
import sys
# Restrictions:
# Integer constants must be short.
# Stack size must not exceed 1024.
# Integer is the only type.
# Logical operators cannot be nested.
class Scanner:
'''The interface comprises the methods lookahead and consume.
Other methods should not be called from outside of this class.'''
def __init__(self, input_file):
'''Reads the whole input_file to input_string.'''
# source code of the program to be compiled
self.input_string = input_file.read()
# index where the unprocessed part of input_string starts
self.current_char_index = 0
# a pair (most recently read token, matched substring of input_string)
self.current_token = self.get_token()
def skip_white_space(self):
'''Consumes all characters in input_string up to the next
non-white-space character.'''
if (self.current_char_index >= len(self.input_string) - 1):
# bad fix for it over-running the end of the file
return
while self.input_string[self.current_char_index].isspace():
self.current_char_index += 1
return
def get_token(self):
'''Returns the next token and the part of input_string it matched.
Returns None if there is no next token.
The characters up to the end of the token are consumed.'''
self.skip_white_space()
# find the longest prefix of input_string that matches a token
token, longest = None, ''
for (t, r) in Token.token_regexp:
match = re.match(r, self.input_string[self.current_char_index:])
if match and match.end() > len(longest):
token, longest = t, match.group()
# consume the token by moving the index to the end of the matched part
self.current_char_index += len(longest)
return (token, longest)
def lookahead(self):
'''Returns the next token without consuming it.
Returns None if there is no next token.'''
return self.current_token[0]
def consume(self, *tokens):
'''Returns the next token and consumes it, if it is in tokens.
Raises an exception otherwise.
If the token is a number or an identifier, its value is returned.'''
if self.current_token[0] not in tokens:
print('Token ' + self.current_token[0] + ' isn\'t in the tokens: ')
for token in tokens:
print(token)
raise Exception('Token is not in tokens this shouldn\'t happen much')
if self.current_token[0] == 'ID':
symbol_table.location(self.current_token[1])
value = self.current_token[1]
elif (self.current_token[0] == 'NUM'):
value = self.current_token[1]
else:
value = self.current_token[0]
self.current_token = self.get_token()
return value
class Token:
DO = 'DO';
ELSE = 'ELSE';
END = 'END';
IF = 'IF';
THEN = 'THEN';
WHILE = 'WHILE';
SEM = 'SEM';
BEC = 'BEC';
LESS = 'LESS';
EQ = 'EQ';
GRTR = 'GRTR';
LEQ = 'LEQ';
NEQ = 'NEQ';
GEQ = 'GEQ';
ADD = 'ADD';
SUB = 'SUB';
MUL = 'MUL';
DIV = 'DIV';
LPAR = 'LPAR';
RPAR = 'RPAR';
NUM = 'NUM';
ID = 'ID';
READ = 'READ';
WRITE = 'WRITE';
OR = 'OR';
AND = 'AND';
NOT = 'NOT';
# The following list gives the regular expression to match a token.
# The order in the list matters for mimicking Flex behaviour.
# Longer matches are preferred over shorter ones.
# For same-length matches, the first in the list is preferred.
token_regexp = [
(DO, 'do'),
(ELSE, 'else'),
(END, 'end'),
(IF, 'if'),
(THEN, 'then'),
(WHILE, 'while'),
(READ, 'read'),
(WRITE, 'write'),
(OR, 'or'),
(AND, 'and'),
(NOT, 'not'),
(SEM, ';'),
(BEC, ':='),
(LESS, '<'),
(EQ, '='),
(NEQ, '!='),
(GRTR, '>'),
(LEQ, '<='),
(GEQ, '>='),
(ADD, '[+]'), # + is special in regular expressions
(SUB, '-'),
(MUL, '[*]'),
(DIV, '/'),
(LPAR, '[(]'), # ( is special in regular expressions
(RPAR, '[)]'), # ) is special in regular expressions
(ID, '[a-z]+'),
(NUM, '[0-9]+'),
]
class Symbol_Table:
'''A symbol table maps identifiers to locations.'''
def __init__(self):
self.symbol_table = {}
def size(self):
'''Returns the number of entries in the symbol table.'''
return len(self.symbol_table)
def location(self, identifier):
'''Returns the location of an identifier. If the identifier is not in
the symbol table, it is entered with a new location. Locations are
numbered sequentially starting with 0.'''
if identifier in self.symbol_table:
return self.symbol_table[identifier]
index = len(self.symbol_table)
self.symbol_table[identifier] = index
return index
class Label:
def __init__(self):
self.current_label = 0
def next(self):
'''Returns a new, unique label.'''
self.current_label += 1
return 'l' + str(self.current_label)
def indent(s, level):
return ' '*level + s + '\n'
# Each of the following classes is a kind of node in the abstract syntax tree.
# indented(level) returns a string that shows the tree levels by indentation.
# code() returns a string with JVM bytecode implementing the tree fragment.
# true_code/false_code(label) jumps to label if the condition is/is not true.
# Execution of the generated code leaves the value of expressions on the stack.
class Program_AST:
def __init__(self, program):
self.program = program
def __repr__(self):
return repr(self.program)
def indented(self, level):
return self.program.indented(level)
def code(self):
program = self.program.code()
local = symbol_table.size()
java_scanner = symbol_table.location('Java Scanner')
return '.class public Program\n' + \
'.super java/lang/Object\n' + \
'.method public <init>()V\n' + \
'aload_0\n' + \
'invokenonvirtual java/lang/Object/<init>()V\n' + \
'return\n' + \
'.end method\n' + \
'.method public static main([Ljava/lang/String;)V\n' + \
'.limit locals ' + str(local) + '\n' + \
'.limit stack 1024\n' + \
'new java/util/Scanner\n' + \
'dup\n' + \
'getstatic java/lang/System.in Ljava/io/InputStream;\n' + \
'invokespecial java/util/Scanner.<init>(Ljava/io/InputStream;)V\n' + \
'astore ' + str(java_scanner) + '\n' + \
program + \
'return\n' + \
'.end method\n'
class Statements_AST:
def __init__(self, statements):
self.statements = statements
def __repr__(self):
result = repr(self.statements[0])
for st in self.statements[1:]:
result += '; ' + repr(st)
return result
def indented(self, level):
result = indent('Statement(s)', level)
for st in self.statements:
result += st.indented(level+1)
return result
def code(self):
result = ''
for st in self.statements:
result += st.code()
return result
class If_AST:
def __init__(self, boolean_expression, then):
self.boolean_expression = boolean_expression
self.then = then
def __repr__(self):
return 'if ' + repr(self.boolean_expression) + ' then ' + \
repr(self.then) + ' end'
def indented(self, level):
return indent('If-Then', level) + \
self.boolean_expression.indented(level+1) + \
self.then.indented(level+1)
def code(self):
l1 = label_generator.next()
return self.boolean_expression.code(l1) + \
self.then.code() + \
l1 + ':\n'
class If_Else_AST:
def __init__(self, boolean_expression, then, _else):
self.boolean_expression = boolean_expression;
self.then = then;
self._else = _else;
def __repr__(self):
return 'if ' + repr(self.boolean_expression) + ' then ' + \
repr(self.then) + ' else ' + \
repr(self._else) + ' end'
def indented(self, level):
return indent('If-Then-Else', level) + \
self.boolean_expression.indented(level+1) + \
self.then.indented(level+1) + \
indent('Else', level+1) + \
self._else.indented(level+1)
def code(self):
l1 = label_generator.next()
l2 = label_generator.next()
return self.boolean_expression.code(l1) + \
self.then.code() + \
'goto ' + l2 + '\n' + \
l1 + ':\n' + \
self._else.code() + \
l2 + ':\n'
class While_AST:
def __init__(self, boolean_term, body):
self.boolean_term = boolean_term
self.body = body
def __repr__(self):
return 'while ' + repr(self.boolean_term) + ' do ' + \
repr(self.body) + ' end'
def indented(self, level):
return indent('While-Do', level) + \
self.boolean_term.indented(level+1) + \
self.body.indented(level+2)
def code(self):
l1 = label_generator.next()
l2 = label_generator.next()
return l1 + ':\n' + \
self.boolean_term.code(l2) + \
self.body.code() + \
'goto ' + l1 + '\n' + \
l2 + ':\n'
class Assign_AST:
def __init__(self, identifier, expression):
self.identifier = identifier
self.expression = expression
def __repr__(self):
return repr(self.identifier) + ':=' + repr(self.expression)
def indented(self, level):
return indent('Assign', level) + \
self.identifier.indented(level+1) + \
self.expression.indented(level+1)
def code(self):
loc = symbol_table.location(self.identifier.identifier)
return self.expression.code() + \
'istore ' + str(loc) + '\n'
class Write_AST:
def __init__(self, expression):
self.expression = expression
def __repr__(self):
return 'write ' + repr(self.expression)
def indented(self, level):
return indent('Write', level) + self.expression.indented(level+1)
def code(self):
return 'getstatic java/lang/System/out Ljava/io/PrintStream;\n' + \
self.expression.code() + \
'invokestatic java/lang/String/valueOf(I)Ljava/lang/String;\n' + \
'invokevirtual java/io/PrintStream/println(Ljava/lang/String;)V\n'
class Read_AST:
def __init__(self, identifier):
self.identifier = identifier
def __repr__(self):
return 'read ' + repr(self.identifier)
def indented(self, level):
return indent('Read', level) + self.identifier.indented(level+1)
def code(self):
java_scanner = symbol_table.location('Java Scanner')
loc = symbol_table.location(self.identifier.identifier)
return 'aload ' + str(java_scanner) + '\n' + \
'invokevirtual java/util/Scanner.nextInt()I\n' + \
'istore ' + str(loc) + '\n'
class Comparison_AST:
def __init__(self, left, op, right):
self.left = left
self.op = op
self.right = right
def __repr__(self):
op = { Token.LESS:'<', Token.EQ:'=', Token.GRTR:'>',
Token.LEQ:'<=', Token.NEQ:'!=', Token.GEQ:'>=' }
return repr(self.left) + op[self.op] + repr(self.right)
def indented(self, level):
return indent(self.op, level) + \
self.left.indented(level+1) + \
self.right.indented(level+1)
def true_code(self, label):
op = { Token.LESS:'if_icmplt', Token.EQ:'if_icmpeq',
Token.GRTR:'if_icmpgt', Token.LEQ:'if_icmple',
Token.NEQ:'if_icmpne', Token.GEQ:'if_icmpge' }
return self.left.code() + \
self.right.code() + \
op[self.op] + ' ' + label + '\n'
def false_code(self, label):
# Negate each comparison because of jump to "false" label.
op = { Token.LESS:'if_icmpge', Token.EQ:'if_icmpne',
Token.GRTR:'if_icmple', Token.LEQ:'if_icmpgt',
Token.NEQ:'if_icmpeq', Token.GEQ:'if_icmplt' }
return self.left.code() + \
self.right.code() + \
op[self.op] + ' ' + label + '\n'
class Expression_AST:
def __init__(self, left, op, right):
self.left = left
self.op = op
self.right = right
def __repr__(self):
op = { Token.ADD:'+', Token.SUB:'-', Token.MUL:'*', Token.DIV:'/' }
return '(' + repr(self.left) + op[self.op] + repr(self.right) + ')'
def indented(self, level):
return indent(self.op, level) + \
self.left.indented(level+1) + \
self.right.indented(level+1)
def code(self):
op = { Token.ADD:'iadd', Token.SUB:'isub',
Token.MUL:'imul', Token.DIV:'idiv' }
return self.left.code() + \
self.right.code() + \
op[self.op] + '\n'
class Number_AST:
def __init__(self, number):
self.number = number
def __repr__(self):
return self.number
def indented(self, level):
return indent(self.number, level)
def code(self): # works only for short numbers
return 'sipush ' + self.number + '\n'
class Identifier_AST:
def __init__(self, identifier):
self.identifier = identifier
def __repr__(self):
return self.identifier
def indented(self, level):
return indent(self.identifier, level)
def code(self):
loc = symbol_table.location(self.identifier)
return 'iload ' + str(loc) + '\n'
class BooleanFactor_AST:
def __init__(self, condition, logic):
self.condition = condition
self.logic = logic
def __repr__(self):
if self.logic == False:
return 'NOT ' + repr(self.condition)
else:
return repr(self.condition)
def indented(self, level):
if self.logic == False:
return indent('NOT ', level) + self.condition.indented(level + 1)
else:
return self.condition.indented(level)
def false_code(self, label):
if self.logic == True:
return self.condition.false_code(label)
else:
return self.condition.true_code(label)
return
def true_code(self, label):
if self.logic == True:
return self.condition.true_code(label)
else:
return self.condition.false_code(label)
class BooleanTerm_AST:
def __init__(self, terms):
self.terms = terms
def __repr__(self):
result = repr(self.terms[0])
for term in self.terms[1:]:
result = result + ' AND ' + repr(term)
return result
def indented(self, level):
result = self.terms[0].indented(level)
for term in self.terms[1:]:
result = result + indent('AND', level)
result = result + term.indented(level)
return result
def code(self, label):
result = ''
for term in self.terms:
result = result + term.false_code(label)
return result
class BooleanExpression_AST:
def __init__(self, expressions):
self.expressions = expressions
def __repr__(self):
result = repr(self.expressions[0])
for expression in self.expressions[1:]:
result = result + ' OR ' + repr(expression)
return result
def indented(self, level):
result = self.expressions[0].indented(level)
indentation = 0
for expression in self.expressions[1:]:
indentation += 1
result = result + indent('OR', level + indentation)
result = result + expression.indented(level + indentation)
return result
def code(self, label):
result = ''
for expression in self.expressions:
result = result + expression.code(label)
return result
# The following methods comprise the recursive-descent parser.
def program():
sts = statements()
return Program_AST(sts)
def statements():
result = [statement()]
while scanner.lookahead() == Token.SEM:
scanner.consume(Token.SEM)
st = statement()
result.append(st)
return Statements_AST(result)
def statement():
if scanner.lookahead() == Token.IF:
return if_statement()
elif scanner.lookahead() == Token.WHILE:
return while_statement()
elif scanner.lookahead() == Token.ID:
return assignment()
elif scanner.lookahead() == Token.READ:
return read();
elif scanner.lookahead() == Token.WRITE:
return write();
else: # error
return scanner.consume(Token.IF, Token.WHILE, Token.ID)
def if_statement():
scanner.consume(Token.IF)
condition = boolean_expression()
scanner.consume(Token.THEN)
then = statements()
if scanner.lookahead() == Token.END:
scanner.consume(Token.END)
return If_AST(condition, then)
else:
scanner.consume(Token.ELSE)
_else = statements()
scanner.consume(Token.END)
return If_Else_AST(condition, then, _else)
def while_statement():
scanner.consume(Token.WHILE)
condition = boolean_expression()
scanner.consume(Token.DO)
body = statements()
scanner.consume(Token.END)
return While_AST(condition, body)
def assignment():
ident = identifier()
scanner.consume(Token.BEC)
expr = expression()
return Assign_AST(ident, expr)
def read():
scanner.consume(Token.READ)
variable = identifier()
return Read_AST(variable)
def write():
scanner.consume(Token.WRITE)
expr = expression()
return Write_AST(expr)
def comparison():
left = expression()
op = scanner.consume(Token.LESS, Token.EQ, Token.GRTR,
Token.LEQ, Token.NEQ, Token.GEQ)
right = expression()
return Comparison_AST(left, op, right)
def expression():
result = term()
while scanner.lookahead() in [Token.ADD, Token.SUB]:
op = scanner.consume(Token.ADD, Token.SUB)
tree = term()
result = Expression_AST(result, op, tree)
return result
def term():
result = factor()
while scanner.lookahead() in [Token.MUL, Token.DIV]:
op = scanner.consume(Token.MUL, Token.DIV)
tree = factor()
result = Expression_AST(result, op, tree)
return result
def factor():
if scanner.lookahead() == Token.LPAR:
scanner.consume(Token.LPAR)
result = expression()
scanner.consume(Token.RPAR)
return result
elif scanner.lookahead() == Token.NUM:
value = scanner.consume(Token.NUM)
return Number_AST(value)
elif scanner.lookahead() == Token.ID:
return identifier()
else: # error
return scanner.consume(Token.LPAR, Token.NUM, Token.ID)
def identifier():
value = scanner.consume(Token.ID)
return Identifier_AST(value)
def boolean_factor():
if scanner.lookahead() == Token.NOT:
scanner.consume(Token.NOT)
logic = False
else:
logic = True
result = comparison()
return BooleanFactor_AST(result, logic)
def boolean_term():
result = [boolean_factor()]
while scanner.lookahead() in [Token.AND]:
scanner.consume(scanner.lookahead())
temp = boolean_factor()
result.append(temp)
return BooleanTerm_AST(result)
def boolean_expression():
result = [boolean_term()]
while scanner.lookahead() in [Token.OR]:
scanner.consume(scanner.lookahead())
temp = boolean_term()
result.append(temp)
return BooleanExpression_AST(result)
# Initialise scanner, symbol table and label generator.
#scanner = Scanner(open('test.txt'))
scanner = Scanner(sys.stdin)
symbol_table = Symbol_Table()
symbol_table.location('Java Scanner') # fix a location for the Java Scanner
label_generator = Label()
# Uncomment the following to test the scanner without the parser.
# This shows a list of all tokens in the input.
#
#token = scanner.lookahead()
#while token != None:
# print(token)
# scanner.consume(token)
# token = scanner.lookahead()
#exit()
# Call the parser.
ast = program()
assert scanner.lookahead() == None
# Uncomment the following to test the parser without the code generator.
# The first line gives back the program by calling __repr__ of the AST classes.
# The second line shows the syntax tree with levels indicated by indentation.
#
#print(ast)
#print(ast.indented(0))
#exit()
# Call the code generator.
# This translates the abstract syntax tree to JVM bytecode.
# It can be assembled to a class file by Jasmin: http://jasmin.sourceforge.net/
print(ast.code())test du fichier bat
python compiler.py <test.txt> Program.j
java -Xmx100m -jar jasmin.jar Program.j
java -Xmx100m Program < testInput.txt > test_output.txtet le langage (BNF)
Program = Statements
Statements = Statement (; Statement)
Statement = If | While | Assignment
If = if Comparison then Statements end
While = while Comparison do Statements end
Assignment = identifier := Expression
Comparison = Expression Relation Expression
Relation = = | != | < | <= | > | >=
Expression = Term ((+ | -) Term)
Term = Factor ((* | /) Factor)
Factor = (Expression) | number | identifier
BooleanExpression = BooleanTerm (or BooleanTerm)*
BooleanTerm = BooleanFactor (and BooleanFactor)*
BooleanFactor = not BooleanFactor | ComparisonJe pense que c'est tout ce qui est pertinent, merci si vous essayez de m'aider sur ce point.
