A simple stack VM and assembler/compiler

LICENSE

The MIT License (MIT)

Copyright (c) 2015 miraculixx

Permission is hereby granted, free of charge, to any person obtaining a copy of
this software and associated documentation files (the "Software"), to deal in
the Software without restriction, including without limitation the rights to
use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of
the Software, and to permit persons to whom the Software is furnished to do so,
subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS
FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR
COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER
IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

microvm.py

"""
This is a quick experiment in implementing a stack machine that 
executes a set of given op codes. Just for fun.
"""
class VMError(Exception):
    def __init__(self, kind):
        self.kind = kind
        self.message = ''
    def __str__(self):
        print "MICROVM", self.kind, self.message
    def __call__(self, message):
        self.message = message
        return self
        
StackError = VMError('StackError')

class Stack(list):
    def __init__(self, parent=None):
        self.parent = parent
    def pop(self):
        try:
            return super(Stack, self).pop()
        except:
            raise StackError('stack is empty')
    
class MicroOp(object):
    """ MicroOp is the implementation of an OpCode """
    def __init__(self, bytecode, mnemonic):
        self.bytecode = bytecode
        self.mnemonic = mnemonic
        
    def execute(self, *args, **kwargs):
        pass
        
class ADD(MicroOp):
    def execute(self, *args, **kwargs):
        stack = kwargs.get('stack')
        opnum = args[0]
        opargs = [stack.pop() for i in range(opnum)]
        result = sum(opargs)
        stack.append(result)
        return result
    
class PUSH(MicroOp):
    def execute(self, *args, **kwargs):
        stack = kwargs.get('stack')
        stack.extend(args)
        
class PRINT(MicroOp):
    def execute(self, *args, **kwargs):
        stack = kwargs.get('stack')
        opnum = args[0] if args else 1
        opargs = [str(stack.pop()).replace('"', '').replace("'", '') for i in range(opnum)]
        print ' '.join(opargs)
    
OPCODES = {
    0x01 : PUSH(0x01, 'PUSH'),
    0x10 : ADD(0x10, 'ADD'),
    0x20 : PRINT(0x20, 'PRINT'),
}

MNEMONICS = { microop.mnemonic : opcode for opcode, microop in OPCODES.iteritems() }
    
class VM(object):
    """ A VM provides the execution context to run a byte code program """
    def __init__(self):
        self.stack = Stack()
    
    def run(self, code):
        for op, args in code:
            OPCODES[op].execute(*args, stack=self.stack)
            
class Compiler(object):
    """ The compiler takes source code and produces byte code """
    def _native_value(self, value):
        tries = [int, float, str]
        for t in tries:
            try:
                return t(value)
            except:
                pass
        return value
        
    def compile(self, source):
        code = []
        for loc, line in enumerate(source.split('\n')):
            if not line:
                continue
            parsed = line.split(' ', 1)
            mnemonic, args = parsed[0], parsed[1:]
            args = [self._native_value(arg) for arg in args] 
            opcode = MNEMONICS.get(mnemonic)
            if opcode:
                code.append((opcode, args))
            else:
                raise SyntaxError("%d: invalid statement >%s<" % (loc, line))
        return code

microvm2.py

"""
Enhanced version of microvm.

This now supports 

* JUMP and JUMP_TRUE supporting absolute and relative jumps
* labeled sections as JUMP targets
* associative RAM to store variables
* STORE,LOAD to transfer to/from RAM/stack
* better exceptions including pointing to the current code location
* smarter micro ops
* comments in code (# as first character in line)
"""
class VMError(Exception):
    def __init__(self, kind, context=None):
        self.kind = kind
        self.message = ''
    def __str__(self):
        s = ["MICROVM {kind} {message}".format(**self.__dict__)]
        if self.vm:
            s.append("==> {loc}: {code}".format(**dict(loc=self.vm.ic, code=self.vm.get_current_code())))
        return '\n'.join(s)
                                               
    def __call__(self, message=None, vm=None):
        self.message = message or self.message or ''
        self.vm = vm
        return self
    
class MicroOpResult(object):
    """
    the result of a micro op
    
    MicroOps pop and push 
    
    :param result: the result of the operator, if any
    :param target: the target of the operation (code memory, if any)
    :param relative: is the target a relative or absolute address, defaults to False
    """
    def __init__(self, result=None, target=None, relative=False):
        self._update(result=result, target=target, relative=relative)
        
    def __call__(self, result=None, target=None, relative=False):
        self._update(result=result, target=target, relative=relative)
        return self
        
    def _update(self, result=None, target=None, relative=False):
        self.target = target
        self.result = result
        self.relative = relative
        
StackError = VMError('StackError')

class Stack(list):
    def __init__(self, parent=None):
        self.parent = parent
    def pop(self):
        try:
            return super(Stack, self).pop()
        except:
            raise StackError('stack is empty')
    def is_empty(self):
        return len(self) == 0
    
class ARAM(dict):
    """ Associative Random Access Memory """
    pass

class MicroOp(object):
    """ MicroOp is the implementation of an OpCode """
    def __init__(self, bytecode, mnemonic):
        self.bytecode = bytecode
        self.mnemonic = mnemonic
        self.autopop = True
    
    def _parseargs(self, *args, **kwargs):
        self._args = args
        self._kwargs = kwargs
        self.vm = kwargs.get('vm')
        return self
    
    @property
    def heap(self):
        return self.vm.heap
    
    @property
    def stack(self):
        return self.vm.stack
    
    def _prepare(self, *args, **kwargs):
        self._parseargs(*args, **kwargs)
        stack = self.vm.stack
        heap = self.vm.heap
        opnum = self.opnum = args[0] if args else 0
        if self.autopop:
            opargs = self.opargs = [stack.pop() for i in xrange(opnum)]
        else:
            opargs = self.opargs = args
        self.result = MicroOpResult()
        return stack, opnum, opargs
    
    def execute(self, *args, **kwargs):
        self.stack, self.opnum, self.opargs = self._prepare(*args, **kwargs)
        pass
    
class ADD(MicroOp):
    def execute(self, *args, **kwargs):
        result = sum(self.opargs)
        self.stack.append(result)
        return self.result(result=result)
    
class JUMP(MicroOp):
    def __init__(self, *args, **kwargs):
        super(JUMP, self).__init__(*args, **kwargs)
        self.autopop = False
        
    def get_target(self):
        target = self.opargs[0]
        if isinstance(target, basestring):
            target = self.vm.labels[target]
        return target
    
    def execute(self, *args, **kwargs):
        target = self.get_target()
        relative = self.opargs[1] == "r" if len(self.opargs) > 1 else False
        return self.result(target=target, relative=relative)
    
class JUMP_IF_TRUE(JUMP):
    def execute(self, *args, **kwargs):
        relative = self.opargs[1] == "r" if len(self.opargs) > 1 else False
        target = self.get_target()
        value = self.stack.pop()
        if value:
            return self.result(result=True, target=target, relative=relative)
        else:
            return self.result(result=False)
    
class CMP(MicroOp):
    def __init__(self, *args, **kwargs):
        super(CMP, self).__init__(*args, **kwargs)
        self.autopop = False
    def execute(self):
        left = self.opargs[0]
        right = self.opargs[1] if len(self.opargs) > 1 else 1
        if isinstance(left, basestring):
            left = self.heap[left]
        else:
            left = self.stack.pop()
        if isinstance(right, basestring):
            right = self.heap[right]
        else:
            right = self.stack.pop()
        result = left == right
        self.stack.append(result)
        return self.result(result=result)
        
class PUSH(MicroOp):
    def __init__(self, *args, **kwargs):
        super(PUSH, self).__init__(*args, **kwargs)
        self.autopop = False
    def execute(self):
        self.stack.extend(self.opargs)
        return self.result
    
class STORE(MicroOp):
    def __init__(self, *args, **kwargs):
        super(STORE, self).__init__(*args, **kwargs)
        self.autopop = False
    def execute(self):
        name = self.opargs[0]
        value = self.opargs[1] if len(self._args) > 1 else self.stack.pop()
        self.heap[name] = value
        return self.result
    
class LOAD(MicroOp):
    def __init__(self, *args, **kwargs):
        super(LOAD, self).__init__(*args, **kwargs)
        self.autopop = False
    def execute(self):
        name = self.opargs[0]
        value = self.heap[name]
        self.stack.append(value)
        return self.result(result=value)
        
        
class PRINT(MicroOp):
    def _prepare(self, *args, **kwargs):
        self._parseargs(*args, **kwargs)
        stack = self.stack 
        opnum = self.opnum = self._args[0] if self._args else 1
        self.opargs = [str(stack.pop()).replace('"', '').replace("'", '') for i in xrange(opnum)]    
        self.result = MicroOpResult()
        return self.stack, self.opnum, self.opargs
            
    def execute(self):
        print ' '.join(self.opargs).strip()
        return self.result
    
# VM opcodes    
OPCODES = {
    0x01 : PUSH(0x01, 'PUSH'),
    0x02 : STORE(0x02, 'STORE'),
    0x03 : LOAD(0x03, 'LOAD'),
    0x10 : JUMP(0x10, "JUMP"),
    0x11 : JUMP_IF_TRUE(0x11, "JUMP_TRUE"),
    0x12 : CMP(0x12, "CMP"),
    0x20 : ADD(0x20, 'ADD'),
    0x30 : PRINT(0x30, 'PRINT'),
}

# compiler mnemonics to opcode mapping
MNEMONICS = { microop.mnemonic : opcode for opcode, microop in OPCODES.iteritems() }
    
class VM(object):
    """ A VM provides the execution context to run a byte code program """
    def __init__(self):
        self.stack = Stack()
        self.heap = ARAM()
        self.ic = 0
        self.code = None
        self.labels = None
        
    def run(self, program):
        code = self.code = program['bytecode']
        labels = self.labels = program['labels']
        while True:
            ic = self.ic
            if ic < len(code):
                op, args = code[ic]
                try:
                    OPCODES[op]._prepare(*args, vm=self)
                    result = OPCODES[op].execute()
                except VMError as e:
                    raise e(vm=self)
                if result.target is not None:
                    if result.relative:
                        self.ic = ic + result.target
                    else:
                        self.ic = result.target
                else:
                    self.ic += 1
            else:
                break
                
    def get_current_code(self):
        mnemonic, args = self.code[self.ic]
        return "%s %s" % (OPCODES[mnemonic].mnemonic, args)
                                            
    
            
class Compiler(object):
    """ The compiler takes source code and produces byte code """
    def _native_value(self, value):
        tries = [int, float, str]
        for t in tries:
            try:
                return t(value)
            except:
                pass
        return value
    
    def tokenize(self, source):
        tokenized = {}
        code = tokenized['code'] = []
        labels = tokenized['labels'] = {}
        label = None
        for loc, line in enumerate(source.split('\n')):
            if not line or line.startswith('#'):
                continue
            if line.endswith(':'):
                label = line.split(':')[0]
                continue 
            parsed = line.strip().split(' ', 1)
            mnemonic, args = parsed[0], parsed[1].split(',')
            code.append((mnemonic, args))
            if label:
                labels[label] = len(code) - 1
                label = None
        return tokenized
    
    def compile(self, source):
        tokenized = self.tokenized = self.tokenize(source)
        tokens = tokenized['code']
        program = {}
        bytecode = program['bytecode'] = []
        labels = program['labels'] = tokenized['labels']
        for mnemonic, args in tokens:
            args = [self._native_value(arg) for arg in args] 
            opcode = MNEMONICS.get(mnemonic)
            if opcode:
                bytecode.append((opcode, args))
            else:
                raise SyntaxError("%d: invalid statement >%s<" % (loc, line))
        return program

sample.py

source = """
PUSH 10
PUSH 20
PUSH 20
ADD 3
PUSH "result is"
PRINT 2
"""

code = Compiler().compile(source)
vm = VM()
vm.run(code)
print "stack at the end", vm.stack

sample2.py

# the following assembler code is the equivalent of the following python program
python_source = """
foo = 0
sum = 0
while foo < 2:
    sum += 10 + 20 + 20
    print "result is", sum
    foo += 1
    if foo < 2:
       print "repeat"
    else:
       print "stop"
print "finished"
"""

    
# assembler. this is hand-written for maximum performance :-)
source = """
STORE foo,0
STORE sum,0
loop:
# cumulative sum
    LOAD sum
    PUSH 10
    PUSH 20
    PUSH 20
    ADD 4
    STORE sum
    LOAD sum
# print result
    PUSH "result is"
    PRINT 2
# increment counter
    LOAD foo
    PUSH 1
    ADD 2
    STORE foo
# compare  counter to condition
    PUSH 2
    CMP foo
# repeat or stop
    JUMP_TRUE stop_it
    PUSH "repeat"
    PRINT 1
    JUMP loop
stop_it:
    PUSH "stop"
    PRINT 1
    JUMP end
end:
    PUSH "finished"
    PRINT 1
"""

compiler = Compiler()
code = compiler.compile(source)
print "tokens"
pprint(compiler.tokenized)
print "bytecode"
pprint(code)
vm = VM()
vm.run(code)

the sample assembler program in sample.py will print

result is 50

the sample assembler program in sample2.py will print

result is 50
repeat
result is 100
stop
finished

bytecode for sample2

{'bytecode': [(2, ['foo', 0]),
              (2, ['sum', 0]),
              (3, ['sum']),
              (1, [10]),
              (1, [20]),
              (1, [20]),
              (32, [4]),
              (2, ['sum']),
              (3, ['sum']),
              (1, ['"result is"']),
              (48, [2]),
              (3, ['foo']),
              (1, [1]),
              (32, [2]),
              (2, ['foo']),
              (1, [2]),
              (18, ['foo']),
              (17, ['stop_it']),
              (1, ['"repeat"']),
              (48, [1]),
              (16, ['loop']),
              (1, ['"stop"']),
              (48, [1]),
              (16, ['end']),
              (1, ['"finished"']),
              (48, [1])],
 'labels': {'end': 24, 'loop': 2, 'stop_it': 21}}

reproduce in sample2.py as

from pprint import pprint
compiler = Compiler()
code = compiler.compile(source)
print "bytecode"
pprint(code)

this is the tokenized source code representation. tokenized means that all comments have been removed and labelled sections are indexed to the first code object they represent. currently the tokenizer has very limited support for parsing arguments, it is intentionally kept very simple. note that the compiler transforms this tokenized program into it's byte code equivalent by looking up the respective MicroOp, then outputting the MicroOp's bytecode.

tokens
{'code': [('STORE', ['foo', '0']),
          ('STORE', ['sum', '0']),
          ('LOAD', ['sum']),
          ('PUSH', ['10']),
          ('PUSH', ['20']),
          ('PUSH', ['20']),
          ('ADD', ['4']),
          ('STORE', ['sum']),
          ('LOAD', ['sum']),
          ('PUSH', ['"result is"']),
          ('PRINT', ['2']),
          ('LOAD', ['foo']),
          ('PUSH', ['1']),
          ('ADD', ['2']),
          ('STORE', ['foo']),
          ('PUSH', ['2']),
          ('CMP', ['foo']),
          ('JUMP_TRUE', ['stop_it']),
          ('PUSH', ['"repeat"']),
          ('PRINT', ['1']),
          ('JUMP', ['loop']),
          ('PUSH', ['"stop"']),
          ('PRINT', ['1']),
          ('JUMP', ['end']),
          ('PUSH', ['"finished"']),
          ('PRINT', ['1'])],
 'labels': {'end': 24, 'loop': 2, 'stop_it': 21}}

reproduce in sample2.py as follows

from pprint import pprint
compiler = Compiler()
code = compiler.compile(source)
print "tokens"
pprint(compiler.tokenized)