import sys
from collections import defaultdict
from itertools import groupby

# TODO: Keep trail of message travelling through the rules

class Element(object):
	def __init__(self):
		self.outputs = []
		
	def display(self, indent):
		print ("\t" * indent) + self.get_description()
		for output in self.outputs:
			output.display(indent + 1)
	
class Bin(Element):
	def __init__(self, name):
		Element.__init__(self)
		self.name = name
		
	def get_description(self):
		return "[Bin] %s" % self.name
		
	def process(self, message):
		self.forward(message)
		
	def forward(self, message):
		for output in self.outputs:
			output.process(message)
	
class Rule(Element):
	def __init__(self, input_):
		Element.__init__(self)
		self.input_ = input_
		
	def process(self, message):
		self.forward(message)
		
	def forward(self, message):
		for output in self.outputs:
			output.process(message)
	
class Filter(Rule):
	def __init__(self, input_, rule):
		Rule.__init__(self, input_)
		self.rule = rule
		
		# Rules:
		#  Boolean 'and' has precedence over 'or'
		#  Enclosure in parentheses means creating a new FilterExpressionGroup
		#  Having 'and' and 'or' operators in the same group means a new FilterExpressionGroup is created for every 'and' chain, to retain precedence
		#  Variable accessors are prefixed with $
		#  Strings are enclosed in "quotes"
		
		rule_length = len(rule)
		idx = 0
		buff = ""
		in_expression = False
		current_element = {}
		element_list = defaultdict(list)
		operator_list = defaultdict(list)
		current_depth = 0
		
		while idx < rule_length:
			char = rule[idx]
			#print len(buff), len(rule), idx, buff
			if char == "(" and in_expression == False:
				# New group encountered
				#print "START GROUP %d" % current_depth
				current_depth += 1
			elif char == ")" and in_expression == False:
				# End statement, Process list of elements
				element_list[current_depth].append(create_filter_expression(buff))
				# Add elements to group object
				group = create_group(element_list[current_depth], operator_list[current_depth])
				element_list[current_depth - 1].append(group)
				
				element_list[current_depth] = []
				operator_list[current_depth] = [] # Clear out lists to prevent working with stale data
					
				#print "-- GR: %s" % group
				buff = ""
				current_depth -= 1
				#print "END GROUP %d" % current_depth
			elif char == '"':
				in_expression = not in_expression
				buff += '"'
			elif not in_expression and char == "o" and idx + 2 < rule_length and rule[idx+1:idx+2] == "r" and len(buff) > 0 and (buff[-1] == " " or buff[-1] == ")"):
				# End statement, Boolean OR
				if buff.strip() != "":
					element_list[current_depth].append(create_filter_expression(buff))
				operator_list[current_depth].append(OR)
				buff = ""
				idx += 1 # We read ahead one position extra
			elif not in_expression and char == "a" and idx + 3 < rule_length and rule[idx+1:idx+3] == "nd" and len(buff) > 0 and (buff[-1] == " " or buff[-1] == ")"):
				# End statement, Boolean AND
				if buff.strip() != "":
					element_list[current_depth].append(create_filter_expression(buff))
				operator_list[current_depth].append(AND)
				buff = ""
				idx += 2 # We read ahead two positions extra
			else:
				buff += char
				
			idx += 1
		
		if current_depth > 0:
			raise Exception("Missing %d closing parenthese(s)." % current_depth)
		elif current_depth < 0:
			raise Exception("Missing %d opening parenthese(s)." % (0 - current_depth))
		
		# If there's anything left in the buffer, it's probably a statement we still need to process.	
		if buff.strip() != "":
			element_list[current_depth].append(create_filter_expression(buff))
			
		if len(element_list[current_depth]) > 1:
			# Multiple elements, need to encapsulate in a group
			root_element = create_group(element_list[current_depth], operator_list[current_depth])
		elif len(element_list[current_depth]) == 1:
			root_element = element_list[current_depth][0]
		else:
			# FIXME: Proper exception
			raise Exception("No root elements?!")
		
		self.root = root_element
		#print repr(root_element)
		
	def get_description(self):
		return "[Filter] %s" % self.rule
		
	def evaluate(self, message):
		return self.root.evaluate(message)
		
	def process(self, message):
		if self.evaluate(message):
			self.forward(message)

def create_group(elements, operators):
	group = FilterExpressionGroup()
		
	# Process operators
	if len(elements) > 1:
		# Check if the operators vary
		operator_discrepancy = (len(set(operators)) > 1)
		
		if operator_discrepancy:
			# We'll need to find the 'and' chains and push them into separate child groups
			idx = 0
			sieve = [True for x in xrange(0, len(elements))]
			final_list = []
			
			for operator, items in groupby(operators):
				items = list(items)
				
				start = idx
				end = idx + len(items) + 1
				relevant_elements = elements[start:end]
				
				if operator == AND:
					for i in xrange(start, end):
						# Mark as processed
						sieve[i] = False
					for i in [x for x in xrange(0, end) if sieve[x] is True]:
						final_list.append(elements[i])
						sieve[i] = False
					final_list.append(create_group(relevant_elements, [AND for x in xrange(0, end - start)]))
					
				idx += len(items)
			
			for element in final_list:
				group.add(element)
				
			group.relation = OR  # Hardcoded, because all AND chains are taken care of above...
		else:
			for element in elements:
				group.add(element)
			group.relation = operators[0]
	else:
		group.add(elements[0])
		
	return group

def create_filter_expression(buff):
	# TODO: Use shlex split because of spaces in strings?
	left, operator, right = [x.strip() for x in buff.split(None, 2)]
	
	if left[0] == '"' and left[-1] == '"':
		left_obj = FilterExpressionString(left[1:-1])
	elif left[0] == "$":
		if "[" in left[1:] and left[-1] == "]":
			name, scope = left[1:-1].split("[", 1)
		else:
			name = left[1:]
			scope = None
			
		left_obj = FilterExpressionVariable(name, scope)
	else:
		raise Exception("Unrecognized operand type") # No other types supported yet...
	
	if right[0] == '"' and right[-1] == '"':
		right_obj = FilterExpressionString(right[1:-1])
	elif right[0] == "$":
		if "[" in right[1:] and right[-1] == "]":
			name, scope = right[1:-1].split("[", 1)
		else:
			name = right[1:]
			scope = None
			
		right_obj = FilterExpressionVariable(name, scope)
	else:
		raise Exception("Unrecognized operand type") # No other types supported yet...
	
	operators = {
		"=": EQUALS,
		"==": EQUALS,
		"!=": NOT_EQUALS,
		">": MORE_THAN,
		"<": LESS_THAN,
		">=": MORE_THAN_OR_EQUALS,
		"<=": LESS_THAN_OR_EQUALS,
		"has": HAS
	}
	
	try:
		operator_type = operators[operator]
	except KeyError, e:
		raise Exception("Invalid operator")
	
	expression = FilterExpression(left_obj, operator_type, right_obj)
	return expression
	# Broken?
	#print expression

class BinReference(Rule):
	def __init__(self, input_, name):
		Rule.__init__(self, input_)
		self.bin_name = name
		
	def get(self):
		try:
			return bins[self.bin_name]
		except KeyError, e:
			new_bin = Bin(self.bin_name)
			bins[self.bin_name] = new_bin
			return new_bin
			
	def get_description(self):
		return "[BinRef] %s" % self.bin_name
		
	def process(self, message):
		# TODO: Actually deposit into bin
		print "DEPOSITED INTO BIN %s" % self.name

class NodeReference(Rule):
	def __init__(self, input_, name):
		Rule.__init__(self, input_)
		self.node_name = name
		
	def get_description(self):
		return "[NodeRef] %s" % self.node_name
		
	def process(self, message):
		# TODO: Actually forward to node
		print "FORWARDED TO NODE %s" % self.node_name
		
class MethodReference(Rule):
	def __init__(self, input_, name):
		Rule.__init__(self, input_)
		self.method_name = name
		
	def get_description(self):
		return "[MethodRef] %s" % self.method_name
		
	def process(self, message):
		# TODO: Actually pass to method
		print "PASSED TO METHOD %s" % self.method_name
		
class DistributorReference(Rule):
	def __init__(self, input_, name, args):
		Rule.__init__(self, input_)
		self.distributor_name = name
		self.args = args
		
	def get_description(self):
		return "[DistRef] %s (%s)" % (self.distributor_name, ", ".join(self.args))
		
	def process(self, message):
		# TODO: Actually distribute
		# TODO: Parse args
		pass#print "DISTRIBUTED TO %s" % self.get_description()

NONE = 0
AND = 1
OR = 2

EQUALS = 3
NOT_EQUALS = 4
LESS_THAN = 5
MORE_THAN = 6
LESS_THAN_OR_EQUALS = 7
MORE_THAN_OR_EQUALS = 8
HAS = 9

class FilterExpression(object):
	def __init__(self, left, operator, right):
		self.left = left
		self.operator = operator
		self.right = right
		
	def evaluate(self, message):
		if self.operator == EQUALS:
			return (self.left.value(message) == self.right.value(message))
		elif self.operator == NOT_EQUALS:
			return (self.left.value(message) != self.right.value(message))
		elif self.operator == LESS_THAN:
			return (self.left.value(message) < self.right.value(message))
		elif self.operator == MORE_THAN:
			return (self.left.value(message) > self.right.value(message))
		elif self.operator == LESS_THAN_OR_EQUALS:
			return (self.left.value(message) <= self.right.value(message))
		elif self.operator == MORE_THAN_OR_EQUALS:
			return (self.left.value(message) >= self.right.value(message))
		elif self.operator == HAS:
			if is_instance(self.left, basestring):
				return (self.right.value(message) in self.left.value(message))  # Substring comparison
			else:
				return (self.right.value(message) in self.left.values(message))  # In-array check
		else:
			# TODO: Log error
			return False
			
	def __repr__(self):
		if self.operator == EQUALS:
			opname = "EQUALS"
		elif self.operator == NOT_EQUALS:
			opname = "NOT EQUALS"
		elif self.operator == LESS_THAN:
			opname = "LESS THAN"
		elif self.operator == MORE_THAN:
			opname = "MORE THAN"
		elif self.operator == LESS_THAN_OR_EQUALS:
			opname = "LESS THAN OR EQUAL"
		elif self.operator == MORE_THAN_OR_EQUALS:
			opname = "MORE THAN OR EQUAL"
		elif self.operator == HAS:
			opname = "HAS"
		else:
			opname = "?"
			
		return "<FE %s [%s] %s>" % (repr(self.left), opname, repr(self.right))
	
class FilterExpressionGroup(object):
	def __init__(self):
		self.elements = []
		self.relation = NONE
	
	def add(self, element):
		self.elements.append(element)
	
	def evaluate(self, message):
		if self.relation == AND:
			for element in self.elements:
				if element.evaluate(message) != True:
					return False
			return True
		elif self.relation == OR:
			for element in self.elements:
				if element.evaluate(message) == True:
					return True
			return False
		else:
			# TODO: Log error
			return False
			
	def __repr__(self):
		if self.relation == AND:
			relname = "AND"
		elif self.relation == OR:
			relname = "OR"
		else:
			relname = "?"
		return "<FEGroup %s (%s)>" % (relname, ", ".join(repr(x) for x in self.elements))

class FilterExpressionElement(object):
	def select_value(self, message, scope, name, multiple=False):
		if scope == "tags":
			return_value = message.tags
		elif scope == "type":
			return_value = message.type_
		elif scope == "source":
			return_value = message.source
		elif scope == "chain":
			return_value = message.chain
		elif scope == "attr":
			return_value = self.select_attribute(message, name, multiple=multiple)
		else:
			raise Exception("Unknown scope")  # FIXME: Proper exception
			
		if isinstance(return_value, basestring):
			return return_value
		elif multiple == False and len(return_value) > 0:
			return return_value[0]
		else:
			raise Exception("No value found") # FIXME: Proper exception
			
	def select_attribute(self, message, query, multiple=False):
		segments = query.split("/")
		current_object = message.data
		
		for segment in segments:
			try:
				current_object = current_object[segment]
			except KeyError, e:
				raise Exception("Unknown attribute")  # FIXME: Proper exception
				
		return current_object
	
class FilterExpressionVariable(FilterExpressionElement):
	def __init__(self, scope, name=None):
		self.scope = scope
		self.name = name
		# TODO: name path parsing
		
	def value(self, message):
		return self.select_value(message, self.scope, self.name, multiple=False)
	
	def values(self, message):
		return self.select_value(message, self.scope, self.name, multiple=True)
		
	def __repr__(self):
		return "<FEVar %s/%s>" % (self.scope, self.name)
	
class FilterExpressionString(FilterExpressionElement):
	def __init__(self, string):
		self.string = string
		
	def value(self, message):
		return self.string
		
	def values(self, message):
		return [self.string]
		
	def __repr__(self):
		return "<FEString \"%s\">" % self.string

def create_rule(buff, input_):
	buff = buff.strip()
	if buff[0] == "*":
		# Node reference
		new_obj = NodeReference(input_, buff[1:])
	elif buff[0] == "@":
		# Method call
		new_obj = MethodReference(input_, buff[1:])
	elif buff[0] == "#":
		# Bin reference
		new_obj = BinReference(input_, buff[1:])
	elif buff[0] == ":":
		# Distributor
		if "(" in buff and buff[-1:] == ")":
			name, arglist = buff[1:-1].split("(", 1)
			args = [x.strip() for x in arglist.split(",")]
		else:
			name = buff[1:]
			args = []
		new_obj = DistributorReference(input_, name, args)
	else:
		# Filter
		new_obj = Filter(input_, buff)
		
	input_.outputs.append(new_obj)
	return new_obj

f = open(sys.argv[1])
rulebook = f.read()
f.close()
rulebook_length = len(rulebook)

# Main parsing loop
idx = 0
tab_count = 0
current_level = 0
current_rule = {}
target_rule = None
statement_cache = None
new_line = True
multiple_statements = False
buff = ""
bins = {}

while idx < rulebook_length:
	char = rulebook[idx]
	
	if char == "\t":
		if buff == "":
			new_line = True
			tab_count += 1
		else:
			buff += char
	else:
		if new_line == True:
			if tab_count > current_level + 1:
				raise Exception("Incorrect indentation")
			
			if rulebook[idx:idx+2] == "=>":
				# Skip over this, it's optional at the start of a line
				idx += 2
				continue
				
			new_line = False
		
		if char == "\r":
			idx += 1
			continue # Ignore, we don't want carriage returns
		elif char == "\n":
			# Process
			if buff.strip() == "":
				# Skip empty lines, we don't care about them
				idx += 1
				tab_count = 0
				continue
				
			current_level = tab_count
			tab_count = 0
			
			if current_level == 0:
				bin_name = buff.strip()
				new_bin = Bin(bin_name)
				current_rule[current_level] = new_bin
				bins[bin_name] = new_bin
			else:
				if multiple_statements == True:
					new_rule = create_rule(buff, statement_cache)
				else:
					new_rule = create_rule(buff, current_rule[current_level - 1])
					
				current_rule[current_level] = new_rule
				
			buff = ""
			new_line = True
			multiple_statements = False
		elif char == "=" and rulebook[idx + 1] == ">":
			# Next rule, same line!
			if multiple_statements == True:
				statement_cache = create_rule(buff, statement_cache)
			else:
				multiple_statements = True
				statement_cache = create_rule(buff, current_rule[tab_count - 1])
			
			buff = ""
			idx += 1 # We read one extra character ahead
		else:
			# TODO: add entire chunks at once for speed
			buff += char
			
	idx += 1
	# TODO: detect infinite loops via bins!

for bin_name, bin_ in bins.iteritems():
	pass#bin_.display(0)



class Message(object):
	def __init__(self):
		self.id_ = ""
		self.type_ = "none"
		self.tags = []
		self.source = ""
		self.chain = []
		self.data = {}
		
	def set_data(self, data):
		self.id_ = data['id']
		self.type_ = data['type'] 
		self.tags = data['tags']
		self.source = data['source']
		self.chain = data['chain']
		self.data = data['payload']
		


import timeit

m = Message()
m.set_data({
	"id": "qwert-yuiop-61238-10842",
	"type": "task",
	"tags": ["convert", "mpeg"],
	"source": "abcde-fghij-00000-00008",
	"chain": ["abcde-fghij-00000-00005", "abcde-fghij-00000-00006"],
	"payload": {
		"command": "convert",
		"category": "video",
		"original_filetype": "mpg"
	}
})

bins['remote'].process(m)
#print min(timeit.Timer("bins['remote'].process(m)", "from __main__ import bins, m").repeat(7, 1000))