transformer.py

import numpy as np
from core import util

class transformer:
	db = None
	def __init__(self, db):
		self.db = db

	def getS(self, obj,len):
		s0 = self.db.get(obj.__class__.__name__ + "?s0")
		s1 = self.db.get(obj.__class__.__name__ + "?s1")
		s2 = self.db.get(obj.__class__.__name__ + "?s2")
		if s0 is None:
			s0 = [0.]*len
			s1 = [0.]*len
			s2 = [0.]*len

		return s0,s1,s2
	def setS(self, obj, s0,s1,s2):
		self.db.set(obj.__class__.__name__ + "?s0",s0)
		self.db.set(obj.__class__.__name__ + "?s1",s1)
		self.db.set(obj.__class__.__name__ + "?s2",s2)

	def transform(self,obj,finalLen):

		ret = []
		if hasattr(obj, "getFeatures"):
			feats = obj.getFeatures()

			for key, val in feats.iteritems():
				if isinstance( val, str ):
					dbKey = obj.__class__.__name__ + ":" + key
					listOfThisType = self.db.get(dbKey)
					if listOfThisType is None:
						listOfThisType = []

					listOfThisType.append(val)
					self.db.set(dbKey, listOfThisType)
					feats[key] = listOfThisType.index(val)+1

			ret = feats.values()
		else:
			for k in obj.__dict__:
				var = obj.__dict__[k]
				if isinstance( var, ( int, long ) ):
					ret.append(var)
			return ret

		#print ret
		s0,s1,s2 = self.getS(obj, len(ret))	
		s0 = np.add(s0,1.)
		s1 = np.add(s1, ret)
		s2 = np.add(s2, np.multiply(ret,ret))

		mean = np.divide(s1,s0)
		if np.sum(s0) == 1.*len(ret):
			std = 1.
		else:
			std = np.divide(
				np.sqrt(
					np.subtract(
						np.multiply(s0,s2),
						np.multiply(s1,s1))),
				s0)

		self.setS(obj,s0,s1,s2)

		return util.transformFeatureVectorToCorrentShape(np.nan_to_num(np.divide(np.subtract(ret, mean), std)),finalLen)