def test_compile_array():
code = expr_lang.compile_expr('(log2 x) + (log2 y)')
env = {'x': np.array([2, 4]), 'y': np.array([4, 8])}
res = code(env)
expected = np.array([3,5])
print "Expected:", expected, "Received:", res
assert np.all(res == expected)
def __init__(self,
feature_exprs,
model,
encoder=None,
window_sizes={
'5s': 50,
'50s': 500
},
raw_features={
'bid': online_features.best_bid,
'offer': online_features.best_offer
},
min_frames_before_prediction=2):
self.raw_features = raw_features
self.window_sizes = window_sizes
self.agg = mk_agg(raw_features, window_sizes)
self.feature_exprs = feature_exprs
# a list of expr_lang functions which map a symbol environment to the
# evaluated feature expression
self.compiled_features = [
expr_lang.compile_expr(expr) for expr in feature_exprs
]
# the set of symbols that need to be present in an environment so that
# all the feature expressions can be evaluated
self.feature_symbols = expr_lang.symbol_set(feature_exprs)
self.encoder = encoder
self.model = model
self.min_frames_before_prediction = min_frames_before_prediction
self.longest_window = np.max(window_sizes.values())
def __setstate__(self, state):
self.__dict__ = state
# reconstruct feature expr functions, since they weren't pickled
self.compiled_features = [
expr_lang.compile_expr(expr) for expr in self.feature_exprs
]
# reconstruct the aggregator which wasn't pickled
self.agg = mk_agg(state['raw_features'], state['window_sizes'])
def test_compiler_env():
env = {'x':3, 'y':4}
code = expr_lang.compile_expr('x+y')
res1 = code(env)
print "Expected 7, got", res1
assert res1 == 7
env['x'] = 4
res2 = code(env)
print "Expected 8, got", res2
assert res2 == 8
def __init__(self,
feature_exprs,
model,
encoder=None,
window_sizes={'5s': 50, '50s': 500},
raw_features={'bid': online_features.best_bid, 'offer': online_features.best_offer},
min_frames_before_prediction=2):
self.raw_features = raw_features
self.window_sizes = window_sizes
self.agg = mk_agg(raw_features, window_sizes)
self.feature_exprs = feature_exprs
# a list of expr_lang functions which map a symbol environment to the
# evaluated feature expression
self.compiled_features = [expr_lang.compile_expr(expr) for expr in feature_exprs]
# the set of symbols that need to be present in an environment so that
# all the feature expressions can be evaluated
self.feature_symbols = expr_lang.symbol_set(feature_exprs)
self.encoder = encoder
self.model = model
self.min_frames_before_prediction = min_frames_before_prediction
self.longest_window = np.max(window_sizes.values())
def test_compiler_nested():
code = expr_lang.compile_expr('(10+4)%2')
res = code({})
print "Received: ", res
assert res == 7
def test_compiler_simple():
code = expr_lang.compile_expr('3+4')
assert code({}) == 7
def __setstate__(self, state):
self.__dict__ = state
# reconstruct feature expr functions, since they weren't pickled
self.compiled_features = [expr_lang.compile_expr(expr) for expr in self.feature_exprs]
# reconstruct the aggregator which wasn't pickled
self.agg = mk_agg(state['raw_features'], state['window_sizes'])
