""" Simple benchmarking of the adaptation manager.

This is not 'enforced' by any tests (i.e. we currently aren't bound to satisfy
any performance criteria - but in the future we might be ;^).

"""


import abc
from pprint import pprint
import time

from traits.adaptation.adaptation_manager import AdaptationManager
from traits.api import Adapter, HasTraits, Interface, provides


N_SOURCES    = 3
N_ITERATIONS = 100
N_PROTOCOLS  = 50

# Create some classes to adapt.
create_classes_to_adapt = """
class IFoo{i}(Interface):
    pass

@provides(IFoo{i})
class Foo{i}(HasTraits):
    pass
"""
for i in range(N_SOURCES):
    exec create_classes_to_adapt.format(i=i)

# The object that we will try to adapt!
foo = Foo1()

# Create a lot of other interfaces that we will adapt to.
for i in range(N_PROTOCOLS):
    exec 'class I{i}(Interface): pass'.format(i=i)

create_traits_adapter_class = """
@provides(I{target})
class IFoo{source}ToI{target}(Adapter):
    pass
"""

#  Create adapters from each 'IFooX' to all of the interfaces.
for source in range(N_SOURCES):
    for target in range(N_PROTOCOLS):
        exec create_traits_adapter_class.format(source=source, target=target)


#### traits.adaptation with Interfaces ########################################

adaptation_manager = AdaptationManager()

register_ifoox_to_ix = """

adaptation_manager.register_factory(
    factory       = IFoo{source}ToI{target},
    from_protocol = IFoo{source},
    to_protocol   = I{target}
)

"""

# We register the adapters in reversed order, so that looking for the one
# with index 0 will need traversing the whole list.
# I.e., we're considering the worst case scenario.
for source in range(N_SOURCES):
    for target in reversed(range(N_PROTOCOLS)):
        exec register_ifoox_to_ix.format(source=source, target=target)

start_time = time.time()
for _ in range(N_ITERATIONS):
    adaptation_manager.adapt(foo, I0)
time_per_iter = (time.time() - start_time) / float(N_ITERATIONS) * 1000.0
print 'apptools using Interfaces: %.3f msec per iteration' % time_per_iter


#### traits.adaptation with ABCs ##############################################

# Create some classes to adapt (using ABCs!).
for i in range(N_SOURCES):
    exec 'class FooABC{i}(object): __metaclass__ = abc.ABCMeta'.format(i=i)
    exec 'class Foo{i}(object): pass'.format(i=i)
    exec 'FooABC{i}.register(Foo{i})'.format(i=i)

# The object that we will try to adapt!
foo = Foo0()

# Create a lot of other ABCs!
for i in range(N_PROTOCOLS):
    exec 'class ABC{i}(object): __metaclass__ = abc.ABCMeta'.format(i=i)

# Create adapters from 'FooABC' to all of the ABCs.
create_abc_adapter_class = """
class FooABC{source}ToABC{target}(object):
    def __init__(self, adaptee):
        pass

ABC{target}.register(FooABC{source}ToABC{target})
"""

for source in range(N_SOURCES):
    for target in range(N_PROTOCOLS):
        exec create_abc_adapter_class.format(source=source, target=target)

# Register all of the adapters.
adaptation_manager = AdaptationManager()

register_fooxabc_to_abcx = """
adaptation_manager.register_factory(
    factory       = FooABC{source}ToABC{target},
    from_protocol = FooABC{source},
    to_protocol   = ABC{target}
)
"""

# We register the adapters in reversed order, so that looking for the one
# with index 0 will need traversing the whole list.
# I.e., we're considering the worst case scenario.
for source in range(N_SOURCES):
    for target in reversed(range(N_PROTOCOLS)):
        exec register_fooxabc_to_abcx.format(source=source, target=target)

start_time = time.time()
for _ in range(N_ITERATIONS):
    adaptation_manager.adapt(foo, ABC0)
time_per_iter = (time.time() - start_time) / float(N_ITERATIONS) * 1000.0
print 'apptools using ABCs: %.3f msec per iteration' % time_per_iter

#### EOF #######################################################################