def node_to_graph(node, namer, namespace, value_hints):
"""Convert Python code to equivalent TF graph mode code.
Args:
node: A Python AST node representing the code to convert.
namer: A naming.Namer object.
namespace: Dict mapping symbol names to their corresponding live objects.
value_hints: A dict containing value hints for symbols like function
parameters.
Returns:
A tuple (node, deps):
* node: A Python ast node, representing the converted code.
* deps: A set of strings, the fully qualified names of object
dependencies that this node has.
"""
# TODO(mdan): Get rid of this.
node = gradients_function.transform(node)
node = access.resolve(node)
node = live_values.resolve(node, namespace, config.PYTHON_LITERALS)
node = type_info.resolve(node, value_hints)
# TODO(mdan): Factor out common elements.
# These include:
# * keeping track of symbols that have been created
# * marking nodes (e.g. py_func wrappers) to suppress further processing
node = print_functions.transform(node)
node = call_trees.transform(node, namer, config.DEFAULT_UNCOMPILED_MODULES)
node = control_flow.transform(node, namer)
node = logical_expressions.transform(node)
node = side_effect_guards.transform(node, namer)
return node
def node_to_graph(node, namer, namespace, value_hints):
"""Convert Python code to equivalent TF graph mode code.
Args:
node: A Python AST node representing the code to convert.
namer: A naming.Namer object.
namespace: Dict mapping symbol names to their corresponding live objects.
value_hints: A dict containing value hints for symbols like function
parameters.
Returns:
A tuple (node, deps):
* node: A Python ast node, representing the converted code.
* deps: A set of strings, the fully qualified names of object
dependencies that this node has.
"""
node = access.resolve(node)
node = live_values.resolve(node, namespace, config.PYTHON_LITERALS)
node = type_info.resolve(node, value_hints)
# TODO(mdan): Factor out common elements.
# These include:
# * keeping track of symbols that have been created
# * marking nodes (e.g. py_func wrappers) to suppress further processing
node = print_functions.transform(node)
node = call_trees.transform(node, namer, config.DEFAULT_UNCOMPILED_MODULES)
node = control_flow.transform(node, namer)
node = logical_expressions.transform(node)
node = side_effect_guards.transform(node, namer)
return node
def node_to_graph(node, namer, namespace, value_hints, nocompile_decorators):
"""Convert Python code to equivalent TF graph mode code.
Args:
node: A Python AST node representing the code to convert.
namer: A naming.Namer object.
namespace: Dict mapping symbol names to their corresponding live objects.
value_hints: A dict containing value hints for symbols like function
parameters.
nocompile_decorators: A tuple containing decorators to be stripped from
functions during conversion.
Returns:
A tuple (node, deps):
* node: A Python ast node, representing the converted code.
* deps: A set of strings, the fully qualified names of object
dependencies that this node has.
"""
# TODO(mdan): Verify arguments for correctness.
# TODO(mdan): Factor out common elements.
# These include:
# * keeping track of symbols that have been created
# * marking nodes (e.g. py_func wrappers) to suppress further processing
# * code move between blocks
# * insertion of new global references
# * visiting blocks in transformers
# Certain steps, especially canonicalization, insert new symbols into the
# tree, which must be accounted. Although less efficient, it is most robust
# to re-run the analysis.
node = _static_analysis_pass(node, namespace, value_hints)
node = decorators.transform(node, nocompile_decorators)
node = break_canonicalization.transform(node, namer)
# Note: sequencing continue canonicalization before for loop one avoids
# dealing with the extra loop increment operation that the for
# canonicalization creates.
node = continue_canonicalization.transform(node, namer)
namespace['len'] = len
node = _static_analysis_pass(node, namespace, value_hints)
node = for_canonicalization.transform(node, namer)
# for_canonicalization may insert new global references.
node = builtin_functions.transform(node)
# builtin_functions may insert new global references.
namespace['print'] = print
node = _static_analysis_pass(node, namespace, value_hints)
node = print_functions.transform(node)
node = call_trees.transform(node, namer, namespace,
config.DEFAULT_UNCOMPILED_MODULES,
nocompile_decorators)
node = control_flow.transform(node, namer)
node = logical_expressions.transform(node)
node = side_effect_guards.transform(node, namer)
return node
def node_to_graph(node, namer, namespace, value_hints):
"""Convert Python code to equivalent TF graph mode code.
Args:
node: A Python AST node representing the code to convert.
namer: A naming.Namer object.
namespace: Dict mapping symbol names to their corresponding live objects.
value_hints: A dict containing value hints for symbols like function
parameters.
Returns:
A tuple (node, deps):
* node: A Python ast node, representing the converted code.
* deps: A set of strings, the fully qualified names of object
dependencies that this node has.
"""
node = access.resolve(node)
node = live_values.resolve(node, namespace, config.PYTHON_LITERALS)
node = type_info.resolve(node, value_hints)
# TODO(mdan): Factor out common elements.
# These include:
# * keeping track of symbols that have been created
# * marking nodes (e.g. py_func wrappers) to suppress further processing
node = for_canonicalization.transform(node, namer)
node = builtin_functions.transform(node)
# The transformation steps above insert new variables. Although less
# efficient, it is most robust to re-run the analysis.
# We also need to ensure the namespace contains any new references that may
# have been created.
namespace['len'] = len
namespace['print'] = print
node = access.resolve(node)
node = live_values.resolve(node, namespace, config.PYTHON_LITERALS)
node = type_info.resolve(node, value_hints)
node = print_functions.transform(node)
node = call_trees.transform(node, namer, config.DEFAULT_UNCOMPILED_MODULES)
node = control_flow.transform(node, namer)
node = logical_expressions.transform(node)
node = side_effect_guards.transform(node, namer)
return node
def test_transform(self):
def test_fn(a):
state_ops.assign(a, a + 1)
return a
node = self._parse_and_analyze(test_fn, {'state_ops': state_ops})
node = side_effect_guards.transform(node, TestNamer())
result = compiler.ast_to_object(node)
setattr(result, 'state_ops', state_ops)
# TODO(mdan): Configure the namespaces instead of doing these hacks.
ops.identity = array_ops.identity
setattr(result, 'tf', ops)
with self.test_session() as sess:
v = variables.Variable(2)
sess.run(v.initializer)
self.assertEqual(3, sess.run(result.test_fn(v)))
def test_transform(self):
def test_fn(a):
state_ops.assign(a, a + 1)
return a
node = self._parse_and_analyze(test_fn, {'state_ops': state_ops})
node = side_effect_guards.transform(node, TestNamer())
result = compiler.ast_to_object(node)
setattr(result, 'state_ops', state_ops)
# TODO(mdan): Configure the namespaces instead of doing these hacks.
ops.identity = array_ops.identity
setattr(result, 'tf', ops)
with self.test_session() as sess:
v = variables.Variable(2)
sess.run(v.initializer)
self.assertEqual(3, sess.run(result.test_fn(v)))