blob: 538a6798087ce3d0ee432bd288e5f5c554aa3e61 [file] [log] [blame]
# Copyright 2019 Google LLC
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# https://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""Checks for dependency cycles in Emboss IR."""
from compiler.util import error
from compiler.util import ir_pb2
from compiler.util import ir_util
from compiler.util import traverse_ir
def _add_reference_to_dependencies(reference, dependencies, name,
source_file_name, errors):
if reference.canonical_name.object_path[0] in {"$is_statically_sized",
"$static_size_in_bits",
"$next"}:
# This error is a bit opaque, but given that the compiler used to crash on
# this case -- for a couple of years -- and no one complained, it seems
# safe to assume that this is a rare error.
errors.append([
error.error(source_file_name, reference.source_location,
"Keyword `" + reference.canonical_name.object_path[0] +
"` may not be used in this context."),
])
return
dependencies[name] |= {ir_util.hashable_form_of_reference(reference)}
def _add_field_reference_to_dependencies(reference, dependencies, name):
dependencies[name] |= {ir_util.hashable_form_of_reference(reference.path[0])}
def _add_name_to_dependencies(proto, dependencies):
name = ir_util.hashable_form_of_reference(proto.name)
dependencies.setdefault(name, set())
return {"name": name}
def _find_dependencies(ir):
"""Constructs a dependency graph for the entire IR."""
dependencies = {}
errors = []
traverse_ir.fast_traverse_ir_top_down(
ir, [ir_pb2.Reference], _add_reference_to_dependencies,
# TODO(bolms): Add handling for references inside of attributes, once
# there are attributes with non-constant values.
skip_descendants_of={
ir_pb2.AtomicType, ir_pb2.Attribute, ir_pb2.FieldReference
},
incidental_actions={
ir_pb2.Field: _add_name_to_dependencies,
ir_pb2.EnumValue: _add_name_to_dependencies,
ir_pb2.RuntimeParameter: _add_name_to_dependencies,
},
parameters={
"dependencies": dependencies,
"errors": errors,
})
traverse_ir.fast_traverse_ir_top_down(
ir, [ir_pb2.FieldReference], _add_field_reference_to_dependencies,
skip_descendants_of={ir_pb2.Attribute},
incidental_actions={
ir_pb2.Field: _add_name_to_dependencies,
ir_pb2.EnumValue: _add_name_to_dependencies,
ir_pb2.RuntimeParameter: _add_name_to_dependencies,
},
parameters={"dependencies": dependencies})
return dependencies, errors
def _find_dependency_ordering_for_fields_in_structure(
structure, type_definition, dependencies):
"""Populates structure.fields_in_dependency_order."""
# For fields which appear before their dependencies in the original source
# text, this algorithm moves them to immediately after their dependencies.
#
# This is one of many possible schemes for constructing a dependency ordering;
# it has the advantage that all of the generated fields (e.g., $size_in_bytes)
# stay at the end of the ordering, which makes testing easier.
order = []
added = set()
for parameter in type_definition.runtime_parameter:
added.add(ir_util.hashable_form_of_reference(parameter.name))
needed = list(range(len(structure.field)))
while True:
for i in range(len(needed)):
field_number = needed[i]
field = ir_util.hashable_form_of_reference(
structure.field[field_number].name)
assert field in dependencies, "dependencies = {}".format(dependencies)
if all(dependency in added for dependency in dependencies[field]):
order.append(field_number)
added.add(field)
del needed[i]
break
else:
break
# If a non-local-field dependency were in dependencies[field], then not all
# fields would be added to the dependency ordering. This shouldn't happen.
assert len(order) == len(structure.field), (
"order: {}\nlen(structure.field: {})".format(order, len(structure.field)))
del structure.fields_in_dependency_order[:]
structure.fields_in_dependency_order.extend(order)
def _find_dependency_ordering_for_fields(ir):
"""Populates the fields_in_dependency_order fields throughout ir."""
dependencies = {}
# TODO(bolms): This duplicates work in _find_dependencies that could be
# shared.
traverse_ir.fast_traverse_ir_top_down(
ir, [ir_pb2.FieldReference], _add_field_reference_to_dependencies,
skip_descendants_of={ir_pb2.Attribute},
incidental_actions={
ir_pb2.Field: _add_name_to_dependencies,
ir_pb2.EnumValue: _add_name_to_dependencies,
ir_pb2.RuntimeParameter: _add_name_to_dependencies,
},
parameters={"dependencies": dependencies})
traverse_ir.fast_traverse_ir_top_down(
ir, [ir_pb2.Structure],
_find_dependency_ordering_for_fields_in_structure,
parameters={"dependencies": dependencies})
def _find_module_import_dependencies(ir):
"""Constructs a dependency graph of module imports."""
dependencies = {}
for module in ir.module:
foreign_imports = set()
for foreign_import in module.foreign_import:
# The prelude gets an automatic self-import that shouldn't cause any
# problems. No other self-imports are allowed, however.
if foreign_import.file_name.text or module.source_file_name:
foreign_imports |= {(foreign_import.file_name.text,)}
dependencies[module.source_file_name,] = foreign_imports
return dependencies
def _find_cycles(graph):
"""Finds cycles in graph.
The graph does not need to be fully connected.
Arguments:
graph: A dictionary whose keys are node labels. Values are sets of node
labels, representing edges from the key node to the value nodes.
Returns:
A set of sets of nodes which form strongly-connected components (subgraphs
where every node is directly or indirectly reachable from every other node).
No node will be included in more than one strongly-connected component, by
definition. Strongly-connected components of size 1, where the node in the
component does not have a self-edge, are not included in the result.
Note that a strongly-connected component may have a more complex structure
than a single loop. For example:
+-- A <-+ +-> B --+
| | | |
v C v
D ^ ^ E
| | | |
+-> F --+ +-- G <-+
"""
# This uses Tarjan's strongly-connected components algorithm, as described by
# Wikipedia. This is a depth-first traversal of the graph with a node stack
# that is independent of the call stack; nodes are added to the stack when
# they are first encountered, but not removed until all nodes they can reach
# have been checked.
next_index = [0]
node_indices = {}
node_lowlinks = {}
nodes_on_stack = set()
stack = []
nontrivial_components = set()
def strong_connect(node):
"""Implements the STRONGCONNECT routine of Tarjan's algorithm."""
node_indices[node] = next_index[0]
node_lowlinks[node] = next_index[0]
next_index[0] += 1
stack.append(node)
nodes_on_stack.add(node)
for destination_node in graph[node]:
if destination_node not in node_indices:
strong_connect(destination_node)
node_lowlinks[node] = min(node_lowlinks[node],
node_lowlinks[destination_node])
elif destination_node in nodes_on_stack:
node_lowlinks[node] = min(node_lowlinks[node],
node_indices[destination_node])
strongly_connected_component = []
if node_lowlinks[node] == node_indices[node]:
while True:
popped_node = stack.pop()
nodes_on_stack.remove(popped_node)
strongly_connected_component.append(popped_node)
if popped_node == node:
break
if (len(strongly_connected_component) > 1 or
strongly_connected_component[0] in
graph[strongly_connected_component[0]]):
nontrivial_components.add(frozenset(strongly_connected_component))
for node in graph:
if node not in node_indices:
strong_connect(node)
return nontrivial_components
def _find_object_dependency_cycles(ir):
"""Finds dependency cycles in types in the ir."""
dependencies, find_dependency_errors = _find_dependencies(ir)
if find_dependency_errors:
return find_dependency_errors
errors = []
cycles = _find_cycles(dict(dependencies))
for cycle in cycles:
# TODO(bolms): This lists the entire strongly-connected component in a
# fairly arbitrary order. This is simple, and handles components that
# aren't simple cycles, but may not be the most user-friendly way to
# present this information.
cycle_list = sorted(list(cycle))
node_object = ir_util.find_object(cycle_list[0], ir)
error_group = [
error.error(cycle_list[0][0], node_object.source_location,
"Dependency cycle\n" + node_object.name.name.text)
]
for node in cycle_list[1:]:
node_object = ir_util.find_object(node, ir)
error_group.append(error.note(node[0], node_object.source_location,
node_object.name.name.text))
errors.append(error_group)
return errors
def _find_module_dependency_cycles(ir):
"""Finds dependency cycles in modules in the ir."""
dependencies = _find_module_import_dependencies(ir)
cycles = _find_cycles(dict(dependencies))
errors = []
for cycle in cycles:
cycle_list = sorted(list(cycle))
module = ir_util.find_object(cycle_list[0], ir)
error_group = [
error.error(cycle_list[0][0], module.source_location,
"Import dependency cycle\n" + module.source_file_name)
]
for module_name in cycle_list[1:]:
module = ir_util.find_object(module_name, ir)
error_group.append(error.note(module_name[0], module.source_location,
module.source_file_name))
errors.append(error_group)
return errors
def find_dependency_cycles(ir):
"""Finds any dependency cycles in the ir."""
errors = _find_module_dependency_cycles(ir)
return errors + _find_object_dependency_cycles(ir)
def set_dependency_order(ir):
"""Sets the fields_in_dependency_order member of Structures."""
_find_dependency_ordering_for_fields(ir)
return []