blob: e4dc2b8a0adbd4c6e63d40564284995c4aadd154 [file] [log] [blame]
# Copyright 2022 The Bazel Authors. All rights reserved.
#
# 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
#
# http://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.
"""Various things common to rules."""
load(":common/cc/cc_helper.bzl", "cc_helper")
load(
":common/python/providers.bzl",
"PyCcLinkParamsProvider",
"PyInfo",
)
load(":common/python/semantics.bzl", "IMPORTS_ATTR_SUPPORTED", "PyWrapCcInfo")
py_builtins = _builtins.internal.py_builtins
platform_common = _builtins.toplevel.platform_common
CcInfo = _builtins.toplevel.CcInfo
cc_common = _builtins.toplevel.cc_common
coverage_common = _builtins.toplevel.coverage_common
# Extensions without the dot
PYTHON_SOURCE_EXTENSIONS = ["py"]
def union_attrs(*attr_dicts, allow_none = False):
"""Helper for combining and building attriute dicts for rules.
Similar to dict.update, except:
* Duplicate keys raise an error if they aren't equal. This is to prevent
unintentionally replacing an attribute with a potentially incompatible
definition.
* None values are special: They mean the attribute is required, but the
value should be provided by another attribute dict (depending on the
`allow_none` arg).
Args:
*attr_dicts: The dicts to combine.
allow_none: bool, if True, then None values are allowed. If False,
then one of `attrs_dicts` must set a non-None value for keys
with a None value.
Returns:
dict of attributes.
"""
result = {}
missing = {}
for attr_dict in attr_dicts:
for attr_name, value in attr_dict.items():
if value == None and not allow_none:
if attr_name not in result:
missing[attr_name] = None
else:
if attr_name in missing:
missing.pop(attr_name)
if attr_name not in result or result[attr_name] == None:
result[attr_name] = value
elif value != None and result[attr_name] != value:
fail("Duplicate attribute name: '{}': existing={}, new={}".format(
attr_name,
result[attr_name],
value,
))
# Else, they're equal, so do nothing. This allows merging dicts
# that both define the same key from a common place.
if missing and not allow_none:
fail("Required attributes missing: " + csv(missing.keys()))
return result
def csv(values):
"""Convert a list of strings to comma separated value string."""
return ", ".join(sorted(values))
def filter_to_py_srcs(srcs):
"""Filters .py files from the given list of files"""
# TODO(b/203567235): Get the set of recognized extensions from
# elsewhere, as there may be others. e.g. Bazel recognizes .py3
# as a valid extension.
return [f for f in srcs if f.extension == "py"]
def collect_cc_info(ctx, extra_deps = []):
"""Collect the CcInfos from deps
Args:
ctx: rule context
extra_deps: list of additional targets to include
Returns:
Merged CcInfo from the targets.
"""
deps = ctx.attr.deps
if extra_deps:
deps = list(deps)
deps.extend(extra_deps)
return collect_cc_info_from(deps)
def collect_cc_info_from(deps):
"""Collect the CcInfos from deps
Args:
deps: (list[Target]) list of all targets to include
Returns:
(CcInfo) Merged CcInfo from the targets.
"""
cc_infos = []
for dep in deps:
if CcInfo in dep:
cc_infos.append(dep[CcInfo])
elif PyCcLinkParamsProvider in dep:
cc_infos.append(dep[PyCcLinkParamsProvider].cc_info)
elif PyWrapCcInfo and PyWrapCcInfo in dep:
# TODO(b/203567235): Google specific
cc_infos.append(dep[PyWrapCcInfo].cc_info)
return cc_common.merge_cc_infos(cc_infos = cc_infos)
def collect_runfiles(ctx, files):
"""Collects the necessary files from the rule's context.
This presumes the ctx is for a py_binary, py_test, or py_library rule.
Args:
ctx: rule ctx
files: depset of extra files to include in the runfiles.
Returns:
runfiles necessary for the ctx's target.
"""
return ctx.runfiles(
transitive_files = files,
# This little arg carries a lot of weight, but because Starlark doesn't
# have a way to identify if a target is just a File, the equivalent
# logic can't be re-implemented in pure-Starlark.
#
# Under the hood, it calls the Java `Runfiles#addRunfiles(ctx,
# DEFAULT_RUNFILES)` method, which is the what the Java implementation
# of the Python rules originally did, and the details of how that method
# works have become relied on in various ways. Specifically, what it
# does is visit the srcs, deps, and data attributes in the following
# ways:
#
# For each target in the "data" attribute...
# If the target is a File, then add that file to the runfiles.
# Otherwise, add the target's **data runfiles** to the runfiles.
#
# Note that, contray to best practice, the default outputs of the
# targets in `data` are *not* added, nor are the default runfiles.
#
# This ends up being important for several reasons, some of which are
# specific to Google-internal features of the rules.
# * For Python executables, we have to use `data_runfiles` to avoid
# conflicts for the build data files. Such files have
# target-specific content, but uses a fixed location, so if a
# binary has another binary in `data`, and both try to specify a
# file for that file path, then a warning is printed and an
# arbitrary one will be used.
# * For rules with _entirely_ different sets of files in data runfiles
# vs default runfiles vs default outputs. For example,
# proto_library: documented behavior of this rule is that putting it
# in the `data` attribute will cause the transitive closure of
# `.proto` source files to be included. This set of sources is only
# in the `data_runfiles` (`default_runfiles` is empty).
# * For rules with a _subset_ of files in data runfiles. For example,
# a certain Google rule used for packaging arbitrary binaries will
# generate multiple versions of a binary (e.g. different archs,
# stripped vs un-stripped, etc) in its default outputs, but only
# one of them in the runfiles; this helps avoid large, unused
# binaries contributing to remote executor input limits.
#
# Unfortunately, the above behavior also results in surprising behavior
# in some cases. For example, simple custom rules that only return their
# files in their default outputs won't have their files included. Such
# cases must either return their files in runfiles, or use `filegroup()`
# which will do so for them.
#
# For each target in "srcs" and "deps"...
# Add the default runfiles of the target to the runfiles. While this
# is desirable behavior, it also ends up letting a `py_library`
# be put in `srcs` and still mostly work.
# TODO(b/224640180): Reject py_library et al rules in srcs.
collect_default = True,
)
def create_py_info(ctx, direct_sources):
"""Create PyInfo provider.
Args:
ctx: rule ctx.
direct_sources: depset of Files; the direct, raw `.py` sources for the
target. This should only be Python source files. It should not
include pyc files.
Returns:
A tuple of the PyInfo instance and a depset of the
transitive sources collected from dependencies (the latter is only
necessary for deprecated extra actions support).
"""
uses_shared_libraries = False
transitive_sources_depsets = [] # list of depsets
transitive_sources_files = [] # list of Files
for target in ctx.attr.deps:
# PyInfo may not be present for e.g. cc_library rules.
if PyInfo in target:
info = target[PyInfo]
transitive_sources_depsets.append(info.transitive_sources)
uses_shared_libraries = uses_shared_libraries or info.uses_shared_libraries
else:
# TODO(b/228692666): Remove this once non-PyInfo targets are no
# longer supported in `deps`.
files = target.files.to_list()
for f in files:
if f.extension == "py":
transitive_sources_files.append(f)
uses_shared_libraries = (
uses_shared_libraries or
cc_helper.is_valid_shared_library_artifact(f)
)
deps_transitive_sources = depset(
direct = transitive_sources_files,
transitive = transitive_sources_depsets,
)
# We only look at data to calculate uses_shared_libraries, if it's already
# true, then we don't need to waste time looping over it.
if not uses_shared_libraries:
# Similar to the above, except we only calculate uses_shared_libraries
for target in ctx.attr.data:
# TODO(b/234730058): Remove checking for PyInfo in data once depot
# cleaned up.
if PyInfo in target:
info = target[PyInfo]
uses_shared_libraries = info.uses_shared_libraries
else:
files = target.files.to_list()
for f in files:
uses_shared_libraries = cc_helper.is_valid_shared_library_artifact(f)
if uses_shared_libraries:
break
if uses_shared_libraries:
break
# TODO(b/203567235): Set `uses_shared_libraries` field, though the Bazel
# docs indicate it's unused in Bazel and may be removed.
py_info = PyInfo(
transitive_sources = depset(
transitive = [deps_transitive_sources, direct_sources],
),
# TODO(b/203567235): Implement imports attribute
imports = depset() if IMPORTS_ATTR_SUPPORTED else depset(),
# NOTE: This isn't strictly correct, but with Python 2 gone,
# the srcs_version logic is largely defunct, so shouldn't matter in
# practice.
has_py2_only_sources = False,
has_py3_only_sources = False,
uses_shared_libraries = uses_shared_libraries,
)
return py_info, deps_transitive_sources
def create_instrumented_files_info(ctx):
return coverage_common.instrumented_files_info(
ctx,
source_attributes = ["srcs"],
dependency_attributes = ["deps", "data"],
extensions = PYTHON_SOURCE_EXTENSIONS,
)
def create_output_group_info(transitive_sources):
return OutputGroupInfo(
compilation_prerequisites_INTERNAL_ = transitive_sources,
compilation_outputs = transitive_sources,
)
_BOOL_TYPE = type(True)
def is_bool(v):
return type(v) == _BOOL_TYPE
def target_platform_has_any_constraint(ctx, constraints):
"""Check if target platform has any of a list of constraints.
Args:
ctx: rule context.
constraints: label_list of constraints.
Returns:
True if target platform has at least one of the constraints.
"""
for constraint in constraints:
constraint_value = constraint[platform_common.ConstraintValueInfo]
if ctx.target_platform_has_constraint(constraint_value):
return True
return False