databuild/rules.bzl
2025-04-20 12:15:01 -07:00

529 lines
16 KiB
Python

RUNFILES_PREFIX = """
# ================= BEGIN RUNFILES INIT =================
# TODO should this be extracted to shared init script
# Get the directory where the script is located
if [[ -z "${RUNFILES_DIR:-}" ]]; then
SCRIPT_DIR="$(readlink -f "${BASH_SOURCE[0]}")"
# Set RUNFILES_DIR relative to the script location
export RUNFILES_DIR="${SCRIPT_DIR}.runfiles"
fi
# --- begin runfiles.bash initialization v3 ---
# Copy-pasted from the Bazel Bash runfiles library v3.
set -uo pipefail; set +e; f=bazel_tools/tools/bash/runfiles/runfiles.bash
source "${RUNFILES_DIR:-/dev/null}/$f" 2>/dev/null || \
source "$(grep -sm1 "^$f " "${RUNFILES_MANIFEST_FILE:-/dev/null}" | cut -f2- -d' ')" 2>/dev/null || \
source "$0.runfiles/$f" 2>/dev/null || \
source "$(grep -sm1 "^$f " "$0.runfiles_manifest" | cut -f2- -d' ')" 2>/dev/null || \
source "$(grep -sm1 "^$f " "$0.exe.runfiles_manifest" | cut -f2- -d' ')" 2>/dev/null || \
{ echo>&2 "ERROR: cannot find $f"; exit 1; }; f=; set -e
# --- end runfiles.bash initialization v3 ---
# ================== END RUNFILES INIT ==================
"""
def databuild_job(
name,
configure,
execute,
visibility = None):
"""Creates a DataBuild job target with configuration and execution capabilities.
Args:
name: Name of the job target
configure: Target that implements the configuration logic
execute: Target that implements the execution logic
deps: List of other job_targets this job depends on
visibility: Visibility specification
**kwargs: Additional attributes to pass to the underlying rule
"""
_databuild_job_cfg_rule(
name = name + ".cfg",
configure = configure,
visibility = visibility,
)
# Create the main rule that serves as a provider for other targets
_databuild_job_exec_rule(
name = name + ".exec",
execute = execute,
visibility = visibility,
)
# Create a job target that configures then executes
_databuild_job_rule(
name = name,
configure = ":%s.cfg" % name,
execute = ":%s.exec" % name,
visibility = visibility,
)
def _databuild_job_cfg_impl(ctx):
configure_file = ctx.executable.configure
configure_path = ctx.attr.configure.files_to_run.executable.path
script = ctx.actions.declare_file(ctx.label.name)
ctx.actions.expand_template(
template = ctx.file._template,
output = script,
substitutions = {
"%{EXECUTABLE_PATH}": configure_path,
"%{RUNFILES_PREFIX}": RUNFILES_PREFIX,
"%{PREFIX}": "",
},
is_executable = True,
)
runfiles = ctx.runfiles(
files = [configure_file],
).merge(ctx.attr.configure.default_runfiles).merge(
ctx.attr._bash_runfiles.default_runfiles,
)
return [
DefaultInfo(
executable = script,
runfiles = runfiles,
),
]
_databuild_job_cfg_rule = rule(
implementation = _databuild_job_cfg_impl,
attrs = {
"configure": attr.label(
doc = "Target that implements the configuration logic",
executable = True,
cfg = "target",
mandatory = True,
),
"_template": attr.label(
default = "@databuild//runtime:simple_executable_wrapper.sh.tpl",
allow_single_file = True,
),
"_bash_runfiles": attr.label(
default = Label("@bazel_tools//tools/bash/runfiles"),
allow_files = True,
),
},
executable = True,
)
def _databuild_job_exec_impl(ctx):
execute_file = ctx.executable.execute
jq_file = ctx.executable._jq
script = ctx.actions.declare_file(ctx.label.name)
# Get the correct runfiles paths
jq_path = ctx.attr._jq.files_to_run.executable.path
execute_path = ctx.attr.execute.files_to_run.executable.path
ctx.actions.expand_template(
template = ctx.file._template,
output = script,
substitutions = {
"%{JQ_PATH}": jq_path,
"%{EXECUTE_PATH}": execute_path,
"%{RUNFILES_PREFIX}": RUNFILES_PREFIX,
},
is_executable = True,
)
runfiles = ctx.runfiles(
files = [jq_file, execute_file],
).merge(ctx.attr.execute.default_runfiles).merge(ctx.attr._jq.default_runfiles).merge(
ctx.attr._bash_runfiles.default_runfiles,
)
return [
DefaultInfo(
executable = script,
runfiles = runfiles,
),
]
# Define the provider
DataBuildJobInfo = provider(
doc = "Information about a DataBuild job",
fields = {
"configure": "Target that implements the configuration logic",
"execute": "Target that implements the execution logic",
"deps": "List of dependencies (other DataBuildJobInfo providers)",
},
)
_databuild_job_exec_rule = rule(
implementation = _databuild_job_exec_impl,
attrs = {
"execute": attr.label(
doc = "Target that implements the execution logic",
mandatory = True,
executable = True,
cfg = "target",
),
"_template": attr.label(
default = "@databuild//job:execute_wrapper.sh.tpl",
allow_single_file = True,
),
"_jq": attr.label(
default = "@databuild//runtime:jq",
executable = True,
cfg = "target",
),
"_bash_runfiles": attr.label(
default = Label("@bazel_tools//tools/bash/runfiles"),
allow_files = True,
),
},
executable = True,
)
def _databuild_job_impl(ctx):
"""Wraps the configure and execute targets in a shell script."""
script = ctx.actions.declare_file(ctx.label.name)
ctx.actions.write(
output = script,
is_executable = True,
content = RUNFILES_PREFIX + """
$(rlocation _main/{configure_path}) $@ | $(rlocation _main/{execute_path})
""".format(
configure_path = ctx.attr.configure.files_to_run.executable.short_path,
execute_path = ctx.attr.execute.files_to_run.executable.short_path,
),
)
runfiles = ctx.runfiles(
files = [ctx.executable.execute, ctx.executable.configure],
).merge(ctx.attr.execute.default_runfiles).merge(ctx.attr.configure.default_runfiles)
return [
DefaultInfo(
executable = script,
runfiles = runfiles,
),
DataBuildJobInfo(
configure = ctx.attr.configure,
execute = script,
),
]
_databuild_job_rule = rule(
implementation = _databuild_job_impl,
attrs = {
"configure": attr.label(
doc = "Target that implements the configuration logic",
mandatory = True,
executable = True,
# TODO all these cdf=execs are probably a problem for deployment
cfg = "exec",
),
"execute": attr.label(
doc = "Target that implements the execution logic",
mandatory = True,
executable = True,
cfg = "target",
),
},
executable = True,
)
def databuild_graph(name, jobs, lookup, visibility = None):
"""Creates a databuild graph target."""
_databuild_graph_lookup(
name = "%s.lookup" % name,
lookup = lookup,
visibility = visibility,
)
_databuild_graph_analyze(
name = "%s.analyze" % name,
lookup = "%s.lookup" % name,
jobs = jobs,
visibility = visibility,
)
_databuild_graph_exec(
name = "%s.exec" % name,
jobs = jobs,
visibility = visibility,
)
_databuild_graph_build(
name = "%s.build" % name,
analyze = "%s.analyze" % name,
exec = "%s.exec" % name,
visibility = visibility,
)
# TODO there feels like a lot of boilerplate around wrapping a target with a script - can this be simplified?
def _databuild_graph_lookup_impl(ctx):
script = ctx.actions.declare_file(ctx.label.name)
ctx.actions.expand_template(
template = ctx.file._template,
output = script,
substitutions = {
"%{RUNFILES_PREFIX}": RUNFILES_PREFIX,
"%{PREFIX}": "",
"%{EXECUTABLE_PATH}": ctx.attr.lookup.files_to_run.executable.path,
},
is_executable = True,
)
runfiles = ctx.runfiles(
files = [ctx.executable.lookup],
).merge(ctx.attr.lookup.default_runfiles).merge(
ctx.attr._bash_runfiles.default_runfiles,
)
return [
DefaultInfo(
executable = script,
runfiles = runfiles,
),
]
_databuild_graph_lookup = rule(
implementation = _databuild_graph_lookup_impl,
attrs = {
"lookup": attr.label(
doc = "Target that implements job lookup for desired partition refs",
mandatory = True,
executable = True,
cfg = "exec",
),
"_template": attr.label(
default = "@databuild//runtime:simple_executable_wrapper.sh.tpl",
allow_single_file = True,
),
"_bash_runfiles": attr.label(
default = Label("@bazel_tools//tools/bash/runfiles"),
allow_files = True,
),
},
executable = True,
)
def _databuild_graph_analyze_impl(ctx):
script = ctx.actions.declare_file(ctx.label.name)
config_paths = {
"//" + job.label.package + ":" +job.label.name:
"$(rlocation _main/" + job[DataBuildJobInfo].configure.files_to_run.executable.short_path + ")"
for job in ctx.attr.jobs
}
config_paths_str = "{" + ",".join(['\\"%s\\":\\"%s\\"' % (k, v) for k, v in config_paths.items()]) + "}"
candidate_job_env_var = "'" + ",".join([
"//" + target.label.package + ":" +target.label.name
for target in ctx.attr.jobs
]) + "'"
env_setup = """
export DATABUILD_CANDIDATE_JOBS="{candidate_job_env_var}"
export DATABUILD_MODE=plan
export DATABUILD_JOB_LOOKUP_PATH=$(rlocation _main/{lookup_path})
""".format(
candidate_job_env_var = config_paths_str,
lookup_path = ctx.attr.lookup.files_to_run.executable.short_path,
)
script_prefix = env_setup
ctx.actions.expand_template(
template = ctx.file._template,
output = script,
substitutions = {
"%{EXECUTABLE_PATH}": ctx.attr._analyze.files_to_run.executable.path,
"%{RUNFILES_PREFIX}": RUNFILES_PREFIX,
"%{PREFIX}": script_prefix,
},
is_executable = True,
)
# Gather the configure executables
configure_executables = [
job[DataBuildJobInfo].configure.files_to_run.executable
for job in ctx.attr.jobs
]
runfiles = ctx.runfiles(
files = [ctx.executable.lookup, ctx.executable._analyze] + configure_executables,
).merge(ctx.attr.lookup.default_runfiles).merge(ctx.attr._analyze.default_runfiles).merge(
ctx.attr._bash_runfiles.default_runfiles
).merge_all([job.default_runfiles for job in ctx.attr.jobs])
# Merge runfiles from all configure targets
for job in ctx.attr.jobs:
configure_target = job[DataBuildJobInfo].configure
runfiles = runfiles.merge(configure_target.default_runfiles)
return [
DefaultInfo(
executable = script,
runfiles = runfiles,
),
]
_databuild_graph_analyze = rule(
implementation = _databuild_graph_analyze_impl,
attrs = {
"lookup": attr.label(
doc = "Target that implements job lookup for desired partition refs",
mandatory = True,
executable = True,
cfg = "exec",
),
"jobs": attr.label_list(
doc = "The list of jobs that are candidates for building partitions in this databuild graph",
allow_empty = False,
),
"_template": attr.label(
default = "@databuild//graph:go_analyze_wrapper.sh.tpl",
allow_single_file = True,
),
"_bash_runfiles": attr.label(
default = Label("@bazel_tools//tools/bash/runfiles"),
allow_files = True,
),
"_analyze": attr.label(
default = "@databuild//graph:analyze",
executable = True,
cfg = "target",
)
},
executable = True,
)
def _databuild_graph_exec_impl(ctx):
script = ctx.actions.declare_file(ctx.label.name)
# Gather the execute executables
execute_executables = [
job[DataBuildJobInfo].execute
for job in ctx.attr.jobs
]
ctx.actions.expand_template(
template = ctx.file._template,
output = script,
substitutions = {
"%{EXECUTABLE_PATH}": ctx.attr._execute.files_to_run.executable.path,
"%{RUNFILES_PREFIX}": RUNFILES_PREFIX,
"%{PREFIX}": "",
},
is_executable = True,
)
runfiles = ctx.runfiles(
files = [ctx.executable._execute] + execute_executables,
).merge(ctx.attr._execute.default_runfiles).merge(
ctx.attr._bash_runfiles.default_runfiles
).merge_all([job.default_runfiles for job in ctx.attr.jobs])
# Merge runfiles from all execute targets
for job in ctx.attr.jobs:
execute_target = job[DataBuildJobInfo].execute
if hasattr(execute_target, "default_runfiles"):
runfiles = runfiles.merge(execute_target.default_runfiles)
return [
DefaultInfo(
executable = script,
runfiles = runfiles,
),
]
_databuild_graph_exec = rule(
implementation = _databuild_graph_exec_impl,
attrs = {
"jobs": attr.label_list(
doc = "The list of jobs that are candidates for building partitions in this databuild graph",
allow_empty = False,
),
"_template": attr.label(
default = "@databuild//graph:go_exec_wrapper.sh.tpl",
allow_single_file = True,
),
"_bash_runfiles": attr.label(
default = Label("@bazel_tools//tools/bash/runfiles"),
allow_files = True,
),
"_execute": attr.label(
default = "@databuild//graph:execute",
executable = True,
cfg = "target",
)
},
executable = True,
)
def _databuild_graph_build_impl(ctx):
"""Wraps the analyze and execute targets in a shell script."""
script = ctx.actions.declare_file(ctx.label.name)
ctx.actions.write(
output = script,
is_executable = True,
content = RUNFILES_PREFIX + """
$(rlocation _main/{analyze_path}) $@ | $(rlocation _main/{exec_path})
""".format(
analyze_path = ctx.attr.analyze.files_to_run.executable.short_path,
exec_path = ctx.attr.exec.files_to_run.executable.short_path,
),
)
runfiles = ctx.runfiles(
files = [ctx.executable.analyze, ctx.executable.exec],
).merge(ctx.attr.analyze.default_runfiles).merge(ctx.attr.exec.default_runfiles)
return [
DefaultInfo(
executable = script,
runfiles = runfiles,
),
]
_databuild_graph_build = rule(
implementation = _databuild_graph_build_impl,
attrs = {
"analyze": attr.label(
doc = "Target that implements the graph analysis logic",
mandatory = True,
executable = True,
cfg = "exec",
),
"exec": attr.label(
doc = "Target that implements the graph execution logic",
mandatory = True,
executable = True,
cfg = "target",
),
},
executable = True,
)
#def _graph_impl(name):
# """
#
# """
#
# # Lets do this
# pass
#
#databuild_graph = rule(
# implementation = _graph_impl,
# attrs = {
# "jobs": attr.label_list(
# doc = "The list of jobs that are candidates for building partitions in this databuild graph",
# allow_empty = False,
# ),
# "plan": attr.label(
# doc = "The binary that is run to produce a `JobGraph` that builds the requested partition refs",
# executable = True,
# cfg = "exec",
# ),
# },
#)