Version: 2.31 (dev)

Common subsystem tasks

Common tasks for Subsystems

Skipping individual targets

Many subsystems allow skipping specific targets. For example, you might have Python files that you want to not typecheck with mypy. In Pants, this is achieved with a skip_* field on the target. This is simple to implement.

Create a field for skipping your tool

from pants.engine.target import BoolField

class SkipFortranLintField(BoolField):
	alias = "skip_fortran_lint"
	default = False
	help = "If true, don't run fortran-lint on this target's code."

def rules():
	return [
		FortranSourceTarget.register_plugin_field(SkipFortranLintField),
	]

Add this field as part of your subsystems opt_out method:

from dataclasses import dataclass

from pants.engine.target import FieldSet, Target


@dataclass
class FortranLintFieldSet(FieldSet):
    required_fields = (FortranSourceField,)

    source: FortranSourceField

    @classmethod
    def opt_out(cls, tgt: Target) -> bool:
        return tgt.get(SkipFortranLintField).value

Making subsystems exportable with their default lockfile

Support depends on language backend of the subsystem

Only some language backends support pants export. These include the Python and JVM backends. Only tools which are themselves written to use a backend with this feature can be exported. For example, a Python-based tool which operates on a different language is exportable.

Make the subsystem a subclass of ExportableTool

Language backends may have done this in their Tool base class.
For example, the Python backend with PythonToolRequirementsBase and JVM with JvmToolBase are already subclasses.
```
from pants.backend.python.subsystems.python_tool_base import PythonToolBase
from pants.core.goals.resolves import ExportableTool

class FortranLint(PythonToolBase, ExportableTool):
    ...
```

def rules():
    return [
        UnionRule(ExportableTool, FortranLint)
    ]

Loading config files

Add an option to toggle config discovery:

from pants.option.subsystem import Subsystem
from pants.option.option_types import BoolOption
from pants.util.strutil import softwrap

class FortranLint(Subsystem):
    config_discovery = BoolOption(
        default=True,
        advanced=True,
        help=lambda cls: softwrap(
            f"""
            If true, Pants will include all relevant config files during runs.

            Use `[{cls.options_scope}].config` instead if your config is in a non-standard location.
            """
        ),
    )

Add an option for the configuration file itself. Several options are useful depending on what types of config files you need: FileOption, FileListOption, DirOption, DirListOption.

from pants.option.subsystem import Subsystem
from pants.option.option_types import FileOption
from pants.util.strutil import softwrap

class FortranLint(Subsystem):
    config = FileOption(
        default=None,
        advanced=True,
        help=lambda cls: softwrap(
            """
            Path to the fortran-lint config file.

            Setting this option will disable config discovery for the config file. Use this option if the config is located in a non-standard location.
            """
        ),
    )

Add a helper function to generate the ConfigFilesRequest. The check_existence field is used for config discovery. specified can also be a list for using one of the list options.

from pants.core.util_rules.config_files import ConfigFilesRequest
from pants.option.subsystem import Subsystem

class FortranLint(Subsystem):
    def config_request(self) -> ConfigFilesRequest:
        return ConfigFilesRequest(
            specified=self.config,
            specified_option_name=f"[{self.options_scope}].config",
            discovery=self.config_discovery,
            check_existence=["fortran_lint.ini"],
        )

Make a request for the config files in a rule for running the tool. Use find_config_file(ConfigFilesRequest(...)) to get the config files. The resulting ConfigFiles instance has a snapshot containing the config files (or will be empty if none are found). You can merge these with the other digests to pass the files to your Process. If a custom value was provided for the config file, you may need to pass that as an argument to the Process. You may also need to register rules from pants.core.util_rules.config_files.

from pants.core.goals.lint import LintResult
from pants.core.util_rules import config_files
from pants.core.util_rules.config_files import ConfigFilesRequest, find_config_file
from pants.core.util_rules.source_files import SourceFilesRequest, determine_source_files
from pants.engine.fs import MergeDigests
from pants.engine.intrinsics import merge_digests
from pants.engine.rules import collect_rules, concurrently, rule

@rule
async def run_fortran_lint(request: FortranlintRequest.Batch, subsystem: FortranLint) -> LintResult:
    sources, config_file = await concurrently(
        determine_source_files(
            SourceFilesRequest(fs.sources for fs in request.elements)
        ),
        find_config_file(subsystem.config_request()),
    )

    input_digest = await merge_digests(
        MergeDigests((sources.snapshot.digest, config_file.snapshot.digest))
    )

    args = []
    if subsystem.config:
        args.append(f"--config-file={subsystem.config}")

    # run your process with the digest and args

def rules():
    return [
        *collect_rules(),
        *config_files.rules(),
    ]

Skipping individual targets​

Making subsystems exportable with their default lockfile​

Loading config files​

Skipping individual targets

Making subsystems exportable with their default lockfile

Loading config files