Skip to main content

argparse_tree

Let the commandline arguments be distributed along the source directory, which can be collected into one with a single collection as a parent before initialising the command.

Installation

Using pip

pip install argparse_tree

Using pip but with a cloned repository

git clone "https://github.com/bvraghav/argparse_tree"
cd argparse_tree
pip install .

Usage

A detailed usage can be found in the folder example. The crux of the matter is illustrated below:

parser = ArgumentParser(
  parents=collect_parsers(
    '*_style1.py', 'style2/*.py',
    parent_package = __package__
  ),
)

parser.add_argument(
  '-v', '--verbose', action='store_true',
  help="Verbosity switch."
)

add_commands(
  parser, '*_command.py',
  parent_package=__package__,
  action=load_module_subparser_action(
    '*_command.py', __package__
  ),
)

Inside of style2/generic.py

parser.add_argument(
  '--style2', choices=collect_keys('style2/*.py'),
  action=load_module_action('*.py', __package__)
)

Four functions namely, collect_parsers, collect_keys, add_commands, and load_module_subparser_action are utilized to achieve the desired behaviour, that is

  1. To have a set of argument groups collected from a set of files following a convenvention in names;
  2. To allow, non-homogeneity in such groups; and
  3. To extend this behaviour for sub-parsers.

Convention

A set of files, each containing a function cli_args that returns parser information, are grouped together using a certain convention in their file names, for example, using a suffix say,*_data.py may represent different datasets. The arguments generic to all datasets may be written to generic_data.py. They are all collected using the function collect_parsers.

The convention may have simple been altered to follow a prefixed format, say data/*.py --- should work equally well.

collect_parsers

collect_parsers(
  *patterns, 
  root=None,
  parent_package=None
)

Glob the ROOT folder with PATTERNS, one at a time, and collect their parsers. If not specified, ROOT is computed, using the inspect API, to be the folder where the caller script resides.

PARENT_PACKAGE is the name of package corresponding to ROOT folder. If not specified, PARENT_PACKAGE is not used.

collect_keys

collect_keys(
  pattern,
  root=None,
  mod_to_key=utils.mod_to_key
)

Glob the ROOT folder with PATTERN and create a key corresponding to each module. Key is computed using MOD_TO_KEY functional, which follows the same signature as utils.mod_to_key.

add_commands

add_commands(
  parser, pattern, 
  *,
  root=None,
  parent_package=None,
  dest='command',
  action=None,
  mod_to_key=utils.mod_to_key
)

Create subcommands to cli using PARSER, one corresponding to each PATTERN. Command name is computed using MOD_TO_KEY functional, which follows the same signature as utils.mod_to_key.

The same convention as collect_parsers is followed for PATTERN, ROOT, and PARENT_PACKAGE.

DEST and ACTION are forwarded to argparse.ArgumentParser.add_subparsers.

load_module_subparser_action

load_module_subparser_action(
  pattern,
  package=None,
  key_to_mod=utils.key_to_mod
)

Create an argparse.Action to load a module corresponding to a user-given key, based on PATTERN, and PACKAGE using a decoder KEY_TO_MOD functional, which follows the same signature as utils.key_to_mod.

In case it is desirable to load a module corresponding to value in user-specified argument, at the time of parsing the args, use this as value of action in argparser.ArgumentParser.add_argument.

Update

Version >= 0.1.3 : Code issues separate actions corresponding to store action and subparser action, as load_module_store_action and load_module_subparser_action respectively. BREAKING: load_module_action is an alias for load_module_store_action.

Motivation

The core idea behind this project is to exploit this argparse functionality:

local_parser = argparse.ArgumentParser()
local_parser.add_argument('--foo')

global_parser = argparse.ArgumentParser(
  parents=[local_parser, ...]
)

The proof of concept can be seen in this code.

For example in machine learning training routines, there can be many models, many solvers, many datasets, each with different set of options. We can have something like:

## generic_dataset.py

def cli_parser() :
  from argparse import ArgumentParser
  parser = ArgumentParser(add_help=False)
  group = parser.add_argument_group(
    'Generic Dataset Options'
  )
  group.add_arguemnt('--foo')

  return parser

And similarly add cli parsers for a_dataset.py, b_dataset.py and so forth. Here, we have followed the norm of suffxing dataset to each python module. We can collect them into the global parser as follows:

from argparse import 
import argparse_tree as Atree

def cli_parser() :
  parser = ArgumentParser(
    parents=Atree.collect(['*_dataset.py'])
  )
  parser.add_argument(
    '--verbose', help="Verbosity switch."
  )

  return parser

We can also have an option of collection from any pattern. For example,

  • data/*.py for any module in folder data.
  • Any other pattern compatible with pathllib.Path.glob

Scratchpad

Can an instance initializer know the caller file path

Here is an experiment which has an affirmative result.

## alpha.py ##
## ----------------------------------------------------
class A :
  def __init__(self) :
    from pathlib import Path
    import inspect
    f = inspect.stack().pop(1)
    # print(f.filename)
    self.p = Path(f.filename)
## beta.py ##
## ----------------------------------------------------
from alpha import A
print (A().p)

The idea is the use this behaviour and search the caller's parent path for the relevant module patterns.

This may be interesting; but it is always better to provide an unambiguous argument instead of inspect!

Metadata

Release files for argparse-tree 0.1.3.post6

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for argparse-tree 0.1.3.post6
File Size Uploaded Provenance Actions
argparse_tree-0.1.3.post6.tar.gz 10.7 kB Without provenance Details for argparse_tree-0.1.3.post6.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for argparse-tree 0.1.3.post6
File Interpreter ABI Platform Actions
argparse_tree-0.1.3.post6-py3-none-any.whl Python 3 none any Details for argparse_tree-0.1.3.post6-py3-none-any.whl

Total release size: 20.4 kB

Release files / argparse_tree-0.1.3.post6.tar.gz

Download URL argparse_tree-0.1.3.post6.tar.gz
Size 10.7 kB
Tags Source
SHA-256 checksum
How to use checksums
e4c3181548e7bc8850890e2e7bd9edc0ceeb80472bfc9f32d1d09e5b59d0aff8
BLAKE2b-256 checksum
How to use checksums
e7794df8442aa01c2a068b1a10b4685b96c4854acda44b81aa6df4cb8df053b3
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/2.0.0 pkginfo/1.5.0.1 requests/2.24.0 setuptools/49.2.0.post20200714 requests-toolbelt/0.9.1 tqdm/4.47.0 CPython/3.8.5

Release files / argparse_tree-0.1.3.post6-py3-none-any.whl

Download URL argparse_tree-0.1.3.post6-py3-none-any.whl
Size 9.7 kB
Tags Python 3
SHA-256 checksum
How to use checksums
a8a498faeb6d9fa5fa30f9535c4cf0e85d3e00e22af87a265995b98f1d416044
BLAKE2b-256 checksum
How to use checksums
88748d8fd4539aaa0dc25440d69ce71540c66ef9b5f4b6fa079bd748a52e52d3
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/2.0.0 pkginfo/1.5.0.1 requests/2.24.0 setuptools/49.2.0.post20200714 requests-toolbelt/0.9.1 tqdm/4.47.0 CPython/3.8.5

Security for argparse-tree 0.1.3.post6

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page