$λ provides an alternative to argparse
based on parser combinators and functional first principles. Arguably, $λ
is way more expressive than any reasonable
person would ever need... but even if it's not the parser that we need, it's the parser we deserve.
Installation
pip install dollar-lambda
Highlights
$λ
comes with syntactic sugar that can make building parsers completely boilerplate-free.
For complex parsing situations that exceed the expressive capacity of this syntax,
the user can also drop down to the lower-level syntax that lies behind the sugar, which can
handle any reasonable amount of logical complexity.
decorator
For the vast majority of parsing patterns,
@command
is the most concise way to define a parser:
from dollar_lambda import command
@command()
def main(x: int, dev: bool = False, prod: bool = False):
print(dict(x=x, dev=dev, prod=prod))
Here is the help text generated by this parser:
main("-h")
usage: -x X --dev --prod
dev: (default: False)
prod: (default: False)
Ordinarily you provide no arguments to main
and it would get them from the command line.
The explicit arguments in this Readme are for demonstration purposes only.
Here is how the main function handles input:
main("-x", "1", "--dev")
{'x': 1, 'dev': True, 'prod': False}
Use the parsers
argument to add custom logic using the lower-level syntax:
from dollar_lambda import flag
@command(parsers=dict(kwargs=flag("dev") | flag("prod")))
def main(x: int, **kwargs):
print(dict(x=x, **kwargs))
This parser requires either a --dev
or --prod
flag and maps it to the kwargs
argument:
main("-h")
usage: -x X [--dev | --prod]
This assigns {'dev': True}
to the kwargs
argument:
main("-x", "1", "--dev")
{'x': 1, 'dev': True}
This assigns {'prod': True}
to the kwargs
argument:
main("-x", "1", "--prod")
{'x': 1, 'prod': True}
This fails because the parser requires one or the other:
main("-x", "1")
usage: -x X [--dev | --prod]
The following arguments are required: --dev
For many programs, a user will want to use one entrypoint for one set of
arguments, and another for another set of arguments. Returning to our example,
let's say we wanted to execute prod_function
when the user provides the
--prod
flag, and dev_function
when the user provides the --dev
flag:
from dollar_lambda import CommandTree
tree = CommandTree()
@tree.command()
def base_function(x: int):
print("Ran base_function with arguments:", dict(x=x))
@base_function.command()
def prod_function(x: int, prod: bool):
print("Ran prod_function with arguments:", dict(x=x, prod=prod))
@base_function.command()
def dev_function(x: int, dev: bool):
print("Ran dev_function with arguments:", dict(x=x, dev=dev))
Let's see how this parser handles different inputs.
If we provide the --prod
flag, $λ
automatically invokes
prod_function
with the parsed arguments:
tree(
"-x", "1", "--prod"
)
Ran prod_function with arguments: {'x': 1, 'prod': True}
If we provide the --dev
flag, $λ
invokes dev_function
:
tree("-x", "1", "--dev")
Ran dev_function with arguments: {'x': 1, 'dev': True}
With this configuration, the parser will run base_function
if neither
--prod
nor --dev
are given:
tree("-x", "1")
Ran base_function with arguments: {'x': 1}
There are many other ways to use CommandTree
.
To learn more, we recommend the CommandTree
tutorial.
Lower-level syntax
@command
and CommandTree
cover many use cases,
but they are both syntactic sugar for a lower-level interface that is far
more expressive.
Suppose you want to implement a parser that first tries to parse an option
(a flag that takes an argument),
-x X
and if that fails, tries to parse the input as a variadic sequence of
floats:
from dollar_lambda import argument, option
p = option("x", type=int) | argument("y", type=float).many()
We go over this syntax in greater detail in the tutorial.
For now, suffice to say that argument
defines a positional argument,
many
allows parsers to be applied
zero or more times, and |
expresses alternatives.
Here is the help text:
p.parse_args(
"-h"
)
usage: [-x X | [Y ...]]
As promised, this succeeds:
p.parse_args("-x", "1")
{'x': 1}
And this succeeds:
p.parse_args("1", "2", "3")
{'y': [1.0, 2.0, 3.0]}
Thanks
Special thanks to "Functional Pearls" by Graham Hutton and Erik Meijer for bringing these topics to life.