requires¶

Install: pip install requires

Decorate that lets you Require/Import dependencies at runtime.

Python dependency management can be mind bottlingly complex. Optional dependencies are pretty common. Why not require the dependency at run time if a function requires said dependency?

This package has come in handy in lambda-land where you only get 250mb (on aws)!

---

Usage:¶

In [1]:

# This will fail
def uno():
    return json.dumps({"a": 1, "b": 2})


try:
    uno()
except NameError as ne:
    print("Error:", ne)

# This will fail def uno(): return json.dumps({"a": 1, "b": 2}) try: uno() except NameError as ne: print("Error:", ne)

Error: name 'json' is not defined

In [2]:

# This will not fail
import requires  # Module is callable! (checkout funkify for more info -- `pip install funkify`)


@requires("json")
def uno():
    return json.dumps({"a": 1, "b": 2})


uno()

# This will not fail import requires # Module is callable! (checkout funkify for more info -- `pip install funkify`) @requires("json") def uno(): return json.dumps({"a": 1, "b": 2}) uno()

Out[2]:

'{"a": 1, "b": 2}'

In [3]:

import requires


@requires("from json import dumps")
def uno():
    return dumps({"a": 1, "b": 2})


uno()

import requires @requires("from json import dumps") def uno(): return dumps({"a": 1, "b": 2}) uno()

Out[3]:

'{"a": 1, "b": 2}'

In [4]:

def dos():
    return dumps({"a": 1, "b": 2})


dos()

def dos(): return dumps({"a": 1, "b": 2}) dos()

Out[4]:

'{"a": 1, "b": 2}'

In [5]:

import requires


@requires(_from="json", _import="dumps")
def dos():
    return dumps({"a": 1, "b": 2})


dos()

import requires @requires(_from="json", _import="dumps") def dos(): return dumps({"a": 1, "b": 2}) dos()

Out[5]:

'{"a": 1, "b": 2}'

In [6]:

import requires


@requires(_import="rapidjson", pip="python-rapidjson", conda_forge="python-rapidjson")
def tres():
    return rapidjson.dumps({"a": 1, "b": 2})


tres()  # Will err if not install with where to install instructions

import requires @requires(_import="rapidjson", pip="python-rapidjson", conda_forge="python-rapidjson") def tres(): return rapidjson.dumps({"a": 1, "b": 2}) tres() # Will err if not install with where to install instructions

Out[6]:

'{"a":1,"b":2}'

In [7]:

# should error
def quatro():
    return path.join("a", "b")


try:
    quatro()
except NameError as ne:
    print("ERROR:", ne)

# should error def quatro(): return path.join("a", "b") try: quatro() except NameError as ne: print("ERROR:", ne)

ERROR: name 'path' is not defined

In [10]:

from requires import Requirement

os_path_req = Requirement(_import="path", _from="os")


@os_path_req
def quatro():
    return path.join("a", "b")


assert isinstance(quatro(), str)

from requires import Requirement os_path_req = Requirement(_import="path", _from="os") @os_path_req def quatro(): return path.join("a", "b") assert isinstance(quatro(), str)

Enforcing requirements¶

In [18]:

import requires

try:
    import alibrary
except ModuleNotFoundError:
    requirement = requires.Requirement(
        _import="alibrary",
        pip=True,
        conda_forge="alibrary-conda-listing",
        details="Install details",
    )
try:
    requirement.raise_error()
except requires.RequirementError as err:
    print("ERROR:")
    print(err)

import requires try: import alibrary except ModuleNotFoundError: requirement = requires.Requirement( _import="alibrary", pip=True, conda_forge="alibrary-conda-listing", details="Install details", ) try: requirement.raise_error() except requires.RequirementError as err: print("ERROR:") print(err)

ERROR:
Module/Package(s) not found/installed; could not import: `import alibrary`
    pip install alibrary
    conda install -c conda-forge alibrary-conda-listing
    Install details

Less verbose version:¶

import requires

try:
    import alibrary
except ModuleNotFoundError:
    requires.Requirement(
        _import='alibrary',
        pip=True,
        conda_forge='alibrary-conda-listing',
        details="Install details"
    ).raise_error()

---

Future ideas?¶

• Adding support for requiring particular package versions?
• Auto install?
• Allow non pip/conda/conda-forge locations?