gist-import 1.0.3

GitHub Gist are handy snippets, which are meant to be copy-pasted into one's code... but what if you could import them?

gist-import

GitHub Gist are handy snippets, which are meant to be copy-pasted into one's code... but what if you could import them?

Installation

pip install gist-import

Usage

The main class is GistImporter. This allows one to import cleanly a gist into a project. Namely it retrieves the gist and executes it with any given additional named arguments in a local context thus not polluting the global namespace (see background). If required it does a special import of the gist's imports (<3.8) thanks to get_imports_in_codeblock. The variables in that namespace are then available as items.

from gist_import import GistImporter
gi = GistImporter('24d9a319d05773ae219dd678a3aa11be')
Safeguard = gi['Safeguard']

In the case of a codeblock that fails because of, say a terminal line, a warning saying so will be printed.

from gist_import import GistImporter
gi = GistImporter('👾👾👾', foo=1, bar=2)  # a gist with a syntax error on the last line will not run...
clean_code = gi.codeblock.split('\n')[:-1] # skip last line
gi = GistImporter.from_code_block(clean_code, foo=1, bar=2)
baz = gi['baz']

The function get_imports_in_codeblock is a helper function that finds any imports and return the modules as values of a dictionary.

There are four entry points:

• Gist id (regular initialisation)
• codeblock as discussed via from_code_block
• text file url via from_url
• GitHub page url via from_github

Background

GitHub Gist are snippets that aren't part of regular GitHub, they are intended to be used in blogs etc. to show code-hightlighting by embedding the gist. If one were to use in it Python, one should copy-paste it or do something convoluted.

Say the gist is nice and isolated, with all the correct imports, then this works fine.

import requests

response = requests.get('https://gist.github.com/matteoferla/d0daee35fe6f598bc720ce0eeebbac97/raw/6f7ba15dde86f1066629af61e0724dbe6a62cceb/transmute_FindMCS_parameters.py')
response.raise_for_status()
exec(response.text)
transmute_FindMCS_parameters()

..but things get messy quickly. As seen in the comment in this Gist:

https://gist.github.com/matteoferla/24d9a319d05773ae219dd678a3aa11be

As a placeholder for the response.text in the following examples a string is used.

The following works:

faux_gist:str = 'greet = lambda who: print(f"Hello {who}")'  # pretend this is the gist from `response.text`
exec(faux_gist)
greet('World')

But as soon as one moves away from the global namespace issues happen. This would be needed were one to want to wrap the gist execution in a function to avoid global namespace pollution.

This will fail:

def nonglobal(faux_gist:str):
    exec(faux_gist)
    return greet  # NameError: name 'greet' is not defined
    
# assign to a variable with a different name
salute = nonglobal('greet = lambda who: print(f"Hello {who}")')
salute('Mars')

But this will pollute the global namespace:

def nonglobal(faux_gist:str):
    exec(faux_gist, globals())
    return greet  # NameError: name 'greet' is not defined
    
# assign to a variable with a different name
salute = nonglobal('greet = lambda who: print(f"Hello {who}")')
salute('Mars')
assert 'greet' not in globals() # AssertionError: 

As globals() returns the actual global namespace, not a copy. If a copy is passed the copy will have the new variable and the namespace will not be polluted.

def nonglobal(faux_gist:str):
    faux_globals= {**globals(), **locals()}
    exec(faux_gist,  faux_globals)
    return faux_globals['beware']
  
# assign to a variable with a different name
enguard = nonglobal('import warnings; beware = lambda who: warnings.warn(f"Beware {who}")')
enguard('Mars')
assert 'beware' not in globals()

This used to not work in Python 3.7 due to the import being lost. Hence the function get_imports_in_codeblock which returns a dictionary of string to module of the imports in the codeblock —star imports included.