PyWart: "Python's import statement and the history of external dependencies"

[Rick Johnson]

On Friday, November 21, 2014 4:25:49 PM UTC-6, Rick Johnson wrote:
> ############################################################
> #                          STEP 3                          #
> ############################################################
> # Make the following changes to the import machinery:      #
> # Before Python reads a module file, Python will clear the #
> # values in "sys.path_extra", OR, query the                #
> # "__search_paths__" variable, if any paths exists in this #
> # list, THEN THESE PATHS MUST BE SEARCHED, AND THEY MUST   #
> # BE SEARCHED BEFORE ANY PATHS IN "sys.path", AND NO       #
> # PEEKING IN "sys.modules" IS ALLOWED!                     #
> ############################################################


Actually, to be more specific, not only should the
__search_path__ variable be unique to each file, Python
should query the value of "__search_path__" variable each
time it encounters an import statement.

So, if we wanted to import multiple modules who share the
same name, but who exist in different namespaces, we would
could do so by changing the values of __search_path__ before
calling the import:

    #
    # import the first coconut module Python can find using
    # the default search path, if not raise an error
    #
    import coconut

    #
    # import the first coconut module python can find using
    # *ONLY* the explicit paths provided, or raise an error.
    #
    __search_path__ = [
        'path/to/a/specific/coconut/tree',
        'might/be/here/also'
        ]
    import coconut

At first i was thinking that *IF* any "explicit paths" were
defined via __search_path__, then Python should look in all
those paths first, then if nothing is found look in
sys.modules, then if nothing is found, exhaust sys.modules,

BUT NOW I AM THINKING THAT IS A BAD IDEA!

When __search_path__ *IS* defined, Python should look *ONLY*
in the provided paths, and if not module is found, raise an
"ExplictImportError".

Because if we're going to be explicit then we need to also be
consistent. If the programmer is requesting that Python
search in a predefined directories, then obviously he
doing so to avoid name clashes, so if Python went on and
returned a module OUTSIDE of those predefined paths, Python
would be circumventing (IMPLICITLY AGAIN!) the wishes of the
programmer.

============================================================
 Succinct Summary Of My New Proposal:
============================================================

1. Use the historical "implicit import" mechanism for most day
to day imports, and let Python do all the heavy lifting.

2. Use the new "explicit import" mechanism for advanced name
resolutions, but realize that since you are now taking
control of import, so you must do more work, and you must be
"exhaustively explicit" about *where* Python searches.