Codecov Report

Merging #590 into master will increase coverage by 2.14%.
The diff coverage is 46.66%.

@@            Coverage Diff             @@
##           master     #590      +/-   ##
==========================================
+ Coverage   41.61%   43.75%   +2.14%     
==========================================
  Files          74       74              
  Lines       16762    17228     +466     
  Branches     4406     4671     +265     
==========================================
+ Hits         6975     7538     +563     
+ Misses       7781     7594     -187     
- Partials     2006     2096      +90

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 5d28f9b...c4953ee. Read the comment docs.

Looks like CPython doesn't do "automatic" dict type checking there, and just allows any type, then checks for None, then manually typechecks for dict.
https://github.com/python/cpython/blob/master/Python/bltinmodule.c#L1000

In that case I've added a macro to do the extra checking.

This leads me to run into the limitations of our definitions of scope. We consider local scope to be the nearest scope to us so top level statements execute with local and global scope being the same thing. In python though this means there is no local scope, creating one doesn't replace the current local==global scope.

This should work, but doesn't:

x=2
exec("print(2)", None, {})

We'll need to sort this out to implement the globals keyword, but I think what we have is okay for now.

Uh... the above is wrong. It looks like creating a new local scope never replaces the existing local scope. (Ie. that code still works even in a function.)

This is all wrong. :(

I was right the first time! :)

Interesting example:

def g():
    seven = "seven"
    def f():
        d = {}
        exec("y = seven", None, d)
        assert d['y'] == seven
    f()
g()