Is there a callable equivalent to f-string syntax?

Question

Everybody loves Python 3.6's new f-strings:

In [33]: foo = {'blah': 'bang'}

In [34]: bar = 'blah'

In [35]: f'{foo[bar]}'
Out[35]: 'bang'

However, while functionally very similar, they don't have the exact same semantics as str.format():

In [36]: '{foo[bar]}'.format(**locals())
---------------------------------------------------------------------------
KeyError                                  Traceback (most recent call last)
<ipython-input-36-b7ef5aead76c> in <module>()
----> 1 '{foo[bar]}'.format(**locals())

KeyError: 'bar'

In particular, str.format() handles getitem syntax very differently:

In [39]: '{foo[blah]}'.format(**locals())
Out[39]: 'bang'

Given the ability to handle full-blown python expression syntax, f-strings are wonderful and I love them. But they have one hitch: they're evaluated immediately, whereas with str.format() I can save the string with its formatting as a template, and format it multiple times in different contexts.

So, is there an equivalent way to save a string as a template, and evaluate it, using f-string semantics, at a later date? Other than defining a function? Is there an equivalent to str.format() for f-strings?

Update:

So, hypothetical interface here as an example:

In [40]: mystr = '{foo[bar]}'

In [41]: make_mine_fstring(mystr, foo=foo, bar=bar)
Out[41]: 'bang'

Answer 1

Brief answer: NO.

You can read PEP-498 regarding these f-strings. It clearly defines their purpose, and the concept: these strings are evaluated in-place. The result is a usual str with formatted content. You cannot store f-strings as a template, as there is no special object for f-strings.

Your specific example is also mentioned in PEP-498's "Differences between f-string and str.format expressions" section.

So, whatever you do, you either use the inline in-place f-strings, or the old s.format() syntax, with different behaviour.

If you want to read an f-string from a file and evaluate it according to the syntax of f-strings, you could use eval:

foo = {'blah': 'bang', 'bar': 'sorry'}
bar = 'blah'

tpl = '{foo[bar]}'
print(tpl)

print(tpl.format(**locals()))  # sorry
print(eval(f'f{tpl!r}'))  # bang

Note how we use the f-string first, but convert the tpl into its own repr for immediate eval. Typically, for simple types, eval(repr(val)) should return val. But instead of just putting repr(tpl) (or {tpl!r}), we convert the repr of the regular string into the f-string, and evaluate it instead.