Issue 28308: Accelerate 'string'.format(value, ...) by using formatted string literals (original) (raw)

For now using formatted string literals (PEP498) is the fastest way of formatting strings.

$ ./python -m perf timeit -s 'k = "foo"; v = "bar"' -- '"{!s} = {!r}".format(k, v)' Median +- std dev: 3.96 us +- 0.17 us

$ ./python -m perf timeit -s 'k = "foo"; v = "bar"' -- 'f"{k!s} = {v!r}"' Median +- std dev: 1.09 us +- 0.08 us

The compiler could translate new-style formatting with literal format string to the equivalent formatted string literal. The code '{!s} = {!r}'.format(k, v) could be translated to

t1 = k; t2 = v; f'{t1!r} = {t2!s}'; del t1, t2

or even simpler if k and v are initialized local variables.

$ ./python -m perf timeit -s 'k = "foo"; v = "bar"' -- 't1 = k; t2 = v; f"{t1!s} = {t2!r}"; del t1, t2' Median +- std dev: 1.22 us +- 0.05 us

This is not easy issue and needs first implementing the AST optimizer.

One thing to be careful of here is that there's one slight difference between how str.format() and f-strings handle indexing of values. f-strings, of course, use normal Python semantics, but str.format() treats indexing by things that don't look like integers as string literals, not variables. It's an unfortunate left-over from the original PEP-3101 specification:

d = {'a':'string', 0:'integer'} a = 0 f'{d[0]}' 'integer' '{d[0]}'.format(d=d) 'integer' f'{d[a]}' 'integer' '{d[a]}'.format(d=d) 'string'

Note that the exact same expression {d[a]} is evaluated differently by the two ways to format.

There's a test for this in test_fstring.py.

Someday, I'd like to deprecate this syntax in str.format(). I don't think it could ever be added back in, because it requires either additional named parameters which aren't used as formatting parameters, or it requires global/local lookups (which isn't going to happen).

i.e., this: '{d[a]}'.format(d=d, a=a)