On 7/20/2015 5:29 PM, Bruce Leban wrote:
On Mon, Jul 20, 2015 at 1:20 PM, Eric V. Smith <eric@trueblade.com <mailto:eric@trueblade.com>> wrote:
So: f'api:{sys.api_version} {a} size{sys.maxsize}'
would become either: f'api:{.api_version} {} size{.maxsize}'.format(sys, a, sys) or f'api:{0.api_version} {1} size{0.maxsize}'.format(sys, a)
Or: f'api:{} {} size{}'.format(sys.api_version, a, sys.maxsize)
Note that format strings don't allow variables in subscripts, so
f'{a[n]}' ==> '{}'.format(a['n'])
Right. But why re-implement that, instead of making it: '{[n]}'.format(a)? I've convinced myself (and maybe no one else) that since you want this: a=[1,2] b={'c':42} f'{a[0]} {b[c]}' being the same as: '{} {}'.format(a[0], b['c']) that it would be easier to make it: '{[0]} {[c]}'.format(a, b) instead of trying to figure out that the numeric-looking '0' gets converted to an integer, and the non-numeric-looking 'c' gets left as a string. That logic already exists in str.format(), so let's just leverage it from there. It also means that you automatically will support the subset of expressions that str.format() already supports, with all of its limitations and quirks. But I now think that's a feature, since str.format() doesn't really support the same expressions as normal Python does (due to the [0] vs. ['c'] issue). And it's way easier to explain if f-strings support the identical syntax as str.format(). The only restriction is that all parameters must be named, and not numbered or auto-numbered. Eric.