Fix + test for string format with %% and keys

[dmoisset]

This fixes #1717 and adds the corresponding test

[ddfisher]

Thanks! This looks pretty good, but I think it would be a little cleaner to instead not create ConversionSpecifiers if they'd have type % (see parse_conversion_specifiers). Would you mind making that change?

[dmoisset]

I made the change and I think its slightly cleaner. I expected it to be better and also allow me to remove the '%' check from replacement_checkers(), but the parser still needs to produce the ConversionSpecifier for the "%*%" check

[ddfisher]

Oh yikes -- I wasn't aware of "%*%" and similar. In the presence of those, I actually like your initial commit more: "%%" really is a full-on conversion specifier (and not merely an escape sequence as I'd initially thought), but it just doesn't need a value.

Sorry for the flip-flopping!

[fabianhjr]

Also took a swing at it with PR: #1905

What I did was store each result in result and if it was the %% literal just skip it.

[dmoisset]

@ddfisher I created a PR with the old version of the fix, #1907 , feel free to apply to the one you prefer (I still think this one is the smallest and cleaner, but at this point I think it's just time to choose one of them (or the one from @fabianhjr ).

[fabianhjr]

Hello @dmoisset, I think you got an off by one error; The PR is #1908

[gvanrossum]

@ddfisher I'll let you choose the PR to merge!

[ddfisher]

Closing in favor of #1908. Thanks!