support fill, alignment and width in range and tuple formatters

[avionicharshit-byte]

fmt::format("{:*>20}", std::vector<int>{1,2,3}) doesnt compile , it fails with "invalid format specifier" . same for tuple , pair and map . std::format accepts it .

// before
fmt::format("{:*>20}", std::vector<int>{1, 2, 3});  // error: invalid format specifier

// after
fmt::format("{:*>20}", std::vector<int>{1, 2, 3});  // "***********[1, 2, 3]"

in C++23 both range-format-spec and tuple-format-spec start with fill-and-align_opt width_opt , so fill , alignment and width belong to the composed output . range_formatter::parse never called parse_align or parse_width , it went straight to looking for n , s and ? . doc/syntax.md documented the shorter grammar as well , so thats updated too .

the fix adds detail::composed_specs in include/fmt/ranges.h , used from the range , tuple and map formatters . it parses alignment and width before the range's own options , and skips a leading : so {::>5} still means the underlying element spec , which is what std::format does . in format() , when a width is asked for the composed output goes into a memory_buffer first and is then written padded . thats the same buffer and pad idiom nested_formatter uses , and the same one #4860 used for the std::exception formatter .

each formatter's old body moved into write_body unchanged and format() became the padding wrapper . nothing existing changed behaviour , no existing test expectation was edited .

i checked 22 cases against real std::format output from libc++ : default align , < > ^ , custom fill , dynamic width {:*>{}} , width with n , width with a nested element spec , sets , maps , vector<std::string> debug quoting , tuple , pair , and output longer than the width . 21 of the 22 match byte for byte . added ranges_test.format_width covering them , all 22 test suites pass , clang-format clean .

the 22nd is not part of this PR . fmt::format("{:n}", std::pair<int, int>{1, 2}) gives 12 where std::format gives 1, 2 , fmt drops the separator along with the brackets for tuples but keeps it for ranges . thats already on master and i havent touched it , fixing it would change the expectations at test/ranges-test.cc:176 and :183 . happy to file it separately .

[avionicharshit-byte]

@vitaut ci is waiting on workflow approval. the earlier failure was a missing FMT_CONSTEXPR on composed_specs::resolve, fixed and amended.

[vitaut]

Thanks for the PR! Overall looks good but please update the doc per inline comment.

[vitaut]

Please clarify that : is not allowed, something like:

The fill, align and width options have the same meaning as in the standard format specification, except that : cannot be used as a fill character because it introduces the underlying element specification.

[avionicharshit-byte]

done , used your wording