Strf Versions Save

Bug fix: Flush buffer on the destructor of basic_string_appender()

• Breaking changes

  • Changed how printable types are to be added

  • Redesigned transcoding API

  • removed surrogate_policy facet

  • width_decumulator is now default constructed with remaining_width() == width_max

  • remove copy-constructor from array_destination

  • printing of integers is not overridable anymore

  • added template parameter to printable_overrider_c

  • Renamed/replaced

    • In PrintingTraits: formatters -> format_specifiers

      • suffix _formatter to _format_specifiers in all names

    • Named type requirements Formatters to FormatSpecifier

    • conv -> transcode

    • value_with_formatters -> value_and_format

    • codepoints_robust_count -> count_codepoints

    • codepoints_robust_count_f -> count_codepoints_f

    • codepoints_fast_count -> count_codepoints_fast

    • codepoints_fast_count_f -> codepoints_fast_count_f

    • codepoints_count_result -> count_codepoints_result<CharT>

    • width_and_pos -> width_and_ptr<CharT>

    • min_space_after_recycle to min_destination_buffer_size

    • basic_cstr_writer to basic_cstr_destination

    • basic_char_array_writer to array_destination

    • discarded_destination to discarder

    • type requirement PrintTraits -> PrintableDef

      • and its member type formatters -> format_specifiers
      • and its member function template make_printer_input -> PrintTraits::make_input

    • removed type requirement PrinterInput

    • struct template print_traits -> printer, and

      • print_traits_of -> printer_of
      • print_traits_tag -> printable_tag

    • function template preview to precalculate

    • class templates

      • print_preview to premeasurements
      • size_preview to size_accumulator
      • width_preview to width_accumulator ( and redesigned )

    • type alias print_size_and_width_preview to full_premeasurements

    • enumerations

      • preview_size to precalc_size
      • preview_width to precalc_width

    • Replace class templates destination_no_reserve, destination_calc_size and destination_with_given_size by printing_syntax

    • Change Transcode requirements

      • change parameters order in transcode
      • add unsafe_transcode and unsafe_transcode_size
      • rename member type aliases:
        • src_char_type -> src_code_unit
        • dst_char_type -> dst_code_unit

  • Redesigned destination:

    • destination<CharT> now is a type alias to output_buffer<CharT, 64>, where output_buffer<CharT, N> derives from output_buffer<CharT, N - 1> and the poscondition of output_buffer<CharT, N>::recycle_buffer() is that space() >= (1 >> N)
    • Renamed tepe membes char_type to value_type
    • Removed deprecated function size

  • Removed deprecated type aliases: basic_outbuff, bin_outbuff, u8outbuff, u16outbuff, u32outbuff, woutbuff

  • Removed deprecated class template width_by_func ( and function template make_width_calculator )

  • Replaced facet invalid_seq_notifer by transcoding_error_notifer

  • When transcoding U+FFFD, just translate it to the replacement character of the destination encoding ( don't call trascoding_error_notifier::unsupported_codepoint anymore )

  • Replacing unsigned integers with signed everywhere. Including:

    • signature of virtual function output_buffer::do_write
    • signature of charset functions

Fixed bugs:

• #34 - Wrong result when rounding up 9.999...
• #35 - Wrong rounding in strf::fixed(0.051).p(1)

• Rename class template basic_outbuff to destination And its member functions:
  pointer() -> buffer_ptr() end() -> buffer_end() space() -> buffer_space() But create aliases with the old names to preserve backward compatibility

• Fix bugs in floating-point printing:

  • segfault when printing in fixed notation with precision a value that is to close to zero ( like strf::fixed(1e-30, 3))
  • wrong aligment when printing strf::gen(120.1, 3) > 10

• Fix some mistakes in CMakeLists.txt:

  • target strf-cuda was equivalent to strf-header-only
  • make exported package relocatable

• Now, by default, the width of the input strings is calculated as specified to std::format
• Now the destructors of narrow_cfile_writer, wide_cfile_write and basic_streambuf_writer call recycle ( or do something equivalent to that )

Breaking changes:

• Now you need to invoke punct format function if you want to apply numeric punctuation
• Made width_t unsigned
• Modified width_t::round() : now it rounds towards zero on ties ( instead of towards even number ).
• Removed class template basic_outbuff_noexcept.
• Added template parameter to narrow_cfile_writer to specify buffer's size.
• Renamed strf::get_facet to strf::use_facet
• Renamed outbuff_garbage_buf to garbage_buff
• Renamed "Char econding" to "charset":
  • char_encoding to charset and
  • char_encoding_id to charset_eid
  • eid_* to scid_*
  • static_char_encoding to static_charset
  • dynamic_char_encoding_data to dynamic_charset_data
  • dynamic_char_encoding to dynamic_charset
  • char_encoding_c to charset_c
  • convert_encoding to convert_charset`
  • convert_from_encoding to convert_from_charset`
  • sanitize_encoding to sanitize_charset
  • sanitize_from_encoding to sanitize_from_charset
  • get_encoding to get_charset
  • CharEncoding to Charset ( type requirement )
  • Charset::char_type to Charset::code_unit
  • Charset::decode_char to Charset::decode_unit
• Renamed charset and width calculation facets: Added the "_t" suffix, and created a constexpr variable, or a variable template, with each of the previous names. For example:
  • Charsets: The types utf<CharT>, utf8<CharT>, strf::iso_8859_6<CharT>, etc were renamed to utf_t<CharT>, utf8_t<CharT>, strf::iso_8859_6_t<CharT>, etc. The old names ( utf<CharT>, utf8<CharT>, etc ) are now constexpr values.
  • Width calculators: The types fast_width, width_as_fast_u32len, and width_as_u32len were renamed to fast_width_t, width_as_fast_u32len_t and width_as_u32len_t respectively. The old names are now constexpr values.
  • Change the default facet of width_calculator_c to width_as_u32len_t.
  • Modified In PrintTraits::make_printer_input signature: added tag<CharT> parameter.

Improvements:

• Add support to all ISO-8859-x and Windows-125x character encodings.
• Create char_array_writer and to_range ( that writes to char* without appending '\0' )

Bug fixes:

• Fix 32bit floating point in hexadecimal ( it was totally wrong )
• Transcoding from UTF-16 to UTF-8 ac49b9e5c5a26114a8dd3af1f578a7c7e8bb6e36
• UTF-8 decoding big codepoints fac0db3d14457e2641cec929b6941309c0ddcd67

Others things:

• Create write and and do_writemember functions in basic_outbuff
• Moved benchmarks to another git repository
• Removed git submodules

• Breaking changes:

  • Removed split (a.k.a. internal ) alignment formatting. The following items have been removed:
    • text_alignment::split
    • format function operator%
    • join_split
  • Renamed:
    • alignment_formatter -> dynamic_alignment_formatter
    • empty_alignment_formatter -> alignment_formatter
  • Removed template parameter from float_formatter. Now there is instead a float_notation member variable in float_format.

• New features

  • Format function pad0 ( similar to printf format flat '0' )
  • Format function fill_sign ( can also use operator~). ( similar to printf format flat ' ' )

• Little improvements:

  • PrintTraits is more flexible now: If PrintTraits::make_printer_input<CharT>(p,facets,x) is ill-formed, then PrintTraits::make_printer_input<CharT>(p,facets,fmt(x)) is used instead. This reduces boilter-plate code when you extend the library ( add a new printable type )

Bugfix: Header <intrin.h> was being included inside namespace strf::detail::ryu

• New features

  • New printable type: char32_t
  • Created strf::preview function template.

• Breaking changes:

  • In basic_outbuff, renamed size to space
  • In class template size_preview, renamed get_size to accumulated_size.
  • In class template width_preview, changed behaviour of subtract_width and removed checked_subtract_width.
  • Modified PrintTraits type requirements.
  • In usual_printer_input, rename fp to facets.
  • Modified float_format_fn, int_format_fn, alignemnt_format_fn
  • Redesigned class digits_group_iterator
  • Bug #19 - Now the argument passed to fill must be of a character type
  • Bug #28 - Now *oct(0) is printed as "0", not "00"
  • Now format function operator* and operator+ when applied to integers when the numeric base is not correct.
  • Removed deprecated format function operator~ ( replaced by operator*)

• Enhance documenation:

  • Created "How to add destinations"
  • Created "How to add printable types"
  • Created "How to override printable types"
  • Document numpunct facet in strf_hpp.html