Script runs and shaping

A shaper such as HarfBuzz shapes one script at a time. script.h splits text into runs of one script, and the optional maul-unicode-harfbuzz library gives HarfBuzz its Unicode data from Maul Unicode.

Script runs

A script iterator resolves Common and Inherited characters (spaces, punctuation, digits, combining marks) into the run around them, narrows a run by Script_Extensions where a character is shared by a few scripts, and keeps a closing bracket in the script of its opening bracket:

muniScriptIterator scripts;
if (muniInitScriptIterator(&scripts, text, length, false) == muni_success)
{
    muniScriptRun run;
    while (muniScriptIterator_NextRun(&scripts, &run) == muni_success)
    {
        // the text up to run.end has script run.script
    }
}

Like the segment iterators, it takes text in pieces through muniScriptIterator_Feed after muni_needMoreText, and its calls refuse an iterator no init call set up with muni_errorInvalid. muniFindScriptRuns writes the runs of a whole text into an array.

A text layout pipeline itemizes by paragraph (bidi), then by script and bidi run, then by font, and shapes each item.

HarfBuzz

Built with -DMAUL_UNICODE_HARFBUZZ=ON, the static library maul-unicode-harfbuzz provides muniCreateHarfBuzzFunctions, which answers HarfBuzz's questions about general category, combining class, mirroring, script, composition and decomposition from Maul Unicode. Shaping, segmentation and bidi then agree on one Unicode version.

hb_unicode_funcs_t* functions = muniCreateHarfBuzzFunctions();
hb_buffer_set_unicode_funcs(buffer, functions);

The functions are immutable and may be shared between threads; release them with hb_unicode_funcs_destroy. Installed, the library is a package of its own: find_package(maul-unicode-harfbuzz) with the target maul-unicode::maul-unicode-harfbuzz, or the pkg-config module maul-unicode-harfbuzz, both of which bring in HarfBuzz.

Not obvious

• Leading Common characters take the script of the first run. A text of Common and Inherited characters only is one run of MUNI_SCRIPT_COMMON.
• A run whose characters are all shared by several scripts could be any of them. It takes the first Script of its own, among its characters, that the run could still be when that character came, while the run can still be that script, and otherwise the script whose tag comes first alphabetically. U+30FC KATAKANA-HIRAGANA PROLONGED SOUND MARK alone, shared by Hiragana and Katakana, is a run of Hiragana.
• The bracket stack holds 32 opening brackets; when it is full the oldest is forgotten, and its closing bracket then joins the run around it like any other neutral character.
• Script tags have the same values as HarfBuzz's hb_script_t, so a run's script passes to hb_buffer_set_script unchanged.