fix short_version (upper/lower) case priority

[Houston4444]

Replaces #971.

short_version is used in narrow strips to limit the number of letters of strip names or plugin names. Reading its code, something jumped out at me and explain why many times there is nearly no link between the original name and the short name.

This function removes characters until the desired limit is reached.
Originally, It removes characters in this order :

• white spaces and punctuation
• lower-case vowels
• upper-case vowels
• lower-case consonants
• upper-case consonants

In this PR, it removes characters in this order:

• white spaces and punctuation
• lower-case vowels
• lower-case consonants
• upper-case vowels
• upper-case consonants

Upper case has priority over lower case ! It explain why
ACE Reasonable Synth gave in 5 characters CRsnS (no link with original, instead of ACRsS, not perfect but better). It allows to use CamelCase and have at least the first letter of each word.

Note that "Y" is considered as a consonant, "S" is never removed, I did not change that.

[x42]

I think we should remove that ancient custom short_version function and instead rely on Pango::EllipsizeMode
which also works for other languages.

[pauldavisthefirst]

My understanding is that Pango ellipsizing is a completely different concept. This one was called "disemvoweling" and comes up with entirely different results. On long text, disemvoweling still makes most languages legible, which ellipszing does not. However, it is possible that on short text (e.g. plugin names), the results are not the same.

[Houston4444]

@x42, If we talk about this enum https://docs.gtk.org/Pango/enum.EllipsizeMode.html, I think we are not talking about the same thing. EllipsizeMode only removes the start, the middle or the end of the string. The short_version mechanism is pretty good to recognize an item with fewer characters.
@pauldavisthefirst, Note that, for a good internationalization, it would be nicer to include vowels with accent but there are many (at least 16 in french).

For me, the biggest problem remain the fact narrow strips limits the strip name to 5 characters (https://tracker.ardour.org/view.php?id=6610). I try since few days a build with this limit set to 10, and I have never seen any truncated name. It probably depends on the font though, but I think this limit could be set to 8 without any risk.

[pauldavisthefirst]

The vowel and consonant lists are translatable and can be any size.

[Houston4444]

The vowel and consonant lists are translatable and can be any size.

Too many characters, too much work to try to cover many latin languages. example: https://altcodeunicode.com/alt-codes-all-latin-vowel-consonant-letters-with-accents-diacritical-marks/
I could make a global priority to uppercase if std::tolower (and toupper) was giving the lowercase of a letter with accent (exemple: if 'É' becomes 'é' with tolower() as in Python), but it is not the case, and if ever we could, it could be very system specific, so, I won't work on this.

[pauldavisthefirst]

The idea would be, e.g. that the French translation includes tfrnaslations for the vowel list, and the Hungarian translation also, with its own list. There's no need or reason for a unviersal list.

[Houston4444]

something like this ?

string lang = std::getenv("LANG");
string sht_lang = lang.substr(0, 2);
string locale_vowels = "";

if (sht_lang == "fr"){
	locale_vowels = "àéèïôùy";
} 
/* others "else if" for other languages */

/* later */
if (locale_vowels.size()){
	while (orig.length() > target_length) {
		if ((pos = orig.find_last_of (_(locale_vowels.c_str()))) == string::npos) {
			break;
		}
		orig.replace (pos, 1, "");
	}
}

[pauldavisthefirst]

we localize strings in ardour with gettext, which means the _() macro

hence _("aeiou") is localizable. no other verbiage necessary.