Re: [Tuxpaint-devel] Font scoring method

[Albert C. <aca...@gm...>]

> I noticed a lot of translators, back when these strings existed
> (but were not utilized) either translated them literally, or put
> the string, as-is, in the "msgstr" field of their POs.

This will work correctly. It is the intended usage, more or less.

> I've attempted to remove all of those (replaced with `msgstr ""`).

I guess this works too, but it sure wasn't intended usage.

> ca.po-msgstr "aa"
> ca.po-msgstr "aa"
> ca.po-msgstr "eèéëcç"
> ca.po-msgstr "EÉÈËCÇ"

needs «»
needs Euro symbol

> ca@valencia.po-msgstr "aa"
> ca@valencia.po-msgstr "aa"
> ca@valencia.po-msgstr "eèéëcÇ"
> ca@valencia.po-msgstr "EÉÈËCÇ"

needs «»
needs Euro symbol

> de.po-msgstr "aa"
> de.po-msgstr "aa"
> de.po-msgstr "äüöß"
> de.po-msgstr "ÄÜÖ"

needs both cases for ßẞ
needs «»
needs Euro symbol
needs center dot (for multiplication)

> hu.po-msgstr "áíűőüöúóé"
> hu.po-msgstr "ÁÍŰŐÜÖÓÉ"

maybe needs Euro symbol
maybe needs low and high quote marks

> is.po-msgstr "aa"
> is.po-msgstr "AA"
> is.po-msgstr "ðéíóúþæö"
> is.po-msgstr "ÐÉÍÓÚÞÆÖ"

maybe needs Euro symbol
maybe needs low and high quote marks

> it.po-msgstr "aa"
> it.po-msgstr "aa"
> it.po-msgstr "èòàì"
> it.po-msgstr "ÈÒÀÌ"

needs «»
needs Euro symbol

> ja.po-msgstr "漢字"
> ja.po:msgstr "<1>spare-1b"
> ja.po-msgstr "あア"
> ja.po:msgstr "<9>spare-9b"

I don't see the Yen, which should be compared with the Yuan to
ensure that they are distinct. (some fonts mix them up, or used to)

I don't see all three Japanese glyph sets: katakana, hiragana, kanji

Is anything else required to downrate fonts that would render the
kanji in the Traditional Chinese, Simplified Chinese, or Korean style?
(FYI, many of these characters were unified by the Unicode Consortium,
but the graphical appearance has distinctions that people care about)

> I'd assume it makes more sense to leave them blank / untranslated
> (`msgstr ""`), rather than do that.  But I didn't want to stomp on
> your work.  Thoughts?  (Albert too, of course!)

There had been the thought originally that the translators could
supply the weight numbers. Maybe the <9> would be parsed, so
it could instead be <23> or <-4> or whatever. Perhaps that is
needlessly complicated. It may be enough to have just the 4 choices.
Increasing that to 9 choices, without getting fancy, is not so hard.
(would likely be 1a, 1b, 1c, 5a, 5b, 5c, 9a, 9b, 9c)

Leaving them untranslated: the repeated characters match,
causing failure, so a harmless 0 is added.

Translating them to an empty string: I think this too might be
a 0, but in any case it is a constant value and thus harmless.
Performance could be a tiny bit better with empty strings.

> Should we drop a note onto the tuxpaint-i18n@ mailing list to ask
> translators to look at this again?

Yes, but maybe wait until the above is all cleared up.