hyphen breaking behaviour

l10n
#1

How we've managed to fix the smart quote issue for our locale (many thanks to Mikhel and Eike!) I've got another headache I would like to tackle. I've asked Eike but he's not entirely sure and suggested I ask the list.

The problem is this: Gaelic restricts the breaking of hyphens across lines if there are less than two letters before the hyphen. In effect, this affects a finite number of prefixes, mostly:
n-WORD (e.g. ar n-aran)
t-WORD (e.g. an t-aran)
h-WORD (e.g. na h-aran)
an-WORD (e.g. seall an-seo)
eu-WORD
mì-WORD
dì-WORD
(and a few more, I can, if necessary supply a full list).

At the moment, at the end of a line a word like an-seo just gets broken into an-
seo. The only known fix is that once you're done writing, you do a global replace of <an-> to <tU+2011> (i.e. a NON-BREAKING HYPHEN) etc.

Is there something one could exploit to force specific instance of HYPHEN to the NON-BREAKING HYPHEN? The main problem I can see would be that if one cannot distinguish between single t and t word finally, something like cat-ghrìosaich might also end up with a non-breaking hyphen.

Eike suggested the following:

Maybe some rule for AutoCorrection could match this, but I'm not very
familiar with. However, the en-GB lists are under
extras/source/autocorr/lang/en-GB/ in the source tree, likely such rules
similar to the ones in
extras/source/autocorr/lang/en-GB/DocumentList.xml would go into
extras/source/autocorr/lang/gd-GB/DocumentList.xml and using something
like

   <block-list:block block-list:abbreviated-name="n-.*" block-list:name="‑"/>

where the second "‑" is a U+2011 NON-BREAKING HYPHEN.

The ".*" seems to indicate "any sequence of non-word-boundary characters
including none", so also "n-" without a following WORD would be matched.
Apparently there is no support for ".+" or other regular expressions.

Michael