Haven't hacked on Kdict since the T::T incident, but K's been at work
updating the content again, so this morning I sync'd the prod-dev
database to the public-facing copy on dict.collapsar.net. A good
example of how things are looking these days can be found here:
http://dict.collapsar.net/kdict/kdict.xi?wid=57
Last night at Too Damned Late o'clock I started writing down the
systematic (as opposed to programmatic) outline for parsing Mint
grammars and how to support all grammars with a single parser. Forever
and a day I've been thinking I needed to use Parse::RecDescent or
something else along those lines, but last night I had the thought
that since the grammar was pretty small, pretty simple, and very
rigid, the simplest thing might be to just tokenize incoming lines by
splitting on whitespace, and then writing my own (horrible) LR
parser. This would also do away with the various manifestations of the
newline problem, and would permit much more accurate error reporting
(including, in some cases, not only telling what line something went
wrong on, but pointing back at a probable cause or reference point on
another line).
This probably won't get done today though. The TODO is long, and I'm
running on 3.5h of sleep.
