With DSwizzy145's thread closed, I thought I might as well post my findings.
First, the good news. The game's font and script are completely uncompressed - the former starts at file address 120000; the latter starts at file address 1C0000. Theoretically, there's nothing stopping a determined newcomer from hacking the daylights out of this game - the biggest hurdle to overcome is the well-established
volume of each. Which brings us to the bad news. The game's kanji font includes around
2,500 symbols, as I mentioned before, and the script runs to 20 banks of text - yes,
everything from 1C0000 onward in the ROM is
text. Out of the 3MB comprising the game ROM,
1.25MB are dedicated to the game's script. To call this a "huge project" would be a dramatic understatement.
Nonetheless, I am sincerely considering moving forward at least as far as making it possible to dump and insert the script. Yes, I'm talking about tackling that monster of a kanji table. Though I do have ideas. For one thing, it's extraordinarily
unlikely that the game uses all 2,500 kanji stored in its font -
the joyo list consists of "only" 2,136, and I find it doubtful that even a lengthy literary work like the one stored in the game's ROM would even
come close to using them all (consider, for comparison, a book that uses over half the words in your favorite dictionary). In other words, all we
really need to do is identify the kanji that the game
does use.
While my efforts on the GBA game
Wan Wan Meitantei (or "Sherlock Puppies," as I intend to call it) have stalled for some time, I did develop a tool for the express purpose of identifying all the kanji in the game's script. The method is simple: recreate the game's text engine, have it display a side-by-side comparison with the script data as rendered using the table, and insert any missing characters manually. Further refinements could simplify the process: I don't remember whether I set it up to flag lines that had missing characters, for one. While the workload will still be significant, having
context for characters should dramatically aid identification.
Of course, translating the script is still the same onerous task - it's an
enormous amount of text, and I suspect testing the end result will be an utter nightmare. Still, it's by a famous Japanese author, so it
must be worth the effort, right? ^_^;
EDIT: And I have just cut the labor in half. A careful examination reveals that the "second half" of the script (26000-end) is actually the first half of the script, with a couple of banks jumbled up (and broken in the process, so it's just padding instead of an alternate script). So double good news: the script is actually only
half the size I estimated (a bit under 640KB), and
the English script can be up to 2.25MB without compression (don't forget, it's a 3MB ROM - it can be expanded to 4MB if necessary, and the game uses bank pointers!)
EDIT #2: I haven't looked deeper into the matter yet but
holy crap the game's got the fixings for a VWF already baked in. I noticed something funny when my mock-up screens were matching up with the genuine article linebreak for linebreak (the game, and hence my mock-up engine, handles them automatically) even though I hadn't been trying to. I realized there was no way it was fitting 16 characters per line and still having margins on the left and right unless it was squishing them together; lo and behold,
it does. So maybe I can take advantage of this black magic.
EDIT #3: So I turn on the fast forward and autofire to get a feel for how long this game's supposed to be. Disadvantage: no way of choosing paths. Other disadvantage:
holy crap I did not need that late at night. Now I'm gonna have to go stare into Ike's eyes to detox. You're welcome.