Now it's telling me that

K = k g ng

is an "Illegal Variable Definition".

Can you attach the actual rules file? I'm totally unable to reproduce this problem, but it's failing during the parse. That error means that when the string K = k g ng is split on \s*=\s*, the result has more or less than 2 parts. Normally, there should be exactly 2, K and k g ng. This error message is slightly better than the one that was originally produced, which is just an IndexOutOfBoundsException.

Here's the full ruleset:

K = k g ng
T = m p b f v n t d s z c j r l ny y
N = ḿ ń ŋ ñ ññ
C = T K
 
V = i ì e è a aì u aa o ou

l > 0 / V_ % Postvocalic l is lost.
om on ong ony onny > ḿ ń ŋ ñ ññ / _ % Relabeling the syllabic nasals.
i ì e è a aì > u u o o aa ou / K?_(K{T N #})? % Front vowels back after syllable-initial velars and before syllable-final velars.
u aa > ì a / {c j Cy}?_({c j Cy}{T N #})? % Back vowels front after syllable-initial 'palatals' and before syllable-final 'palatals'.

No, you need to attach the rules file itself. I think there is an encoding problem, because I cannot reproduce this error when you just copy-paste the content.

My money's on "oh no bad linebreak!"

It might be a bad UTF-8 BOM but line breaks shouldn't ever be an issue. The command is getting trimmed and then split on \s+=\s+. Now, that probably should use \s*, but that's not an issue for what we are doing here.

The dreaded BOM

Ok, here's the file.

Ok, congratulations, I was able to reproduce the error. Your file is encoded as Big Endian UCS-2. I don't really want to require people to use UTF-8, but these fixed-size encodings can be trouble.

IT'S A BOM

But seriously, it must be:

[user@milosz Downloads]$ cat modern\ southlandic\ rules.txt
��K = k g ng
T = m p b f v n t d s z c j r l ny y
[…]

After reopening it in Kate1, and unchecking the BOM box, and saving, it look ok:

[user@milosz Downloads]$ cat modern\ southlandic\ rules.txt
K = k g ng
T = m p b f v n t d s z c j r l ny y
[…]

I updated the edited file here, so you can check if the error still occurs.

 * * *

Edit: Sorry, I didn't realize Morrigan replied. I am probably wrong.

I don't think requiring UTF-8 is too much to ask tbh. It's not like it's hard to select

I'm using regular old Notepad. Resaving the file as UTF-8 fixed the problem.

You should look into Notepad++. It loses none of the functionality of Notepad1 while gaining a whole lot more, without losing the ability to just use it simply. And it's a hell of a lot faster and more stable at opening huge files than Notepad.

Does Haedus support metathesis yet?

That depends on what you need it to do. It lets you do some things trivially.

Also, Notepad++ is totally boss.

So, the parent language has a lot of final syllabic nasals, and I want these to undergo metathesis with the preceding consonant(s).

as in CN > NC ? Because that rule right there should cover a lot of cases. Only problem is that it might be slow.

Oh, so it'll actually remember the exact values of the variables? Cool.

Apparently it doesn't remember the exact values, because "fraadonny" (frädň before metathesis) becomes "fräḿy" when it should be "fräňd".

I'll check that out. It should keep them lined up, but I haven't really tested this. It might be worth making a special case for metathesis. Basically, it should generate all combinations (which is why it's slow), but that should keep both sides lined up; I guess not quite.