mecab_controller outputs garbled characters for some kanji.
For the expression "粗末な家に住んでいる" in the examples, I got the following, debugger at basic_mecab_controller.py:110 :

(Pdb) outs
b'\xe7\xb2<ajt__component_separator>\xe7\xb2<ajt__component_separator>\xa5\xd2\xa5\xab\xa5\xac\xa5\xdf<ajt__component_separator>\xcc\xbe\xbb\xec<ajt__component_separator><ajt__node_separator>\x97<ajt__node_separator>\xe6\x9c\xab<ajt__node_separator>\xe3\x81\xaa<ajt__node_separator>\xe5\xae\xb6<ajt__node_separator>\xe3\x81\xab<ajt__node_separator>\xe4\xbd\x8f<ajt__node_separator>\xe3\x82\x93\xe3\x81\xa7\xe3\x81\x84\xe3\x82\x8b<ajt__node_separator><ajt__footer>'
(Pdb) outs.rstrip(b'\r\n').decode('utf-8', 'replace')
'�<ajt__component_separator>�<ajt__component_separator>�ҥ�����<ajt__component_separator>̾��<ajt__component_separator><ajt__node_separator>�<ajt__node_separator>末<ajt__node_separator>な<ajt__node_separator>家<ajt__node_separator>に<ajt__node_separator>住<ajt__node_separator>んでいる<ajt__node_separator><ajt__footer>'

I grabbed the command that would be run from self._mecab_cmd and ran in my terminal:

echo '粗末な家に住んでいる' | /opt/homebrew/bin/mecab --dicdir=/opt/homebrew/lib/mecab/dic/ipadic --rcfile=/Users/MY_PATH_TO_MECAB_CONTROLLER/mecab_controller/support/mecabrc --userdic=/Users/MY_PATH_TO_MECAB_CONTROLLER/mecab_controller/support/user_dic.dic --input-buffer-size=819200
        ̾,,*,*,*,*,,ҥ,ҥ
        ,,*,*,*,*,*
末      ̾,̾,ȿ,*,*,*,*
な      ̾,,*,*,*,*,*
家      ̾,,*,*,*,*,*
に      ̾,,*,*,*,*,*
住      ̾,,*,*,*,*,*
んでいる        ̾,,*,*,*,*,*
EOS

Platform: Apple Silicon Mac
OS: macos 14.1
Installed mecab and mecab-ipadic with Homebrew.

I'm not too familiar with mecab, is this how it is supposed to output if it can't parse a token?
If so, the section gets split into components at mecab_controller.py:57, the value ends up ['�', '�', '�ҥ�����', '̾��', '']. Should this be handled in this error block?

except ValueError:
    # unknown to mecab, gave the same word back
    word, headword, katakana_reading = components * 3
    part_of_speech, inflection = None, None

Hello, I came here after using a great Anki plugin of yours called PitchAccent.
I've noticed the issue when trying to convert pitch pattern to hiragana that it doesn't handle long vowel mark ー properly.
Turns out that it isn't that easy to convert katakana to hiragana because of the fact that there are two ways to make vowel longer. If we would simply try to reverse "ー" character based on the preceding vowel it would make words like せんせえ (if the original data is written as センセー).

It would be the best to reverse the conversion workflow, make accents originally in hiragana and then it would be possible to convert to katakana deterministically, right?
For that you need to have the original data in hiragana but from what I've seen the accent_dict data contains fields only in katakana, perhaps you cut out hiragana fields?

I prefer to use hiragana in pitch pattern so I can simply use that instead of vocab kana field in Anki.
If it's too hard - don't mind it.
Thanks for your hard work. よろしくお願いいたします。

The path for ipadic if installed via brew install mecab-ipadic in the code is correct only for Intel-based Macs. On Apple silicon, that would default to /opt/homebrew/lib/mecab/dic/ipadic/.

Likewise, if mecab itself had been installed via brew install mecab, subsequently installing mecab-ipadic-neologd by following its install instructions (since as of today there is no Homebrew formula to install it with brew) will leave the dictionary installed at:

• /usr/local/lib/mecab/dic/mecab-ipadic-neologd/ if on an Intel-based Mac
• /opt/homebrew/lib/mecab/dic/mecab-ipadic-neologd/ if on an Apple silicon based Mac.

neither of which are contemplated in the code.