Checking the whole buffer

[rrthomas]

In the last few days I've been using jinx for "industrial spell-checking", you might say: checking multi-thousand-word documents all in one go.

It's usable, but it's more work than ispell. I wonder whether I could do better. The workflow I have adopted is to bind jinx-next and jinx-previous, and then repeatedly go to the next word, do jinx-correct, and so on.

Obviously, compared to ispell, the problem is that I need to explicitly run jinx-correct for every misspelled word, as there's no "correction mode". Being able to use jinx-next and jinx-previous from within jinx-correct don't help me, as far as I can see.

I have tried making a macro for jinx-next, jinx-correct, add to local file (because the text I am checking has few errors, so this is usually what I want to do), and that helps a bit, but I then notice that for some reason it doesn't always visit even the visible errors in order: it sometimes goes backwards to a word earlier in the text.

Since, as I said, the text I'm checking has few errors, and mostly I'm adding words to a per-file wordlist, one speedup I have found is to simply use my macro fast, without looking at the words it adds to the LocalWords list, and then going to look at the list and finding words that shouldn't be there. But this is a trick that only works in this sort of situation, and wouldn't work in a more typical (at least for me) scenario where I'm actually expecting a fair number of errors.

One might ask why I don't simply use ispell. Fair question: its speed disadvantage doesn't really matter when one is using ispell-buffer. The reason is that it doesn't do its own word division, and Enchant's word division just isn't very good. Also, I have become used to using jinx, so I'd much rather not have to remember the keystrokes for a different correction interface.

(Mea culpa for the poor word division, obviously, my being the Enchant maintainer; but also I don't really regard doing word division as my problem, and the built-in algorithm is very much "best effort". It could perhaps be improved or at least simplified; at present, it treats hyphenated words as compound, and, until the other day, was treating numbers as words; it still considers any token with acceptable characters and at least one letter to be a word. I could use ispell-buffer directly with aspell.)

[minad]

Just to make sure - you are aware that jinx-correct is a dispatcher command, and there is also jinx-correct-all?

[rrthomas]

Further, it's probably also mentioning jinx-correct-all explicitly, since it's not bound by default. The mention you refer to is in the section about suggested bindings. This could be achieved briefly by amending my suggestion above to "starts spell-checking the entire buffer (jinx-correct-all)". It would be perhaps useful to add similarly a note to the documentation for C-u C-u M-$ saying which command it invokes.

[minad]

Okay, I expanded the README.

[rrthomas]

That's super, many thanks! I think Jinx's ability as a "traditional" or "batch-mode" spell-checker is undersold, and I wish I'd found jinx-correct-all before putting so much work in to get by without it!

[minad]

Hmm, I hadn't thought like this. In my eyes Jinx wouldn't be a "real" spell checker or an worthy Ispell substitute without this feature. Users simply expect this feature to be present, so I find it a little odd that you assumed that Jinx doesn't have it. Jinx had this feature basically from the beginning. It was mentioned in the README, even if a little obfuscated. Also with Vertico or Helm it should be quick to discover the list of available Jinx commands.

So I think this is just one of the cases where you took a wrong turn early on and then stuck to your wrong assumption for long without revalidating it. This is not unusual, I've also made such mistakes and I often see that users who create issues based on false assumptions, even if the README already had an answer, but maybe not sufficiently explicit.

[rrthomas]

Indeed, I had a strong impression that jinx was first and foremost a flyspell replacement, and that its design assumes that ad hoc correction is the dominant mode for most users. This doesn't contradict its being a replacement for ispell, at least to me.

I didn't assume that it didn't have this functionality, I simply couldn't see it documented anywhere. I was in fact surprised I couldn't find it. I double-checked by doing various searches of the README. I searched the issues and discussions. Then I posted.

As you say, it's a pretty important feature, so worth highlighting. Maybe I differ from you on this point: it's precisely because it's so important that it's worth spelling out (pardon the pun!).

And in any case, the README should try to forestall such misunderstanding or overlooking. People will, I guess, come to jinx from many directions. I hope so; I'd still very much like to see it supplant ispell, which I find increasingly dated in design.