@poire-z this is 4 lines long, 50% of which are filler ;) say something

Not competent enough to say if it's a good idea or not.
It feels this is supposed to be sdcv job - and not our job to give it 2 words and have it spend 100% more time doing the lookup.
Btw, you could check if word == word:lower() and don't add it if they are equal, to avoid the uneeded work and strip of duplicate resulsts if this happens.

It feels this is supposed to be sdcv job

Indeed. The problem (if it is one) is that there's full fuzzy or nothing but not something in between like "ignore capitalization and diacritics".

In this form it seems to somewhat sabotage the preferences of people who are against fuzzy search.

I agree with Frenzie here; it does seem to go a bit against the principle of exact search.

Btw, I did a quick test of the dictionary preset PR (#13774) and the 'One-time search with preset' feature is pretty cool.
That made me thinking, wouldn't it be awesome if the presets also allowed to select fuzzy search or non-fuzzy search as part of the setting? Then a fuzzy search preset could be selected in cases where a user who normally prefers exact searches wants to look up a word that is capitalized/conjugated.

And to access the presets from the highlight dialog, I would suggest that long-pressing Dictionary opens a dialog to select the list of dictionary presets for a one-time search.

Let's try to get some context around here, with the help of my friend Claude:

In a typical 100,000-word novel, roughly 8,000-12,000 words would be capitalised.
This breaks down approximately as:

* Proper nouns: 3,000-5,000 words (character names, place names, brand names, etc.)
* Sentence beginnings: 4,000-6,000 words (assuming average sentence length of 15-25 words)
* Dialogue and emphasis: 1,000-2,000 words (shouting, emphasis, chapter headings, etc.)

The percentage varies significantly based on genre - fantasy novels with lots of invented proper 
nouns will be higher, whilst literary fiction might be lower. Dialogue-heavy books also tend to have 
more capitalised words due to sentence fragments and exclamations.
This is roughly 8-12% of the total word count, which aligns with general English text patterns.

now our target is the last two bullets there, proper nouns is what Frenzie calls "sabotage the preferences of people who are against fuzzy search". I for one, have become very adamant of not using fuzzy searches as, and please forgive me here, I find it to not be fussy enough (I mean that in the sense of not being selective enough, "picky"). But back to the matter at hand.

It feels this is supposed to be sdcv job

I agree with this, however, sdcv seems to move at tectonic plate speeds, their last commit appears to have been made in Aug 2024 and in that year, it was the only commit committed. sorry. Here is a request for diacritic agnostic searches with a whopping zero replies Dushistov/sdcv#94, so we shouldn't rely on the goodness of their heart.

and not our job to give it 2 words and have it spend 100% more time doing the lookup.

I like to think of it as course correction, it's not really an interference, only making sure 5-8K potential capitalised words get a valid response. Also, we are not really increasing the search by 100%, sure we do if the word is capitalised but the majority of searches will continue to be of normal lower case words, as that makes up about 90% of a text, so although correct, in practise it will only add a bit of time in every 1 out 10 queries.

Btw, you could check if word == word:lower() and don't add it if they are equal

what? how could they be equal? genuinely asking, I mean we are checking for capital letters word:match("^%u") so they should be different, no?

it does seem to go a bit against the principle of exact search.

only it doesn't, or it does but only for those 3-5K proper nouns, but you still get the exact match though (if it exists), you just happen to get another one on top.

The problem (if it is one) is that there's full fuzzy or nothing but not something in between like "ignore capitalization and diacritics".

I was trying to avoid adding another menu entry, I still believe this is a cleaner approach but if that is a preferred solution, I wouldn't be against it. The diacritic though, I don't know about that part, it think that might be best left to a smart, intelligent frenchman.

Not competent enough to say if it's a good idea or not.

well, you need to become competent then, as it's your reader dictionary. ;)

Btw, you could check if word == word:lower() and don't add it if they are equal

what? how could they be equal? genuinely asking, I mean we are checking for capital letters word:match("^%u") so they should be different, no?

Oh, right :) (It wasn't as explicite as my brain state needed, but you are right.)

as it's your reader dictionary. ;)

It's not mine - and I'm so used to fuzzy search, so I have no idea how non-fuzzy people work :)

Well you’re the gatekeeper ;)

The problem with word:lower() is that uppercase letters not available in the ASCII set (Ñ, for example) won't be converted.

Ñapa won't be converted to ñapa. That's why I used Utf8Proc.lowercase(word) in the patch I posted here

that is a valid point I hadn't consider.

Sidenote 'cause I had to leave home but yhe isUppercase function in that patch has the same issue. %l is just a-z

Edit:

> print(("n"):match("%l"))
n
> print(("N"):match("%l"))
nil
> print(("ñ"):match("%l"))
nil
> print(("é"):match("%l"))
nil

in the spirit of user patching, could I do something like this?

if not self.removed_by_patch and index == 1 then

That's probably better if you extract the entire thing into a method?