[Cuis-dev] [IMPROV] Code completion for backslash Unicode input

[Juan Vuletich]

Hi Gerald,

(inline and abridged)

On 7/30/2023 12:48 PM, Gerald Klix via Cuis-dev wrote:
> Hi juan,
>
> please see below.
>> Hi Gerald,
>>
>> When I tried loading this in a fresh Cuis image and tried to use it I 
>> saw several issues:
>> - Just doing [\] [tab] would not open the completion list
> That is the intended behavior, but as long as the list of names 
> remains short,
> the whole list can be displayed in this case, probably it should. What 
> do you think?

I think all of them should be displayed. For instance, in an empty 
browser enter [7] [a] [tab]. The list will get shorter as you press 
additional keys.

>> - [\] [o] [tab]. Expected a list including \omega, but got a walkback
> Shouldn't happen (of course). Just retested it.
> The issue is that the completer's instance variable 
> "possibleInvalidSelectors"
> is obviously set as a side effect of the selector computation, which 
> does compute
> nothing in this case. Of course there is a solution -- check for nil.
> Maybe Hernán can comment on this one.
>> - Don't remember what I did with the walkback, but when I got rid of 
>> it, the completion list had many entries that didn't match what I had 
>> entered ('\o')
> It uses #correctAgainstDictionary:continuedFrom:, because a I thought
> the names are hard to guess, unless you know Latex.
> Of course this can be changed.
> Probably exact prefix matches should be displayed first.
> Again, what's your opinion?

I'd display _only_ entries matching prefix. Just like any other list 
shown by the completer. I think it is what everybody expects. If one 
doesn't know how to spell something, we enter a reasonable prefix and 
look for it in the list.

When you are at it, please make '\o' match both 'omega' and 'Omega'. And 
please sort them alphabetically by name, or by code point (not sure what 
is better).

>> - Found no way to use it with your new hex input
> Jup, won't work. Actually if you have '\+' the list would be too long,
> comprising more than one code plane (17 according to Wikipedia).
> IHMO, Completion for numerical input does not make sense.
> Maybe -- I am not sure -- if you have '\+*61' we could interpret
> this as a pattern and show 61, 161, 261, ... 9161 ...
> Probably still to long a list.
> I strongly prefer not to implement completion for this case,
> until we know better.

Ok. It just wasn't clear to me.

> I also thought about something like '\+UNICODE_NAME' where 
> UNICODE_NAME is
> taken from UnicodeData.txt: '\, so one write
> "\+LATIN_CAPITAL_LETTER_A_WITH_TILDE" and it will be replaced with "Ã".
> For such a form of input, completion would make sense;
> actually it is unusable without completion.
> Of course, this should not be in the standard image,
> because it will consume a lot of memory.

Maybe... But perhaps a specific dialog that can filter by code page or 
kind of code point, do searches, etc would be better. It could also 
handle code points both in hex and decimal.

> BTW: Is UnicodeData.txt still needed at the image location?
>
>
> Looking forward to your feedback,
>
>
> Gerald

Thanks!

-- 
Juan Vuletich
cuis.st
github.com/jvuletich
researchgate.net/profile/Juan-Vuletich
independent.academia.edu/JuanVuletich
patents.justia.com/inventor/juan-manuel-vuletich
linkedin.com/in/juan-vuletich-75611b3
twitter.com/JuanVuletich