tadashi-aikawa
commented
2 years ago Hi, @chrisgrieser. Thank you for your FR :)
Your suggestion is very amazing, but it would be too hard to implement it for now.
As an alternative, how about a terminal-like feature that completes up to common forward matching characters?
For example.
- organize
- orchestra
- organizing
- organizers
- organizer
- organization
- organizations
If we push org<TAB> , it completes to organiz and shows suggestions as follows.
- organize
- organizing
- organizers
- organizer
- organization
- organizations
In addition, the following issue may also help.
https://github.com/tadashi-aikawa/obsidian-various-complements-plugin/issues/141
chrisgrieser
Drakemoor
vanbang9710
asknet
The Problem After accumulating more words in my custom dictionaries, I noticed a lot of completions actually share a similar "string beginning". For example I have quite a bunch of completions that look like this:
The problem here is that I would have to scroll through quite the list if I wanted to select "innovation studies". Narrowing down the list by typing more letters also is quite inefficient, since I would have to type out "innovation s" reduce the number of completions sufficiently. However, both methods would defeat the purpose of an autocompletion mechanism, since the very idea of of autocompletion is to reduce the amount of typing.
I think the problem is specific to autocompletion of natural languages, since conjugations or plural forms result in words with a similar stem ("organize", "organizing", "organization", ...)
A Suggestion for a Solution I think the best solution to tackle this would be some sort of "two-step-completion". If Various Complements notices that there are many items that have the same beginning, then it should (partially) merge the items into a new item that is consists of the shared string beginning. Selecting that item would complement that partial string, and instantly trigger a second, follow-up suggester, displaying the remaining string of the previously merged items. Let me illustrate with an example. Suppose you have a custom dictionary with the following terms:
Typing "or" would give you seven suggestions, a lot to scroll through if you want to select "organizations". But as six of the items share a string beginning – "organiz" – the suggester could summarize and display two items:
Now, if I select "orchestra", various complements would work as usual. However, if I select "organiz...", Various complements would complement "organiz" and then display a follow-up suggester, displaying these items, which are the "remaining" strings of the items from above, if you cut "organiz".
Now, I could type "a" to reduce the list to two items, and then easily select one of them.
Evaluation of the Suggested Solution Now we can compare the different approaches to see how the suggestion would improve the autocompletion mechanism. Suppose we want to select "organizations" from our list of seven items from above.
or, hitarrow downsix times, hitenter→ 9 keystrokesorganiza, then hitarrow downone time, hitenter→ 10 keystrokesor, hitenterto select "organiz...", typea. You now have "ation" and "ations" left, and pressarrow downandenterto select right word. → 5 keystrokesThat's a reduction of keystrokes by ~50%. And that does not even factor in the fact that the suggested solution also reduces the amount of text your eyes see (the redundant "organiz" is only displayed once, not six times), which probably also makes the "scanning" for the right item quicker. And on top of that, this is even a fairly simple example. In cases like with the screenshot above (all the words starting with "innovat"), the words are longer and there are more items; meaning you would save even more keystrokes in cases like that.
Implementing this feature request would make Various Complements reflect the characteristics of natural languages, instead of simply applying the autocompletion-logic from code editors without considering that text is different from code. This would make Various complements a truly smart completion engine – one that I even do not know in any other app.