Virtual Keyboard: Bug when extending autosuggested words

asked 2016-02-01 02:09:03 +0200

G.Ohrner
191 ●6 ●11 ●19

updated 2016-08-17 16:34:52 +0200

jiit
2678 ●32 ●58 ●59

  admin

I observed a behaviour with the virtual keyboards word completion that I consider to be a bug. Possible it might even cause garbled words to be included in the dictionary, but I'm not sure.

I'm struggling a bit with finding a good example using the English keyboard layout, but the problem becomes pretty obvious in the German language.

I'll give an English example first - we want to type "Automobile" into the Jolla notepad application.

Remember that this is no proper example, but it demonstrates the bug. Maybe some reader might come up with a more realistic scenario in the English language, in German language there are plenty due to the language's property of combining several words into single large ones.

1. Switch VK to "en".
2. Enter Au
3. Auto will be suggested, select it from the suggestion list so the entered text now reads Auto.
4. As we want to write Automobile, press backspace to get rid of the trailing space and re-enable suggestion mode.
5. Enter m so the entered Text reads Autom

Now the suggested completions get wired:

• If I enter Autom directly, without the intermediate completion step described above, the first suggestion on my phone is Auto and all subsequent ones start with the prefix I entered, Automatic, Automatically, Automated and so on.
• Using the strategy described above, the first suggestion is Any (although the Text reads Autom at this point) and the second suggestion is Aunt, Aim, Amy, ... .

Long story short, what happens is that the word-suggestion behaves exactly as if I had entered Aum in one go - ie. when reenabling the suggestion mode after pressing backspace, it takes the text which I entered before using the completion (which was Au in this case) and appends all newly added letters to it (so the text it works with becomes Aum) instead of to the word which is actually shown in the text (which is Autom).

I hope you could follow... ;)

In German with words like "computeranimierte Fußballspieler" (combined from the words "Computer", "animierte", "Fuß", "Ball" and "Spieler"), the mentioned iterative completion strategie would accelerate text input a lot if it would yield correct results.

The other question is if the garbled words internally used by the VK in this process (like Aumobile instead of Automobile in my example above) might even somehow end up in and pollute the user dictionary...