Hello everyone, there was one feature I missed on my Xperia 10 III, and I somehow dealt with it, but with my new Jolla Phone, which arrived week ago, I decided to resolve that: Czech text prediction. Czech language is not supported via XT9, so presage looked like a straight forward alternative for that.
Important note: if you need only spellchecking, then with SFOS 5.2, there is following package available:
jolla-keyboard-hunspell-corrector-cs-0.10.4-1.18.1.jolla.aarch64
Install it, and you have the spellchecking fully functional with default Czech keyboard.
But my goal was different, to have working text prediction. So I spent some time on it this weekend, and as of today, I have working solution ready for 5.1 and 5.2, see my MR (just merged):
With tools within sailfish-keyboard · GitHub , the process was quite straightforward.
The most interesting part was to decide which data should be used as a source for text prediction. Then I remembered about Czech novelist and journalist Karel Čapek, who sadly died too early (1938), but until that time wrote many novels, plays, and articles. Based on these circumstances, all his life work is public domain for quite a while. And thanks to the effort of Municipal library of Prague, we have it available in many modern formats 
Specifically following ones were used:
- Katalog | Městská knihovna v Praze (Od člověka k člověku III, 1932-1938)
- Katalog | Městská knihovna v Praze (R.U.R.)
- Katalog | Městská knihovna v Praze (Válka s Mloky)
- Katalog | Městská knihovna v Praze (Hovory s TGM)
- Katalog | Městská knihovna v Praze (Továrna na absolutno)
- Katalog | Městská knihovna v Praze (Bílá nemoc)
- Katalog | Městská knihovna v Praze (Krakatit)
All books are explicitely marked as released under public domain with the exceptions of graphics and quoatation page, which are released under CC BY-NC-SA 3.0. For the purpose of database creation only public domain data were used.
When packages are available in Chum, you’ll need to use:
keyboard-presage-cs_CZ
presage-lang-cs_CZ
(Until that time, its is possible to build the packages using scripts within the repo.)
See its behaviour in messaging app on Jolla Phone:
Also, I added new keyboard layout for Czech language together with the database. The default keyboard on SFOS for Czech contains dead keys. With dead keys used, Q letter is directly above A letter, and so on, which is (at least for me) very unnatural and leads to many typos, so I decided to do minor adjustment and create my own.
I hope someone may find it useful, it was fun
Side note: If you use presage on SFOS 5.2 with any other keyboard, you might noticed prediction doesn’t work. I hit the same problem, it is due to path change for keyboard handlers in 5.2, which should be fixed by following MR (it worked in my case):
