For those who are using open-source Presage-based predictive keyboard, the latest version is available in Chum repositories. This includes a fix required for work on SFOS 4.2.
Is the fix also in openrepos?
No, it is not. Feels rather dull to start uploading this large number of RPMs there. So, I haven’t made myself to do so and don’t plan to upload it. Other maintainers of the keyboard may take the time for it, I cannot comment on that.
So, if a person, say me, has presage installed from openrepos, which I have since 3.4, I will have to switch to chum?
I’m not sure I’m d’accord. I mean, I really like the idea of source → repo (git → chum → openrepos) but I’m going to try to keep continuity.
All software maintenance is dull, isn’t it? I only ‘really’ enjoy coding. Which I haven’t had any time for for bloody ages. In. No. Small. Part. Because. Of. Chum?
EDIT: Ps. No stress, I’m just really stressed at the moment and trying to keep 3 deployment mechansims (actuall 4. I try to upload to github, too). in play on top of my day job is becoming a bit much.
Re switching to Chum: if you want it now then yes. I have started discussion on whether cut our losses and drop OpenRepos distribution for the keyboard at https://github.com/sailfish-keyboard/sailfishos-presage-predictor/pull/37 . So far, no reply.
As we have dependencies in the compiled plugins (Marisa for example), uploading to OpenRepos would require uploading those as well. Which will end up as having multiple copies of the lib trees all over the place, at least my first impression of it.
After repackaging all the keyboards, databases and dictionaries to make it possible to build them at OBS and testing that it works, I think it is OK to ask users to make a step and
- add Chum repos
zypper ref && zypper up.
- check that the packages which have to switch organization are ok and then force update them.
Now whether these packages will appear at OpenRepos depends on:
- someone will upload them
- OpenRepos and Chum integration will suddenly go full steam ahead
Let’s see if any of these scenarios will happen.
In principle, given the needs of developers, I agree. Given what users are already faced with, I’m skeptical. I’m going to reach out to Basil to see if I can help him. As it happens, I’ve done drupal stuff in the past, so I may actually be able to support him.
In the long run, chum has to feed harbour.
I’m trying to make a Polish keyboard but I’m stuck. I already create some corpus from two books which I’d like to play with. I have hunspell and presage rpm files. But with keyboard I’m stuck.
I followed this: https://github.com/sailfish-keyboard/sailfishos-presage-predictor/tree/master/utils/keyboard
the problem I have is that I can select the keyboard I made but when I open it example in messages I see it’s empty. in journalctl I see nothing.
So does anyone knows how to debug this? How to see what the actual problem is - empty keyboard?
Can you post somewhere the keyboard RPM? Do you have some other presage-based keyboard installed (English, for example)? Does it work?
@rinigus es, I have English and it works fine. Actually PL is based on the English one.
here’s the keyboard file: Upload files for free - keyboard-presage-pl_PL-1.0.0-1.noarch.rpm - ufile.io
There is not available Polish corpus due to copyrights but if you want big word base you can use subcorpuses. NationalCorpusOfPolish - Computational Linguistics in Poland
Anyway I made corpus on my own I have the presage-lang file and hunspell file.
The only problem I have for now to get this at least somehow working is the keyboard which is not showing, there’s only empty square where the kybd should be visible.
Some time ago I was thinking about making Polish keyboard for Presage. So I wrote to authors of Polish corpus and ask them if it’s available. Answer was that is not available due to copyrights of magazines, books etc. Thay send me link to subcorpuses that can be used for keyboard purposes.
oh ok… well I’ve made my own now I’d like to play with it but I can’t get the keyboard working i.e. to show anything
Leave for others.
turns out that I need to restart keyboard service doing:
systemctl --user restart maliit-server
After that I have keyboard with autocompletion - awsome!
I am sure that maliit restart was suggested in some README or we may even recommended reboot. Great that you’ve got it sorted.
I would like to ask you to refrain sending PMs unless there is something that involves your privacy. For example, bug reports including routes from your home, IMEI as a part of submitted logs and such. Even then, issue should be opened publicly and privacy sensitive data sent via PM. Note that when you send publicly, others may respond as well and participate in resolving the issues and save time for developers. If every user of our apps, ports will be sending all issues as PMs, developers will just not have time to develop. So, I am going to ignore your PM as it does not really contain anything that would warrant sending it privately.
I wonder if they have n-grams available as those shouldn’t be hit with the same copyright issue. For example, Russian keyboard n-gram DB was based on published n-grams. Or maybe they can generate one for us using their corpus?
The corpus i found -even thought creating a KB didn’t work out- was from extracting wikipedia with a script.
This might work for you.
Well I’ve made a PL corpus with keyboard just as a sample, based on two books I have. Of course it’s very very small but I’m quite surprised how it works. Still the biggest problem is SFOS keyboard which is not very user friendly. Need to play with it a bit but what I have right now, for me is not bad.