Wikimedia Bugzilla is closed!

(In reply to comment #0)
> Some combination of letteres with vowel are incorrect in  Sinhala Wijesekara
> scheme.
> 
> (These are mailed by Sinhala wikipedian)
> Currently:
> 'sf' produce 'ෙජ', it should be 'ජෙ'
> 'fc' produce 'ෙජ', it should be 'ජෙ'
> 'fi' produce 'ෙස', it should be 'සෙ'

This is not correct jv,
'sf' produce 'ෙජ', it should be 'ජෙ'

f=ෙ is correct.
but when it join with other letters like this

'fc' produce 'ෙජ', it should be 'ජෙ'
'fi' produce 'ෙස', it should be 'සෙ'
'fl' produce 'ෙක' it should be 'කෙ'

Yes; Also letter combination isn't working.
eg: " ;|j " gives "ත|ව", instead of "ත්‍ව"
    " u|u " gives "ම|ම" instead of "ම‍්ම"
    " o|O " gives "ද|ධ" instead of "ද‍්ධ"
    " k\O " gives "න\ධ" instead of "න්‍ධ"
There're more cases on this pattern.

Wijesekara is a fixed layout so the character should be types as we want.

(In reply to comment #1)
> 'fc' produce 'ෙජ', it should be 'ජෙ'
> 'fi' produce 'ෙස', it should be 'සෙ'
> 'fl' produce 'ෙක' it should be 'කෙ'

I think the above typing is incorrect. We will be adding the vowel sign 'ෙ' to existing consonant(s). Rather than typing 'fc', 'fi', 'fl', we should type 'cf', 'if', 'lf' producing correct combinations 'ජෙ', 'සෙ', 'කෙ'.


(In reply to comment #2)
> Yes; Also letter combination isn't working.
> eg: " ;|j " gives "ත|ව", instead of "ත්‍ව"
>     " u|u " gives "ම|ම" instead of "ම‍්ම"
>     " o|O " gives "ද|ධ" instead of "ද‍්ධ"
>     " k\O " gives "න\ධ" instead of "න්‍ධ"
> There're more cases on this pattern.

From existing link, http://www.ucsc.lk/ltrl/services/layout/, I have no guide to implement this patterns. Are there any complete implementation of this layout as a help?

(In reply to comment #3)
> From existing link, http://www.ucsc.lk/ltrl/services/layout/, I have no guide
> to implement this patterns. Are there any complete implementation of this
> layout as a help?

GNU/Linux Distros ship this layout for Sinhala. We can refer its behavior or read the source code(package  m17n-contrib, File: /usr/share/m17n/si-wijesekera.mim)

What remains to be done to resolve this issue?

Changed priority and type.

I am starting to review this layout and to add tests for it. First commit: https://gerrit.wikimedia.org/r/15531 . If I find that the current implementation is correct according to the documents that describe the layout, I'll close the bug, and if not, then I'll fix them.

I just checked the m17n version of this input tool. 

> 'sf' produce 'ෙජ', it should be 'ජෙ'
> 'fc' produce 'ෙජ', it should be 'ජෙ'
> 'fi' produce 'ෙස', it should be 'සෙ'

All these are valid. This input method is smart to apply the vowel signs even it is written to the left or right. That means. both cf and fc produce ජෙ . or in otherwords, one can type in unicode order or visual order. Now this is something very complex and non standard. But it is implemented in m17n with lot of code. I doubt Narayam is capable of doing this, but we need to try with some regex to start with.

I'll use these references to check the correctness of the implementation:

* http://www.icta.lk/attachments/658_658_SLS%201134.pdf
* http://cvs.savannah.gnu.org/viewvc/m17n-db/MIM/si-wijesekera.mim?root=m17n&view=markup

I'm reading the documents and I think that I understand the problem now: It is supposed to be a visual layout, where typing order is not necessarily the same as the logical order. I'll write tests and implementation accordingly.

Partial patch submitted here:
https://gerrit.wikimedia.org/r/#/c/16019/

It only fixes the typing of consonants with inherent vowel, pure consonants and vowel letters. Other types of characters will be done later.

This can be better handled by callback(s): bug 35457

(In reply to comment #11)
> https://gerrit.wikimedia.org/r/#/c/16019/
Status Merged

Merged Gerrit change #16254 links here