Template talk:mr-decl-noun

direct and oblique form directly visible

In the minimal view I would like to see the direct (nominative) form AND the oblique form of a noun. When expanded I want to see all possible 8 "cases" properly displayed Are there any objections against it? Or why does this template have the actual form? Why is the "ablative case" missing? Undekagon30 (talk) 01:06, 23 December 2023 (UTC)

@Kutchkutch FYI: you seem to be the only one still working on marathi grammar templates and already helped me a lot. I felt today comfortable enough to start editing this template code. I think it can later be reused for the marathi (noun) declension module to come (sooner or later). I renamed the input parameters, and formatted the underlying code to make it better readable, and will also in the next go automize the transliterations, so that this can be a proper base for ongoing work in this area.

What do you think about it? For me it made sense to first document and test the already existing declinations (and learn wiktionary formatting) by smaller steps, rather than changing everything in one go.

It is still a big and quite challenging task to write a (highly configurable) module and get everything in there right. A good quality documentation and better readability doesn't hurt in any case.

I hope I haven't overlooked any transclusions errors which could be introduced by this refactor rename approach (though I checked everything double). If so, please let me know and I fix them immediately. Undekagon30 (talk) 19:18, 22 January 2024 (UTC)

Okay, these were much bigger changes than originally planned. Please feel free to notify me if something has changed to a disadvantage or is not working properly anymore, so that I can fix it, if necessary. I especially would like to keep the automatic transiliteration from the default marathi transliteration module {{Module:mr-translit}} instead of all the different manually maintained ones (in the derived templates). (Maybe it has to be tweaked a bit?) Undekagon30 (talk) 13:55, 23 January 2024 (UTC)

@Undekagon30:

I felt today comfortable enough to start editing this template code...What do you think about it?

Any improvement is helpful considering that it has not been worked on for over five years.

For me it made sense to first document and test the already existing declinations (and learn wiktionary formatting) by smaller steps, rather than changing everything in one go

Seems reasonable to me

It is still a big and quite challenging task to write a (highly configurable) module

That is definitely true considering the complexity of the Hindi module at Module:hi-noun. Hopefully you will be able to finish the task in the coming weeks and months. Kutchkutch (talk) 05:08, 24 January 2024 (UTC)

I already learned so much by messing up everything. First I messed up the whole formatting in my local copy of the template. I am quite happy that it is so easy to undo changes or just copy the original again. One of the recent days I didn't put the category syntax at the right place in the documentation of a template for which I had added the documentation and suddenly all Marathi nouns had no gender anymore (at least the list of entries in the category was empty). It was easy fixable (rollback), but it made me read the docu about noinclude and similar much faster and understand things much quicker, knowing that some 1000 entries have gone wrong in under one second ... I have already made another template variant for the Marathi noun declension where I use for the show/hide button the appropriate way with NavHeader I found somewhere in the docu. It seems to work, only the colour is now grey instead of dark blue. I also have splitted the genitive declension and the basic declensions into two differnt similar structured subtemplates. your link subtemplate also got a twin brother for these new templates. At the moment I keep the explicit transliterations in the 10 submodules and this subtemplate link intact. Either I manage to adapt the transliteration module to do what I want and what the IPA module wants at the same time, or I have to roll back and manually keep this different set of transliterations somewhere stored (which I strongly dislike, but if I reach my limits of coding, understanding and time ore ressources I properly give up and roll back to something which looks mostly how it looked before my changes in terms of transliteration). I at least try my best to get things better and not worse. Undekagon30 (talk) 16:16, 24 January 2024 (UTC)

@Undekagon30:

There is an extra space before the comma in the Devanagari before the genitive neuter singular in the templates at CAT:Marathi noun inflection-table templates. आजोबा#Declension is an unmarked ā-stem that is plural-only. This would be the same as चहा but possibly without the singular.

Regarding transliteration:

It may be better to have the transliteration for ergative plural ंशी to be nśí instead of uśi and the transliteration for locative plural ंत to be t instead of nta. The a at the end of the transliteration for word-final consonant clusters is so that a final ə can be shown in the IPA using

-- schwa force

translit = gsub(translit, "%*", "a")

at Module:mr-IPA. The anusvara ं is transliterated as u before श (ś) due to:

local nasal_assim = {

[…]

['श'] = 'उ', ['ष'] = 'उ', ['स'] = 'उ', ['ह'] = 'उ', }

at Module:mr-translit. Kutchkutch (talk) 16:50, 23 January 2024 (UTC)

@Kutchkutch the spacing is fixed. With आजोबा I will have a look. It is a direct template call. And with the transliteration I will think about how to achieve this best. Maybe I have some further questions to you there ... Undekagon30 (talk) 17:07, 23 January 2024 (UTC)

I still wonder why the transliteration has to have so many mainly phonetically tweaks in it, because at other places it is more straightforward rules which are mostly mapping letters from one alphabet to the other and much less context based rules. What you have done her looks to me more like a phonetical transcription and less like an orthography based transliteration. (see definitions on Wikipedia, but I guess you are familar with them). As a user I would expect to see one of the typical transliteration systems in place (IAST, ISO-15919,

Also that the transliteration module {{mr-translit}} serves as a helper module for the {{mr-ipa}}-transcription is in my personal opinion no reason to implment it the way as it is done, but these things are that way for some years and I will not question the consensus behind it. The book you aree citying there (Nalkar) is from the year 1880(?) and th author is describing the purpos of his romanization method as phonological (for me that is more on the side of transcription, not transliteration). I have never seen the u transliterated for anusvara before certain letters, and he is there also suggestin nv and not u. I especcialy find it confusing and against the idea of transliteration (reversibility), that the result "haus" can stem from different original words (sorry, I have no devanagari keyboard on my laptop here) ... whatever Nalkar doesn't tell how to transliterate the anusvar at ends of words (typical alternativ nutr eending of nouns nowadays in spoken form). In old dictionaries (like Molesworth and Berntsen this was also no issue because these words there always ended in a combination of e-matra and anusvar.) I added three lines of test cases to the transliteration tests, but I don't know what to expect how these 3 words should be transliterated in consisteency with the current approach with what the module is doing now.

I would either suggest consistency to one transliteration standard everywhere or making the module in a way configureable to give one time output which can be used as an intermediate step for IPA-transcription on the one hand and in other contexts (like this one of the declination tables) being as closed to what the user would expect to see compared to other scientific resources transliteerating marathi. EWikipedia has transliteration modules which get an extra parameter which transliteration method to use, I think default there is the ISO standard (?). There is also stated that Devenagari transliteration should be language independent (=same for Sanskrit, Hindi, Marathi, and others). With this last point I also don't agree, because getting some extra phonological information (schwa-dropping differences between these languages, differentiation between nasalized vowels and the different nasal consonants, the different phonological representations of c, j, jh, all of these are valuable informations. It is a balance between reversibility and context dependent mapping according to phonological praxis (which can also differ between speakers region and /or social group, right?)

I am not sure I got my point accross, I would either weaken the phonological exactness of the actual transcription and more towards a point to common usage, what do other modern authors now do. Or second way to fix this problem

Why am I concerned with this question now? Because the moment where all of these declination tables are in a module, I want that the automatic transliteration works, means represents an output near to how it was now manually placed in the 10 different modules. And only in very special exceptional cases (noun stem changes which are transliterated in a certain way or so, I would create an option to manually overwrite the default output). That is at least my idea of it, and getting thee transliteration module doing the right thing would be one step on the road of this module. Another would be to have a proper hyphenation module (shouldn't be that hard). so that the final module doesn't have to do weird regexp stuff to split a word into stem and ending. or at least put that code into a specialized module, dealing with words, their orthography and representation. I see the 3 things (IPA, transliteration and hyphenation more connected with each other, and more like a toolbox), and the final grammar module (noun / adjective declination, verb conjugation, a bit of from them). The first 3 deal with Marathi phonology (and morphology), the second more with syntax.

(And sorry for writing such a long text, I couldn't put all my thoughts together in a shorter way). Undekagon30 (talk) 12:25, 24 January 2024 (UTC)