@Aryamanarora: Thanks, and I appreciate all the work you've done on Module:mr-decl! Since Module:mr-decl is the replacement for T:mr-decl-noun, that could mean I won't be able contribute as much to the implementation side unless T:mr-decl-noun is still useful. Since it's called Module:mr-decl without '-noun' does that mean it would ideally handle adjective and pronoun declension too?
Module:mr-decl would ideally have the genitive case, but it's fine if it's not handled yet. At Template:mr-decl-noun, I was thinking of removing the locative case for simplicity since it's just a postposition, but that can be done easily at any time.
An old source from Google Books says the default way to decline u-final masculine words is without an oblique, and the oblique is an alternative for some words such as विंचू(viñcū). If the oblique is used for विंचू(viñcū), I see a many search results for विंचवाला, विंचवांना, विंचवाने, विंचांनी, etc without the virama/halant before व.
It looks like you used Dhongde & Wali to make Module:mr-decl, and Dhongde & Wali doesn't say much on plural cases other than Direct Plural. Also, it looks the module can be manually told if a consonant-final feminine nouns is in the exception class with elseif g == "f-e" then . Kutchkutch (talk) 01:42, 1 November 2017 (UTC)Reply
@Kutchkutch: Okay, thanks for the help! I think I will make a function for alternate obliques. Also, I totally forgot about the genitive, I'll implement that as well. And yes, you're right about "f-e". —Aryaman(मुझसे बात करो)10:37, 1 November 2017 (UTC)Reply
@Aryamanarora: Thanks for your willingness to continue! Figuring out how each paradigm works is confusing since direct singular, direct plural, oblique singular and oblique plural could all be different. Kutchkutch (talk) 10:51, 12 November 2017 (UTC)Reply
@Aryamanarora: I can't thank you enough for the work you've done thus far on everything. It's the least I can do so that the focus is more on to how to Luacise this rather than what to Luacise. I noticed that you've even 'adopted' this project to the point that you put {{currently|Marathi decl and translit}} on your userpage. Since User:Vinayak.patwardhan originally created Template:mr-decl-noun when I wasn't around, this seemed like a natural starting point.
I combined the direct plural and & oblique charts in Dhongde & Wali into a single chart in the 'Regular Paradigms' section for clarity. Of course there may be ways to improve this guide, but it appears to be comprehensive for the most part. I'm still wondering if pronoun declension should be handled automatically or entered manually. Kutchkutch (talk)
@Aryamanarora: Sure, this looks great. Kutchkutch's detailed rules are also very helpful. The desclension structure (and the associated schwa deletion) are very similar to Konkani, only the suffixes differ. As already pointed out, the feminine nouns aren't declining properly -- माधवपंडित (talk) 01:54, 15 November 2017 (UTC)Reply
The oblique plural is the hardest 'box' to ascertain, but a quick search shows this is the case for लाट(lāṭ) since even Google suggests 'Did you mean: लाटांवर' when लाटेंवर is searched. Kutchkutch (talk) 02:22, 15 November 2017 (UTC)Reply
There appears to a delay between changes at a module and transmitting those changes to where that module is being used so doing a page preview of a page shows the latest version (The mobile version takes hours).
@Aryamanarora: The त in भक्त appears to be missing.
Also, I realised that oblique plural of शाळा is शाळां- instead of शाळें-. That was an error on my end. Dhongde & Wali don't say much about the oblique plural other than the ं. I wish I would have caught that much earlier. Kutchkutch (talk) 03:00, 15 November 2017 (UTC)Reply
@Aryamanarora: I originally thought C cluster-ending might be unnecessary, but this shows it is necessary.
[1] has an enormous list of words tagged with an example word to show how it's declined, which would be useful for the ambiguous cases if I can figure out the correspondences between their example words and Dhongde & Wali's example words. भक्त is an example word used for tagging in that that list to show that a C cluster-ending masculine word is declined exactly like भक्त. There's other resources at [2], but I haven't figured out a use for them yet. Kutchkutch (talk) 22:30, 15 November 2017 (UTC)Reply
@Kutchkutch: I was actually just now on that site, using their wonderful IndoWordNet tool to fill up the descendants for Sanskritजानाति(jānāti). That list could actually be used in our code if C clusters are too irrgular, much like how {{zh-new}} can automatically generate Mandarin and Cantonese pronunciations (which have no obvious rules) using a database. —Aryaman(मुझसे बात करो)22:35, 15 November 2017 (UTC)Reply
@AryamanA: (Off topic, but It appears your username has changed, which changed the title of this page…)
Thanks for suggesting the idea if it becomes necessary!
The tags in that list includes समई(samaī), which perhaps suggests my guide may not be as complete as I originally thought, but I think the only difference is the lack of the halant/virama in the oblique marker ्या.
Another puzzling tag is वकील(vakīl), which suggests there could be more exceptions to the final closed syllable rule in which case your suggestion might come in handy. However, in that list वकील(vakīl) is the tag for वडील(vaḍīl) so perhaps the list didn't account for the final closed syllable rule.
The module even in it's current state is a spectacular wonder. I know it takes a lot of patience to code, and you've clearly shown that. As I realise how the additional paradigms work, I can update the guide. Kutchkutch (talk) 23:46, 15 November 2017 (UTC)Reply
@Kutchkutch: Out of curiosity I Googled वकलाला(vaklālā) and only got 3 hits and "Did you mean: वकिलाला(vakilālā)". So the stem does change, but I wonder if it is irregular because it's a Persian-derived word.
But then I looked up वडलाशी(vaḍlāśī), and got very few hits. I tried वडिलाशी(vaḍilāśī) and got many thousands of hits. That makes me think the stem is just weakened, not entirely lost, and maybe वडलाशी(vaḍlāśī) dropping the vowel entirely is a colloquial or a recent variant (I only got hits after 2015 for it).
I'll probably implement alternative forms eventually since it seems Marathi is undergoing some standardization right now.
@AryamanA: Earlier, I misunderstood the tag for वडील(vaḍīl). Compounds containing वडील(vaḍīl) such as वाडवडील have the <वकील(vakīl)> tag and वडील(vaḍīl) itself has the <पाटील(pāṭīl)> tag. The more I look at that list, the large amount of exceptions perhaps favours integrating features of the tagging system. What I'll do now is try to combine the advantages of both the Dhongde & Wali and the tagging system.
Calling it a tagging system probably doesn't affect the testcases already implemented because it's just another way of looking at the same system with more descriptive names for each paradigm. For example, <देऊळ(deūḷ)> is a tag in the tagging system and पाऊल(pāūl) has the <देऊळ(deūḷ)> tag so since the <देऊळ(deūḷ)> tag is already implemented perhaps nothing needs to done about it. For an ambiguous case such as consonant-ending feminine nouns, 'f-e' is currently the tag used to tell the module to use the exceptional paradigm. Some of the identical paradigms in Dhongde & Wali are merged into a single tag. For example, खांब(khāmba) and भक्त(bhakta) in Dhongde & Wali are effectively the same paradigm so the tagging system merged them into the <भक्त(bhakta)> tag.
@AryamanA: As already mentioned, the Dhongde & Wali system appears to work well for the unambiguous and common paradigms. The 'CFILT_Tags' system is perhaps a 'Version 2' of the Dhongde & Wali system for the ambiguous and less common paradigms for exception handling. It's interesting to think about how a module could handle those. The tags in the last section are ones that I'm still thinking about. If there's a use for them, I'll add information about them. If they're not useful they can be ignored because there are a few duplicate tags in the CFIL list. Kutchkutch (talk)
@AryamanA: Thanks for your work on the Eyelash र today! The schwa is currently still in the transliteration of the oblique for चेहरा and नोकरी. The Eyelash र is an orthographic way of indicating that the र has moved to the onset of the following syllable with another consonant following it before the next vowel. Kutchkutch (talk) 22:10, 17 November 2017 (UTC)Reply
Note: Some nouns can be declined using the non-exceptional or the exceptional paradigm. The word साळू(sāḷū, “porcupine”) is rare, but despite Dhongde & Wali's choice of word the paradigm it represents is valid.