Template talk:sa-conj

Accent in Thematic Verbs

@Dragonoid76: The accent for Class I thematic verbs is wrongly being placed on the thematic vowel, as can be seen in तपति (tápati). --RichardW57m (talk) 10:09, 15 September 2023 (UTC)

@Dragonoid76: There are a number of things going wrong.

1. There's no specification of the writing system of the inputs. The assumption in the code is SLP1, whereas the natural assumption, used in quite a few calls, is that it is IAST. This messes up the detection of accents - the use of normal accents will not be detected!

2. The detection test for accented thematic vowels is:

oxy = match(prov, "(" .. sa_utils.accent .. "?)ti$") and "/" or ""

For any 3s ending in -ti, this will yield "/", which will be transliterated to a combinining acute.

If we convert the inputs to SLP1, or at least what we call SLP1, I think we can then correct that expression to

oxy = match(prov, "(" .. sa_utils.accent .. "?)ti$")[1] or ""

but that needs to be checked, especially for the very rare, probably non-existent case of the thematic vowel having an independent svarita. It seems that we don't use the standard SLP1, but (a) swap the notations anudatta and svarita around; (b) put them after rather than before the vowel!

There are so many layers of obfuscation around that I may have missed something. --RichardW57m (talk) 17:23, 15 September 2023 (UTC)

I based the accent detection on Module:sa-decl/data, and it should be converting IAST to SLP1 in the first step of sc_cache.tr. Dragonoid76 (talk) 20:04, 15 September 2023 (UTC)

You're right that this module needs a rework, though. One other issue currently is that for atmanepada verbs, {{sa-conj|pres|gamyáte}} won't work—we have to do {{sa-conj|pres|gamyáti|n=m}} to get the module to detect the stem.

When I get some free time soon, I'll try to simplify the module and address these things. Dragonoid76 (talk) 20:07, 15 September 2023 (UTC)

I just quickly made a couple of changes. Check out the latest edit of Module:sa-verb/data. This addresses

• Now able to do {{sa-conj|pres|gamyáte}} and have it automatically detect atmanepada and place the accent correctly

Present: गम्यते (gamyáte)
	Active			Mediopassive
	Singular	Dual	Plural	Singular	Dual	Plural
Indicative
Third	- -	- -	- -	गम्यते gamyáte	गम्येते gamyéte	गम्यन्ते gamyánte
Second	- -	- -	- -	गम्यसे gamyáse	गम्येथे gamyéthe	गम्यध्वे gamyádhve
First	- -	- -	- -	गम्ये gamyé	गम्यावहे gamyā́vahe	गम्यामहे gamyā́mahe
Imperative
Third	- -	- -	- -	गम्यताम् gamyátām	गम्येताम् gamyétām	गम्यन्ताम् gamyántām
Second	- -	- -	- -	गम्यस्व gamyásva	गम्येथाम् gamyéthām	गम्यध्वम् gamyádhvam
First	- -	- -	- -	गम्यै gamyaí	गम्यावहै gamyā́vahai	गम्यामहै gamyā́mahai
Optative/Potential
Third	- -	- -	- -	गम्येत gamyéta	गम्येयाताम् gamyéyātām	गम्येरन् gamyéran
Second	- -	- -	- -	गम्येथाः gamyéthāḥ	गम्येयाथाम् gamyéyāthām	गम्येध्वम् gamyédhvam
First	- -	- -	- -	गम्येय gamyéya	गम्येवहि gamyévahi	गम्येमहि gamyémahi
Participles
	- -			गम्यमान gamyámāna

• The issue with the accent in tápati was not with SLP1/IAST but an accent-detection issue in the code. This is fixed now:

Present: तपति (tápati), तपते (tápate)
	Active			Mediopassive
	Singular	Dual	Plural	Singular	Dual	Plural
Indicative
Third	तपति tápati	तपतः tápataḥ	तपन्ति tápanti	तपते tápate	तपेते tápete	तपन्ते tápante
Second	तपसि tápasi	तपथः tápathaḥ	तपथ tápatha	तपसे tápase	तपेथे tápethe	तपध्वे tápadhve
First	तपामि tápāmi	तपावः tápāvaḥ	तपामः tápāmaḥ	तपे tápe	तपावहे tápāvahe	तपामहे tápāmahe
Imperative
Third	तपतु tápatu	तपताम् tápatām	तपन्तु tápantu	तपताम् tápatām	तपेताम् tápetām	तपन्ताम् tápantām
Second	तप tápa	तपतम् tápatam	तपत tápata	तपस्व tápasva	तपेथाम् tápethām	तपध्वम् tápadhvam
First	तपानि tápāni	तपाव tápāva	तपाम tápāma	तपै tápai	तपावहै tápāvahai	तपामहै tápāmahai
Optative/Potential
Third	तपेत् tápet	तपेताम् tápetām	तपेयुः tápeyuḥ	तपेत tápeta	तपेयाताम् tápeyātām	तपेरन् táperan
Second	तपेः tápeḥ	तपेतम् tápetam	तपेत tápeta	तपेथाः tápethāḥ	तपेयाथाम् tápeyāthām	तपेध्वम् tápedhvam
First	तपेयम् tápeyam	तपेव tápeva	तपेम tápema	तपेय tápeya	तपेवहि tápevahi	तपेमहि tápemahi
Participles
	तपत् tápat			तपमान tápamāna

Dragonoid76 (talk) 21:34, 15 September 2023 (UTC)

Present stems in -j

@Dragonoid76: If the 3s ends in -kti and the 3p in -janti, should the logic deduce that the strong stem ends in -j? A case in point is

{{sa-conj|pres|yunákti|yuñjánti}}

Present: युनक्ति (yunákti), युङ्क्ते (yuṅkté)
	Active			Mediopassive
	Singular	Dual	Plural	Singular	Dual	Plural
Indicative
Third	युनक्ति yunákti	युङ्क्तः yuṅktáḥ	युञ्जन्ति yuñjánti	युङ्क्ते yuṅkté	युञ्जाते yuñjā́te	युञ्जते yuñjáte
Second	युनक्षि yunákṣi	युङ्क्थः yuṅktháḥ	युङ्क्थ yuṅkthá	युङ्क्षे yuṅkṣé	युञ्जाथे yuñjā́the	युङ्ग्ध्वे yuṅgdhvé
First	युनज्मि yunájmi	युञ्ज्वः yuñjváḥ	युञ्ज्मः yuñjmáḥ	युञ्जे yuñjé	युञ्ज्वहे yuñjváhe	युञ्ज्महे yuñjmáhe
Imperative
Third	युनक्तु yunáktu	युङ्क्ताम् yuṅktā́m	युञ्जन्तु yuñjántu	युङ्क्ताम् yuṅktā́m	युञ्जाताम् yuñjā́tām	युञ्जताम् yuñjátām
Second	युङ्ग्धि yuṅgdhí	युङ्क्तम् yuṅktám	युङ्क्त yuṅktá	युङ्क्ष्व yuṅkṣvá	युञ्जाथाम् yuñjā́thām	युङ्ग्ध्वम् yuṅgdhvám
First	युनजानि yunájāni	युनजाव yunájāva	युनजाम yunájāma	युनजै yunájai	युनजावहै yunájāvahai	युनजामहै yunájāmahai
Optative/Potential
Third	युञ्ज्यात् yuñjyā́t	युञ्ज्याताम् yuñjyā́tām	युञ्ज्युः yuñjyúḥ	युञ्जीत yuñjītá	युञ्जीयाताम् yuñjīyā́tām	युञ्जीरन् yuñjīrán
Second	युञ्ज्याः yuñjyā́ḥ	युञ्ज्यातम् yuñjyā́tam	युञ्ज्यात yuñjyā́ta	युञ्जीथाः yuñjīthā́ḥ	युञ्जीयाथाम् yuñjīyā́thām	युञ्जीध्वम् yuñjīdhvám
First	युञ्ज्याम् yuñjyā́m	युञ्ज्याव yuñjyā́va	युञ्ज्याम yuñjyā́ma	युञ्जीय yuñjīyá	युञ्जीवहि yuñjīváhi	युञ्जीमहि yuñjīmáhi
Participles
	युञ्जत् yuñját			युञ्जान yuñjāná

, which gets seven forms wrong and would get the subjunctive wrong if it were included. The hack

{{sa-conj|pres|yunájti|yuñjánti}}

Present: युनक्ति (yunákti), युङ्क्ते (yuṅkté)
	Active			Mediopassive
	Singular	Dual	Plural	Singular	Dual	Plural
Indicative
Third	युनक्ति yunákti	युङ्क्तः yuṅktáḥ	युञ्जन्ति yuñjánti	युङ्क्ते yuṅkté	युञ्जाते yuñjā́te	युञ्जते yuñjáte
Second	युनक्षि yunákṣi	युङ्क्थः yuṅktháḥ	युङ्क्थ yuṅkthá	युङ्क्षे yuṅkṣé	युञ्जाथे yuñjā́the	युङ्ग्ध्वे yuṅgdhvé
First	युनज्मि yunájmi	युञ्ज्वः yuñjváḥ	युञ्ज्मः yuñjmáḥ	युञ्जे yuñjé	युञ्ज्वहे yuñjváhe	युञ्ज्महे yuñjmáhe
Imperative
Third	युनक्तु yunáktu	युङ्क्ताम् yuṅktā́m	युञ्जन्तु yuñjántu	युङ्क्ताम् yuṅktā́m	युञ्जाताम् yuñjā́tām	युञ्जताम् yuñjátām
Second	युङ्ग्धि yuṅgdhí	युङ्क्तम् yuṅktám	युङ्क्त yuṅktá	युङ्क्ष्व yuṅkṣvá	युञ्जाथाम् yuñjā́thām	युङ्ग्ध्वम् yuṅgdhvám
First	युनजानि yunájāni	युनजाव yunájāva	युनजाम yunájāma	युनजै yunájai	युनजावहै yunájāvahai	युनजामहै yunájāmahai
Optative/Potential
Third	युञ्ज्यात् yuñjyā́t	युञ्ज्याताम् yuñjyā́tām	युञ्ज्युः yuñjyúḥ	युञ्जीत yuñjītá	युञ्जीयाताम् yuñjīyā́tām	युञ्जीरन् yuñjīrán
Second	युञ्ज्याः yuñjyā́ḥ	युञ्ज्यातम् yuñjyā́tam	युञ्ज्यात yuñjyā́ta	युञ्जीथाः yuñjīthā́ḥ	युञ्जीयाथाम् yuñjīyā́thām	युञ्जीध्वम् yuñjīdhvám
First	युञ्ज्याम् yuñjyā́m	युञ्ज्याव yuñjyā́va	युञ्ज्याम yuñjyā́ma	युञ्जीय yuñjīyá	युञ्जीवहि yuñjīváhi	युञ्जीमहि yuñjīmáhi
Participles
	युञ्जत् yuñját			युञ्जान yuñjāná

works. Should it be encouraged? Incidentally, it would be helpful if the documentation said something about the marking of accents. --RichardW57m (talk) 11:22, 15 September 2023 (UTC)

Currently the method of doing this is with {{sa-conj|pres|yunákti<yunáj>|yuṅkté<yuñj>}}. Needs to be documented better. Dragonoid76 (talk) 19:59, 15 September 2023 (UTC)

@Dragonoid76: Ah, the solution is that stems can be provided honestly as a supplement to the 3s. That information was buried in an example, moreover, in what was apparently an example of specifying the passive. The error message needs to be much improved. --RichardW57m (talk) 14:44, 15 September 2023 (UTC)

Common talk with module

Should we redirect the talk page for Module:sa-verb to this talk page? --RichardW57m (talk) 11:35, 15 September 2023 (UTC)

Seems like a good idea. I've done it. — Eru·tuon 05:52, 16 December 2023 (UTC)

Athematic imperfect 3p middle

I'm assuming that the talk page of Module:sa-verb/data will also be redirected here.

In Module:sa-verb/data, I corrected the athematic imperfect 3p middle from -atAm to -ata. With a bit of hacking for roots not ending in a dental stop or a vowel, it now generates the forms correctly, e.g.

{{sa-conj|impf|áyunajt|áyuñjan}}

Imperfect: अयुनक् (áyunak), अयुङ्क्त (áyuṅkta)
	Active			Mediopassive
	Singular	Dual	Plural	Singular	Dual	Plural
Indicative
Third	अयुनक् áyunak	अयुङ्क्ताम् áyuṅktām	अयुञ्जन् áyuñjan	अयुङ्क्त áyuṅkta	अयुञ्जाताम् áyuñjātām	अयुञ्जत áyuñjata
Second	अयुनक् áyunak	अयुङ्क्तम् áyuṅktam	अयुङ्क्त áyuṅkta	अयुङ्क्थाः áyuṅkthāḥ	अयुञ्जाथाम् áyuñjāthām	अयुङ्ग्ध्वम् áyuṅgdhvam
First	अयुनज्तम् áyunajtam	अयुञ्ज्व áyuñjva	अयुञ्ज्म áyuñjma	अयुञ्जि áyuñji	अयुञ्ज्वहि áyuñjvahi	अयुञ्ज्महि áyuñjmahi

. (The hack is to specify the 3s as the 'underlying' áyunajt instead of the near-surface áyunak so that the strong form is identified correctly.) --RichardW57m (talk) 13:37, 15 September 2023 (UTC)

@Dragonoid76: I finally found that there is a way of honestly specifying the stem. See above. --RichardW57m (talk) 14:46, 15 September 2023 (UTC)

Sandhi in Sigmatic Aorist

@Dragonoid76 At हरति (harati), the module is somehow generating aorist अहाःसीत् (ahāḥsīt) from an input áhārṣīt. That looks wrong to me. --RichardW57 (talk) 21:55, 16 December 2023 (UTC)

@RichardW57 pushed an update. The Module:sa-utilities module upon which the whole conjugation thing is based is a bit weird for verbs since it was designed for nouns. I fixed an issue that was forcing that sandhi to happen. Dragonoid76 (talk) 04:50, 18 December 2023 (UTC)

@Dragonoid76: That's because the noun endings starting with consonants in the weak forms undergo external sandhi, whereas verbs undergo internal sandhi. --RichardW57 (talk) 12:39, 18 December 2023 (UTC)

Palatal Sibilant Sandhi

@Dragonoid76: I'm having a problem with

{{sa-conj|nonf|ākruśya|ākruṣṭá}}

Forms of Sanskrit verbs are numerous and complicated. The following conjugation shows only a subset of all forms and should be treated as a guide.

Nonfinite Forms: आक्रुष्टुम् (ākruṣṭum)
	Undeclinable
Infinitive	आक्रुष्टुम् ākruṣṭum
Gerund	आक्रुष्ट्वा ākruṣṭvā́
	Participles
Masculine/Neuter Gerundive	आक्रुश्य / आक्रुष्टव्य / आक्रुशनीय ākruśya / ākruṣṭavya / ākruśanīya
Feminine Gerundive	आक्रुश्या / आक्रुष्टव्या / आक्रुशनीया ākruśyā / ākruṣṭavyā / ākruśanīyā
Masculine/Neuter Past Passive Participle	आक्रुष्ट ākruṣṭá
Feminine Past Passive Participle	आक्रुष्टा ākruṣṭā́
Masculine/Neuter Past Active Participle	आक्रुष्टवत् ākruṣṭávat
Feminine Past Active Participle	आक्रुष्टवती ākruṣṭávatī

especially with the infinitive, for which I get ākruśtum whereas I would expect something more like ākruṣṭum, and, unless I am overlooking something, as the plain root kruś has infinitive kroṣṭum (source: Whitney's Roots), I would hope to get ākroṣṭum. What is the correct invocation? --RichardW57 (talk) 15:32, 17 December 2023 (UTC)

@RichardW57 The point that at least I got from Wiktionary:Beer_parlour/2023/August#Sanskrit Lemmas is that these non-lemma forms do not belong to the verb since that just makes everything confusing and the non-finite forms are very often completely different from the way the present is formed. The non-finite forms instead belong to the root, and I'm trying to replace older entries with {{sa-conj|nonf}} (e.g. हरति (harati), right now) with Template:sa-root deriv (e.g. तप् (tap), which neatly contains all the non-finite forms, and तपति (tapati), which just contains verbal present forms).

Just to clarify, though, I do think there's an inf= parameter to note what the infinitive should be to override. Dragonoid76 (talk) 04:58, 18 December 2023 (UTC)

@Dragonoid76: I've found the overrides now. However, that doesn't solve the gerundive in -tavya. However, I think there may be some gaps in what actually exists, so perhaps there's more work to do with eliminating forms that don't exist. It looks as though I should bite the bullet and enter some more roots. --RichardW57 (talk) 05:26, 18 December 2023 (UTC)

@RichardW57 Actually, now I've just edited so that हृ (hṛ) is a good example of the new format for roots. Everything is more consistent and now the aorist, future, etc are each their own lemma form with their own conjugations. Dragonoid76 (talk) 05:18, 18 December 2023 (UTC)

Formatting for output of sa-root deriv

@Dragonoid76 I have several issues with the formatting of{{sa-root deriv}}.

1. The output of {{rootsee}} is distinctly ugly, especially compared with the rest of the outputs. I would say that if root entries are maintained, it is better to have |noderiv=1 as the usual setting (not necessarily the default).
2. For the positional parameters, wasn't the plan to have nominal derivatives, '/', '(derived) prefixed roots', '/', 'other terms'? If so, we need a heading for 'other terms'. It's embarassingly missing for special:diff/77238869 (वच्)

There are other problems which make simply invoking {{rootsee}} undesirable.

1. For Devanagari, all the verb's lemmas should end up there, which will make unpleasant duplication.
2. For terms in other scripts, should the root be in Devanagari or the script of the term? On the principle that information should not be duplicated across pages (except for matching links), I think we should not mention roots for non-Devanagari terms unless the term cannot appear in Devanagari. This may be the case for local Sanskrit words, if there be any. For Pali, I push the view that roots should always be in Roman script, as they are abstractions. (On the other hand, names of roots can appear in all scripts for which they are attestable - they are inflected words conforming to the phonology of the language.) I am undecided on whether to bring all the verb forms of a root in the non-principal script together under the root in that script. --RichardW57 (talk) 16:15, 18 December 2023 (UTC)

Class V Imperative

The 2s imperative active in Classical Sanskrit ends in either -hi or is a bare weak stem, according to https://en.wikisource.org/wiki/Sanskrit_Grammar_(Whitney)/Chapter_IX §703. We are showing it as शृणुधि (śṛṇudhi) / शृणुतात् (śṛṇutāt) at शृणोति (śṛṇoti), while the latter apparently isn't attested. An example of शृणु (śṛṇu) can be found as Mahabharat 13.135.12 and I am planning to use it in a quotation. Notifying @Dragonoid76 as the most likely to fix the inflection table correctly. --RichardW57 (talk) 15:26, 4 February 2024 (UTC)

(Discussion on test cases and the need for them has been moved to Module talk:sa-verb/testcases#Requirement. --RichardW57m (talk) 11:22, 6 March 2024 (UTC)

@Exarchus: I've opened a debate on -tāt, with limited scope, at Wiktionary:Requests_for_verification/Non-English#रिणक्ति. --RichardW57m (talk) 12:02, 16 February 2024 (UTC)

@Dragonoid76 The quotation is now installed at ஶ𑍃ணோதி (śṛṇoti). It took me a lot of effort to tease out a tolerable translation. --RichardW57 (talk) 00:21, 7 February 2024 (UTC)

@RichardW57 why is the display of ligatures in the Tamil script so off? I saw this discussion occurring in the other thread—was this not fixed? Dragonoid76 (talk) 18:06, 9 February 2024 (UTC)

@Dragonoid76 Can you please raise the issue at Module_talk:sa-convert, and say more precisely what the problem is. We can generally get characters to appear in the right order, though the fine positioning of the non-Tamil script diacritics leaves something to be desired,and the dotted circles are undesirable. I'm principally monitoring the situation at Module:sa-convert/testcases/Tamil. --RichardW57 (talk) 19:36, 9 February 2024 (UTC)

क्रामति

(Notifying AryamanA, Bhagadatta, Svartava, JohnC5, Kutchkutch, Inqilābī, Getsnoopy, Rishabhbhat, Dragonoid76): For the most part, this word has separate stems for the present active and present middle. Is there a better solution than separate tables for the active and present? The combination looks quite ugly. On the other hand, I can't give any good similar examples. --RichardW57 (talk) 12:20, 10 February 2024 (UTC)

Weak Stems

As I develop the test cases, I'm seeing a lot of problems that could be solved, or greatly alleviated, by having two weak stems - one for before consonants, and another for before vowels.:

1. For हन्ति (hanti) (class 2, oddball): ha- before consonants, ghn- before vowels. For this, there is actually a third variant, for before m, v and y: han-. The environment of -y is perhaps unusual - that could be handled instead by specifying optative 3rd singulars and propagating from them.
2. For reduplicating presents (class 3), such as बिभर्ति (bibharti). Before consonants, we have an unaccented weak stem, bibhR-, but before vowels we have an accented(!) weak stem bíbhr-.
3. For class 5 stems with roots ending in consonants, such as आप्नुति (āpnuti), we may need a pre-consonantal weak stem āpnu- and a prevocalic weak stem āpnuv-. It might be possible to specialise the sandhi rules to just use the former, but I can imagine irresoluble cases, though again, a special sandhi flag might solve them.

For now, I'm just sharing my thoughts - let's get the tests set up first before we change the code. --RichardW57m (talk) 12:41, 14 February 2024 (UTC)

And also:

4. Class 9 weak stems - their final ī drops before vowels. --RichardW57m (talk) 13:28, 14 February 2024 (UTC)

Unusual 3rd Plurals

Two thoughts occur to me here. Firstly, we should recognise 3rd plurals in -uḥ and -ati. Secondly, when the weak root is given explicitly, we should still use the supplied 3d plural, at least if its voice is right, to generate the 3rd person plural rather than generating it from the weak root. The latter trick could also help with athematic 3s active. --RichardW57m (talk) 12:43, 14 February 2024 (UTC)

@Dragonoid76, Exarchus In a similar vein, class 6 (imperfect) injunctives aren't being recognised as thematic. (In coders' terms, imperfect 3s -át is not being recognised as thematic.) ---RichardW57m (talk) 18:11, 15 February 2024 (UTC)

We can also add -án for augmentless athematic imperfects. --RichardW57 (talk) 01:24, 16 February 2024 (UTC)

Weak lemmas in -Cyanti and -Cvanti etc.

@Dragonoid76 Have you tried extracting the weak stems as -Ci and -Cv instead of -Cy and -Cv? We probably need a large test suite before we try it, though. --RichardW57m (talk) 12:47, 14 February 2024 (UTC)

@RichardW57m can you give an example of a word like this? Dragonoid76 (talk) 14:00, 14 February 2024 (UTC)

All class 5 and 8 verbs, e.g. शृणोति (śṛṇoti); जुहोति (juhoti); stretching the boundary condition एति (eti) (though its compounds may meet the condition, e.g. उदेति (udeti)), and extending to other semivowels, बिभर्ति (bibharti). For the reduplicating verbs amongst them, it may be less worthwhile, as we may end up with two weak stems for accentual reasons. --RichardW57m (talk) 14:47, 14 February 2024 (UTC)

@Dragonoid76: Indeed, you added such a manual stem override for तनोति (tanoti). I don't understand what your 3rd argument 'átanvnt<atanu>' for the imperfect is - the remnant of a discarded idea? --RichardW57m (talk) 15:56, 14 February 2024 (UTC)

@RichardW57m Ah okay I understand what you were meaning now. The manual override stem with <> is the current way to do this, and `átanvnt` is a discarded idea. It is possible to just call with

{{sa-conj|impf|átanot|<átanu>}}

The <> indicates that we're supplying the stem rather than the third-person form.

This usually works, but again, this does need to be better documented and tested. Dragonoid76 (talk) 16:30, 14 February 2024 (UTC)

You're rather assuming that the 3rd plural can be formed from the weak stem, which isn't always the case. Have you been looking at the test case results? At present, I'm just working through the tabulated verbs in Whitney, which is worth doing in its own right. I think we will also have to compare against the verbs shorn of irregularities, and I am thinking about a method of avoiding duplication of complete verb paradigms. For example, I might have a subtable impra for the recorded imperative active, and impra_reg for the regular imperative active where this differs, and a function {{code|lua|regular()} that selects the regular forms for comparison.

I fear we also need a guide to creating the template calls for the conjugation tables, though working through Whitney will make it clearer what is needed. I have several times thought about documenting what I do for Pali, but there I have the benefit of a list of attested canonical Pali terms, and I'm only documenting Pali aorists as I meet them. I really do think that for Sanskrit, we want to be able to supplement regular forms rather than just replacing them. For that, in Pali, I have an additional parameter to say whether the additional forms replace or supplement the derived forms. For Pali, it defaults to supplement and list after. For Sanskrit, I suspect we're locked into making replacement the default. Having the option to supplement also makes it easier to change our minds about what the regular 2nd person imperative active is. --RichardW57m (talk) 10:06, 15 February 2024 (UTC)

It fails even for the present tense. If I use {{sa-conj|pres|tanót<tanó>|<tanu>}}, the weak forms don't get an accent. --RichardW57 (talk) 01:54, 16 February 2024 (UTC)

have module detect class of verb?

@RichardW57m Could it be an idea to rewrite the whole module and have it detect the class the verb is in, based on the category a page is in? (or else ask for it) And then write conjugation rules per class (and combine where possible). Otherwise I fear the code will be a very convoluted affair. Exarchus (talk) 19:23, 16 February 2024 (UTC)

@Exarchus: Page category alone is unlikely to work - I believe two differently conjugated verbs may have the same third singular. Detecting the class seems to have been considered, but I think we will need overrides for that to be available. In the present system, most works from 3 stems - strong, weak before consonants and weak before vowels. The two weak stems are usually, but not always, the same. As I've said above, ignoring the 3p form seems crazy. Classes 5 and 8 have some syncopation rules - optional for num/v and mandatory for ruv and ruy. I'd like to explore the possibility of generalising them to all stems. I think we could benefit from having individual overrides for the optative active and optative middle, allowing alternative forms. It seems a bit off to have a mandatory override for the 2s imperative active of all class 9 roots ending in consonants.

I think more study of class 3 and 7 roots is needed, at least from me. --RichardW57 (talk) 22:40, 16 February 2024 (UTC)

@Exarchus I suggest recommending |class=0 as the value to prevent misidentifying उत्स्नाति (utsnāti, “to step out of water”) as a class 9 verb. --RichardW57 (talk) 01:01, 11 March 2024 (UTC)

@RichardW57 I've also been thinking about using the 3p after all, which would exclude such cases (-ānti vs. -anti). But then there's this silly thing that I wouldn't be able to use the 'detect_strong / _weak' function for two conditions at once. Exarchus (talk) 08:03, 11 March 2024 (UTC)

Now I see I've been doing exactly that (using 'detect_' twice) for deponent class 3 verbs...

But there's a general issue with athematic deponent verbs: neither 3s or 3p tells us what the strong stem is for the 1st person imperative. Supplying the strong stem with the 3s currently works. It would obviously be easier if people just gave the active forms, adding "n=m". Exarchus (talk) 10:25, 11 March 2024 (UTC)

I would by the way not be surprised if none of these deponent verbs have this 1st person imperative/subjunctive actually attested... Exarchus (talk) 12:42, 11 March 2024 (UTC)

The only attestation in the DCS that I could find (from the athematic ātmanepada verbs on Wiktionary) is 'āsāmahai' from 'āste', where the same stem happens to be used. What can be considered is to not produce the 1st person imperative for deponent verbs, unless the stem is specifically supplied with the 3s. Exarchus (talk) 13:18, 11 March 2024 (UTC)

@Exarchus, Dragonoid76: The absence of strong form parameter should already result in the non-generation of strong forms. For example, this happens automatically with {{sa-conj|pres|inddhe|indhate}}. Or is that new behaviour? --RichardW57m (talk) 18:01, 11 March 2024 (UTC)

@RichardW57m this is new, previously the so-called 'strong stem' (middle 3s) was used Exarchus (talk) 18:04, 11 March 2024 (UTC)

For identification, I don't think we need to distinguish classes 5 and 8 - we can just include 'r' in the list of consonants before 'oti'. --RichardW57 (talk) 01:01, 11 March 2024 (UTC)

@Exarchus: Why are you detecting Class 5/8 verbs using the regexp "[nR]o/?$"? One should really be working from the weak form, especially as only weak forms are available from present system deponents. It also goes badly wrong with karoti, which has kur- in the weak forms. --RichardW57m (talk) 18:08, 11 March 2024 (UTC)

I didn't add karoti there because it's irregular anyway, and the deponents are treated further below (where I have to add this yet) Exarchus (talk) 18:34, 11 March 2024 (UTC)

PeriphrasticFuture

@Dragonoid76 Is the documentation page simply wrong about how to display the periphrastic future? I can't work out how to generate it with {{sa-conj|fut}}, but I can generate it with the unlisted {{sa-conj|pfut}}. --RichardW57 (talk) 21:38, 16 February 2024 (UTC)

Anomaly in Generation of -tāt Imperatives

@Exarchus, Dragonoid76 Before we remove this suffix from the imperative proper, we might want to work out how we are getting लीढाट् (līḍhāṭ) from लेढि (leḍhi). The final retroflex is wrong, or at least it contradicts the Laghukaumudīvyākaraṇa at https://www.dsbcproject.org/canon-text/content/881/3011, corresponding to Pāṇini 3.4.112. --RichardW57 (talk) 23:55, 22 February 2024 (UTC)

@RichardW57 fixed in sandhi module Exarchus (talk) 08:35, 23 February 2024 (UTC)

3s Perfect Active

According to the grammars I have consulted, where one would expect the the form CaCaCa, the 3s actually takes the form CaCāCa (in accordance with w:Brugmann's law). However, @Dragonoid76 has generally positively asserted the existence of the 3s form CaCaCa, e.g. चकर (cakara) and ततप (tatapa) at the entries for चकार (cakāra) and तताप (tatāpa). What is the evidence of the existence of these as 3s forms?

I have just implemented Brugmann's Law for the perfects in [[Module:sa-verb/data]! (I'm not very happy with the discordant processing encoded, but I couldn't see a closer approximation to the current style.) I was wondering whether we should derive strong perfect stem CaCaC from CaCāCa rather than telling editors to do it, e.g. by {{sa-conj|perf|cakāra<cakar>|cakruḥ}}. (That verb also has some anomalous weak forms that we haven't yet addressed.) आप (āpa) would have been the only downside I had identified to such processing. When I checked on what we had, I saw these surprising 3s forms, which suggests I may have to back out this change entirely! That would be frustrating, as the changes to the perfect got the number of errors down from 351 to 331.

(Notifying AryamanA, Bhagadatta, Svartava, JohnC5, Kutchkutch, Getsnoopy, Rishabhbhat, Dragonoid76): , @Exarchus. --RichardW57 (talk) 22:15, 3 March 2024 (UTC)

@RichardW57 Looking at the Digital corpus of Sanskrit, there is no evidence for those forms Exarchus (talk) 07:29, 4 March 2024 (UTC)

@Exarchus, Dragonoid76 I've cleaned out the alleged 3s perfects in CaCaCa. --08:36, 7 March 2024 (UTC) RichardW57 (talk) 08:36, 7 March 2024 (UTC)

Vedic forms

A specific Vedic form which needs to be added: active 2/3sg. s-aorist without ī (Whitney §888: "The forms without ī are the only ones found in RV. and K.").

Note that ánaiṣīt from the test forms has Vedic 3sg. 'nait' instead of regular (augmentless) 'nais' Exarchus (talk) 11:07, 5 March 2024 (UTC)

Instead of or in addition to? --RichardW57m (talk) 12:26, 5 March 2024 (UTC)

Whitney actually mentions 'nais' a bit above, but I can't find in the DCS. Anyway, these forms on -t are exceptions, similar to what can happen in the imperfect of verbs from class 2 or 7 (Whitney §555). Exarchus (talk) 13:11, 5 March 2024 (UTC)

How did you get to that DCS link? I couldn't find it when I tried.

Indeed, they're almost so exceptional that I'm not sure they meet the spirit of the Criteria for Inclusion (CFI). However, they do seem regular enough that they may be able to avail themselves of an exception to the principle for regular but rare forms - like the Latin contracted perfects, the ones that shed -ve-, -vē- or -vi-. --RichardW57m (talk) 14:26, 5 March 2024 (UTC)

I look for 'nī' in the query Exarchus (talk) 14:48, 5 March 2024 (UTC)

Another Vedic form definitely to be mentioned: the 1st plural ending -masi instead of -mas (Whitney §548a) Exarchus (talk) 17:13, 5 March 2024 (UTC)

@Exarchus: Or should that be treated as an irregularity of individual verbs? --RichardW57 (talk) 01:34, 6 March 2024 (UTC)

@RichardW57 Certainly not, I'd think, it's simply a property of the oldest language (five times more common than -mas in the Rigveda). Like Epic Greek having infinitives on -μεναι, where Attic Greek has -ναι (e.g. ἔμμεναι (émmenai) vs. εἶναι (eînai)).

Something that could be considered, for both nouns and verbs, is a 'novedic' parameter if words clearly don't occur in Vedic language, to avoid giving those forms. Exarchus (talk) 08:24, 6 March 2024 (UTC)

Set for perfect

I think we need to use |set= to control the linking -i- forms for the perfect. It would control the use of linking -i- in the weak forms, probably with special treatment for the 2s active and 3p middle. However, the value rules would need to be different to other 'tenses', i.e.

• default = true
• other values: "both"

(Notifying AryamanA, Bhagadatta, Svartava, JohnC5, Kutchkutch, Getsnoopy, Rishabhbhat, Dragonoid76, RichardW57, Exarchus): --RichardW57m (talk) 12:43, 5 March 2024 (UTC)

@Exarchus, Dragonoid76 Set v. anit seems not to apply to the perfect active participle. Rather, the rules seems to be based on a syllable count for the weak stem - Whitney 803. We might have to require the use of |mono= to override the count when the verb has a prefix. --00:05, 11 March 2024 (UTC)

I think we should consider giving the three (masculine,feminine and neuter) nominative singulars for the active participles, or possibly the Wiktionary stem in place of the neuter. The non-perfect nominative singular neuter is the Wiktionary stem. The problems at present are that the non-perfect nominative singular masculine is not predictable from the nominative singular neuter and that the weak stem of the perfect active participle is not predictable from strong stem. See Whitney 447 and 461 for examples. --RichardW57 (talk) 00:05, 11 March 2024 (UTC)

@RichardW57m For the 2s perfect of jagā́ma and jaghā́na, the testcases give 'jagántha jagámitha' and 'jaghántha jaghánitha', and this is what sanskrit.inria.fr gives, but according to the rules given by Whitney (800 h), this should rather be 'jagántha jagmithá' and 'jaghántha jaghnithá' (so the last forms from the weak stem).

Only the forms without 'i' are attested in the DCS (although I suspect this corpus is not exhaustive). Exarchus (talk) 18:40, 14 March 2024 (UTC)

But Whitney §794g is not so clear about this and could suggest that 'jajñithá' is not accurate either. I'm inclined to restrict using the weak stem before -itha to the forms on -au (roots on -ā) and to the verbs like tatāna (weak stem 'ten'). Exarchus (talk) 23:54, 14 March 2024 (UTC)

@Exarchus: Whitney 794g says 'in general'. That means that there are exceptions, and he lists several at 794e and 794f. There's a claim that weak stems in -e- were generalised from roots starting with 's', so it's not surprising if older forms preserving the consonant reduplication survived into Vedic. I do however agree that the rule needs to be thoroughly checked. I think a lot of 2s forms will require manual overrides, which is why I want to allow addition of forms rather than always forcing complete replacement. (And for Pali, I do prevent duplication of forms.) --RichardW57m (talk) 12:01, 15 March 2024 (UTC)

The form जगमिथ (jagamitha) does seem to be a grammarians' form. Whitney's Roots has it in square brackets, as prescribed but unattested. Will you raise the RfV? I'm surprised the grammarians' endorsement seems not to have caused it to be used. --RichardW57m (talk) 11:35, 15 March 2024 (UTC)

I added the weak stem to जगाम (jagāma), and we've now got that form twice in the table. We may need some anti-duplication logic adding to the compilation of forms. The logic will need to at least warn if duplicates have different notes. --RichardW57m (talk) 11:35, 15 March 2024 (UTC)

@RichardW57m Only seeing this now. Well, the way the module currently works is that 'set' doesn't need to be added manually. Rather, when a verb is 'anit' (the 8 verbs in Whitney §797c), the weak root like occurring in 1p should be given, like: "cakrus<cakṛ>", or "dudruvus<dudru>" (or 'cakruḥ' etc.). So if dadhā́ra is to be considered a set-verb, then just {{sa-conj|perf|dadhā́ra|dadrúḥ}} will suffice (3s can in fact also be given as 'dadhára', but I would not recommend this).

So I added a quick fix and now there's 2s 'jaghántha jaghánitha'.

What isn't there (yet) is the functionality to have both with and without 'i'. I would in any case like to see a few examples of verbs actually having both forms before implementing this. Exarchus (talk) 20:25, 15 March 2024 (UTC)

for clarity: the values given for 'set' in the test module are simply the exit values of this parameter Exarchus (talk) 20:39, 15 March 2024 (UTC)

@RichardW57m I had to abandon this idea of marking perfect anit verbs with <cakṛ> etc., as sometimes the weak stem will need to be given for the active participle (for e.g. 'nināya', this is the only form where 'ninī' shows up).

So the idea is now to manually indicate 'set=true' for the relevant roots on -ṛ or -u. Exarchus (talk) 11:06, 17 March 2024 (UTC)

The form "jaghanitha" is not only supported by grammarians, but also turns up in a few texts - unless Google keeps finding a homograph or typo. --RichardW57m (talk) 11:35, 15 March 2024 (UTC)

Replacement of kṣ-aorist by sa-aorist

@Exarchus, Dragonoid76 The concept of the kṣ-aorist had utility for conjugation - especially as the sandhi utility, internal_sandhi in Module:sa-utilities, used not to support the dropping of -s- between stops. (Are you sure that adding support has broken nothing in noun declension? We really need testcases for the noun declension as well - which may highlight a lot of problems with participles.) We need to check the workings of the s-aorist for aorists in -kṣīt and -tsīt (active has test case in acchaitsīt) and their middle-only forms. In particular, I think the detection of the middle of the kṣ-aorist should be restored and extended to support aorist middles in -tta. (For categorisation, these should usually be s-aorists - I don't think any are root-aorists.) --RichardW57m (talk) 10:57, 6 March 2024 (UTC)

You are correct about the medial aorist, I made an edit there. (Is a '-ps' aorist possible?)

I can't really think of something where dropping -s- between stops could cause issues in noun declension. Exarchus (talk) 11:27, 6 March 2024 (UTC)

I think 'akḷpta' (medial) mentioned by Monier-Williams would qualify as 'ps' aorist.

A tricky case is 'arautsīt', medial 'aruddha', which you might add to the testcases.

"sa-conj|aor|arautsīt<araudh>" produces the active correctly, but the medial needs extra code Exarchus (talk) 11:59, 6 March 2024 (UTC)

the fake medial 'arutta<arudh>' would work: "sa-conj|aor|arautsīt<araudh>||arutta<arudh>" Exarchus (talk) 12:09, 6 March 2024 (UTC)

We also want to dig out a tricky s-aorist in -gh or -g, if such exist. I am now suspecting that the old kṣ-aorist was only partially working, and we need to revise the s-aorist medial's regular expressions. --RichardW57m (talk) 13:20, 6 March 2024 (UTC)

@Exarchus: I thought I'd ruled -ps out, but Whitney 834d (under root aorist) gives atapthās, alipta and asṛpta as ambiguous between root aorist and s-aorist. So include it. It does little to no harm. giving the middle of अच्छैत्सीत् (acchaitsīt), which is why I faked the 3s middle as acchitsta in the test cases. (At that point, I hadn't developed check_aorist.) --RichardW57m (talk) 12:13, 6 March 2024 (UTC)

@RichardW57m One thing is that there currently is no middle conjugation for root-aorists (as it generally does not occur), but Whitney is inclined to classify several middle aorists (adhita, ásthita) as root aorists. Those are traditionally classified as s-aorists with omission of s, which is far fetched in the case of roots ending on a vowel.

So a concrete question in the case of 'alipta': is the middle 3p 'alipsata' (s-aorist) or 'alipata' (root)? This is probably in the realm of pure speculation, so I'm not sure what to do here. Exarchus (talk) 13:23, 6 March 2024 (UTC)

I do find 'alipsata' in the Digital corpus of Sanskrit ... Exarchus (talk) 13:33, 6 March 2024 (UTC)

@Exarchus: Whitney's collaborator Lanman didn't seem to find such forms to be firm evidence (https://www.jstor.org/stable/282859), so supporting both interpretations seems to be the right approach. --RichardW57m (talk) 16:59, 7 March 2024 (UTC)

It's clear that 'áyukta' should be classified as root-aorist as those forms are attested (ayuji 1s, not ajukṣi; áyujran 3p, not ayukṣata). So either the aorist-type needs to be indicated here, or the 3p form. Exarchus (talk) 13:52, 6 March 2024 (UTC)

Keeping in mind that all those (potential) middle root-aorists are Vedic. Exarchus (talk) 14:13, 6 March 2024 (UTC)

@RichardW57m I added some code so that middle forms on -ta can be conjugated as root aorist with "aor=root" (the default for [ktp]ta being s-aorist). I've taken -ata as the standard ending for 3p but -ran is also frequent. Forms on -ddha etc. would still need a fake 3s to make it work, for now at least. Exarchus (talk) 08:19, 7 March 2024 (UTC)

I added support for medial aorists on -gdha/ddha/bdha (defaulting to s-aorist), and as an aorist 3s on -ddha necessarily means the stem ends on -dh, this is automatically detected (but manually adding the stem as <...> is still supported) Exarchus (talk) 10:24, 7 March 2024 (UTC)

Incidentally, I had a lot of trouble with my keyboard mapping for Module:sa-verb/testcases - ibus would time edits out, so I lost a lot of final characters, especially diacritics, and to compound it the font didn't support acutes on macrons well. Proofing rather relied on the table generation being correct, which it often isn't. --RichardW57m (talk) 10:57, 6 March 2024 (UTC)

Optional Cluster Reduction

Form generation does not handle optional cluster reduction, which particularly affects class 7 (nasal infix) where we get sequences of plosives after nasals. For testing in Module:sa-verb/testcases, I am applying function adddegem to supplement the generated forms , but ultimately we need to add it somewhere, probably Module:sa-verb/data. We probably need to add a note to these forms, as it seems that they don't occur in modern or European Devanagari printing. Making this fix (and correcting revealed typos) eliminated 8 error reports. (Notifying AryamanA, Bhagadatta, Svartava, JohnC5, Kutchkutch, Getsnoopy, Rishabhbhat, Dragonoid76, RichardW57, Exarchus): --RichardW57m (talk) 12:04, 7 March 2024 (UTC)

dadhāti

@RichardW57m Should I make an extra parameter 'no_bartholomae' for this one verb or will we use overrides? Exarchus (talk) 12:19, 8 March 2024 (UTC)

For just one verb, I am inclined to use overrides. Its accent location is also very anomalous. (Notifying AryamanA, Bhagadatta, Svartava, JohnC5, Kutchkutch, Getsnoopy, Rishabhbhat, Dragonoid76, RichardW57, Exarchus): --RichardW57m (talk) 12:28, 8 March 2024 (UTC)