User:Shenyileirob/Sanskrit inflection/Sanskrit declension: numerals

Compounds comprised of tens and units (11~99) are generally formed with the accented units (1, ..., 9) + the unaccented tens (20, ..., 90) (exocentric accentuation for addition). Here alternative stems dvā́- for 2, tráyas- for 3, and aṣṭā́- for 8 are used apart from 8X, where these alternative stems are impossible and the normal ones are used instead. Note that ékādaśa- for 11, ṣóḍaśa- for 16, and ṣáṇṇavati- for 96 are irregular. Note also the absence of n-retroflection in 9X and the voicing alternation in X6.

Above 40, the compound can also be formed with the units taking the normal stem, thus giving rise to the following alternatives:

Simple hundreds, thounsands, etc. can be formed with the unaccented digit value (2, ... , 9) + the accented place value (100, 1000, ...) (endocentric accentuation for multiplication): dviśatá- for 200, trisahásra- for 3000, etc.

1~10 and cardinals depending on these stems are adjectives, and agree in case, gender and number with the nouns they modify, if any. All inflect for case; only 1~4 inflect for gender. (1 is singular, 2 is dual, and the rest are plural.)

20~90 and cardinals depending on these stems are feminine nouns, and can take a noun in the genitive plural as a dependent, or a noun in plural in an agreeing case as an appositive.

100+ and cardinals depending on these stems are neuter nouns, and can take a noun in the genitive plural as a dependent, or a noun in plural in an agreeing case as an appositive.

Declined like sarva-, yet singular only. Acute-accented on the initial syllable.

Declined like ā-adjectives, yet dual only. Acute-accented on the initial syllable.

The acute accent in gen.pl always falls on the final syllable.

3, 4: Note that the ṛ in gen.pl stays short.

6: Note that the second ṇ in gen.pl is not predicted by sandhi rules.

5, 7~19: ▥ is the stem (páñca, saptá, ...); ◇ is the unaccented stem with the final a dropped (pañc, sapt, ...).

8: In addition to the more general paradigm for 5, 7~19, there is an alternative paradigm (*).