Releases: vibzart/panchang

v0.2.10 — Batch panchanga primitive + faster tests

19 Apr 04:09

@saucam saucam

py_compute_panchanga_at_jds(jds) — batch primitive that computes tithi/nakshatra/yoga/karana at many arbitrary instants in a single FFI round-trip. Benchmarks: 8760 JDs in ~1.1s, vs ~45s for a per-JD loop. Designed for muhurat engines (vivah, etc.) that sample panchanga hourly across a year.
panchang.core.panchang.compute_panchanga_at_jds(jds) — Python wrapper for ergonomic use.

Test suite runs in ~70s instead of ~290s (×ばつ faster) via session-scoped fixtures for expensive year-long computations (festivals, batch, ekadashis). No coverage loss.

Assets 2

19 Apr 03:37

@saucam saucam

compute_ekadashis: Kshaya Ekadashis (where tithi 11/26 never spans sunrise within a lunar month) are now correctly resolved to the preceding Dashami day per Smarta convention, instead of being silently dropped. Prabodhini Ekadashi 2026 (Nov 20) was the motivating case.
compute_ekadashis: Lunar-month labels are now correct. The old sankranti-anchored ±20-day search sometimes returned Ekadashis with the wrong lunar_month when search windows overlapped (e.g., Jan 29 2026 was returned as "Pausha Putrada" when it is actually "Jaya"). The new Amant-bounded search prevents this class of mis-labeling.