sample_at

Python

Python does not export this function.

Julia

sample_at(p::Provider, t::Real)
sample_at(p::Provider)

sample_at(p::Provider, t::Real) -> NativeDataset
sample_at(p::Provider) -> NativeDataset

The dataset refresh(p, t) returns — data variables AND coordinates, with the same snap-down to the cadence anchor, the same file-local record, the same records_per_sample = 2 bracket and the same end-of-data rules — and it raises what refresh(p, t) raises (spec/prefetch.md §1). It depends on t only through the anchor t snaps to, so two times in one cadence interval give equal samples. The no-argument form is a [CONST] provider’s [materialize](p).

It is safe to call from several threads at once on one provider, which is what [PrefetchGroup] does. In the Python and Rust tracks sample_at is new because their refresh updates a decoded-file buffer; the Julia [Provider] holds no buffer at all (immutable, decoding per sample with its records pushdown), so here it is the same function as [materialize], under the name the cross-track contract gives it. The shared state it does touch is thread-safe: the cache takes a per-blob lock to fill a miss, and every NetCDF-C access is serialised by one process-wide lock.

Source

Rust

Rust does not export this function.