Added divisor_method() to more easily generate divisor functions for prop_allocation() by name.

prop_allocation() is now faster.

Added sps_iterator() to make an iterator that draws a SPS one unit at a time.

Added a new vignette to show how inclusion probabilities are calculated.

inclusion_prob() and sps() now use kit::topn() when available for partial sorting.

inclusion_prob() is now faster with a single stratum, improving the performance of sps() and expected_coverage().

By default, sps_repweights() automatically picks the smallest tau to keep the replicate weights from being negative. Inspired by svrep::make_gen_boot_factors().

Now requires R >= 4.1, which has been the case for a while (closing #2).

Added becomes_ta() to determine the sample size when a unit enter the take-all stratum.

Internal changes to the way classes are instantiated. No user-visible changes.

Updated maintainer email.

Added a vignette.

A single sample size is now recycled for each stratum when drawing samples or calculating inclusion probabilities.

Added a cutoff argument that puts units with sizes above the cutoff into a take-all stratum.

sps objects now inherit from numeric instead of integer.

Replacement methods for length and levels no longer mangle sps objects.