Changes in version 0.4.0                        

  - Functional cut boundaries are evaluated within each requested group.
    Fixed boundaries retain their numerical meaning. Input rows, class
    and grouping are preserved; grouped data.frame inputs no longer
    become tibbles implicitly.
  - cut3 and cut3_quantile support explicit weights for functional
    boundaries. Weighted quantiles use inverse empirical CDF/type 1;
    unweighted quantiles retain R's type 7. Weights do not affect or get
    read for fixed boundaries.
  - Explicit quantile probabilities are passed once. Duplicate
    boundaries collapse and invalid weight or label contracts fail
    clearly.
  - vars_to_date handles leap years, annual/monthly/quarterly periods,
    EViews labels, numeric origins and existing Date values. Kept
    component columns retain their original values. Missing or
    impossible components fail explicitly.
  - describe always returns a data.frame for single-variable input.
    Mixed summaries align by statistic name and Date summaries retain
    calendar precision.
  - Unpacking operators validate targets, lengths and locked/active
    bindings before assigning. Explicit environments and warning flags
    work as documented.
  - db_connect and the deprecated dbConnect alias are exported.
    PostgreSQL driver options are forwarded. Trusted SQL Server skips
    credentials; explicit credentials take precedence over keyring, and
    multiple accounts require an explicit username.
  - Loading the package no longer changes global conflict preferences or
    emits the unrelated countrypops warning. R >= 4.1 is declared
    explicitly.
  - Public API regression tests, bilingual guides and reference,
    source/binary packaging and manual deployment workflows are
    included.

Functional grouped and weighted cuts can change classifications relative
to 0.3.11. Reproduce historical analyses with recorded versions and
quantile rules. Unexported historical time-series and plotting helpers
remain internal.