0 Releases 1 Tag

RSS Feed

• v0.4.0 d2caa6de97

  Compare

  uscogdata 0.4.0

  jared released this 2026-08-10 19:44:34 -04:00 | 21 commits to main since this release

  First release with the full reader surface tuned and measured.

  Features:

  • Cohorts named by state/type predicate instead of a 40k-id IN list (#58).
    4.8x, and at the no-filter floor.
  • limit/offset on cog_gov_search() and cog_balances(), completing the
    pagination surface begun in 0.3.0 (#57).
  • USCOGDATA_DUCKDB_THREADS / _MEMORY_LIMIT so a server can bound the
    connection without reaching into the package namespace (#60).

  Fixes:

  • cog_gov_search() now orders by a total order; population_acs alone left
    ties in scan order, which made a paged sweep unsound.
  • An unknown state abbreviation reaches its curated error instead of base
    R's 'subscript out of bounds'.

  Documentation:

  • The corpus-access table is re-measured against the published corpus. A
    local mirror is 60-80x faster than the remote default, and opening the
    session is the largest single remote cost.

  Reads the published corpus at pipeline_commit 3d28ddd (schema v7,
  FY1967-FY2024, 46,148,034 rows).

  Verified: 1084 tests passing, R CMD check --as-cran clean on 4 platforms.

  Downloads

  • Source Code (ZIP)
  • Source Code (TAR.GZ)