LoginSign Up
← Back to all sparks
F

fastpos

INFRA · APIS
Velocity0.0

Finds the Critical Sequential Point of Stability for a Pearson Correlation

github.com/johannes-titz/fastpos ↗

Dormant three years, then a two-line release nobody using it would notice

simulationcorrelationsample-sizercppdormant
◆Current state
fastpos finds the critical point of stability for a Pearson correlation, a simulation problem whose C++ implementation is the entire reason the package exists. Its substantive work concluded in 2022 with 0.5.0, the release prepared during R Journal review, which renamed the precision parameters, moved multicore work to pbapply and let users set corridor limits directly. After three years of silence, 0.6.0 removes an internal restriction to cpp11 and changes index_pop to an integer.
◆Where it's heading
This is a finished piece of research software in low-effort upkeep. The changelog's centre of gravity is the R Journal review process, and once that concluded the package stopped changing. The single release since is toolchain work of the kind that keeps a package compiling rather than anything a user would see.
◆Prediction
Expect nothing beyond occasional compilation or CRAN-check fixes unless the accompanying paper draws requests for other correlation types or resampling schemes.

◆Recent moves

  1. 10mo ago

    cpp11 restriction removed; index_pop now an integer

    Two internal changes after three years of silence, both at the compilation level. Nothing about how the package is called or what it returns changes.

    View source ↗
  2. 4y ago

    Multicore fix for pbmcapply; Windows limited to one core

    Corrects which pbmcapply functions are called and documents that Windows users get a single core. A follow-up patch to the parallelism rework five days earlier.

    View source ↗
  3. 4y ago

    R Journal review release: renamed precision parameters, manual corridor limits

    The package's last substantive release, driven by R Journal review: precision parameters renamed with the originals deprecated, corridor limits settable directly through lower_limit and upper_limit, and parallelism moved from futures to pbapply. Breaking for existing code, and the point at which the API was fixed.

    View source ↗
  4. 5y ago

    More informative test failures; interactive-only progress bars

    Test diagnostics and a progress-bar change required by the R extensions manual. No user-facing effect beyond quieter non-interactive runs.

    View source ↗
  5. 6y ago

    Multicore support via the future package

    First parallel execution, through the future package and an n_cores parameter, with gains visible only at high study counts because of overhead. Superseded two releases later when 0.5.0 moved to pbapply.

    View source ↗
  6. 6y ago

    Exact-rho population generation and C++ progress bar fixes

    create_pop() now generates a bivariate normal with an exact rho, trading speed for accuracy, and the RcppProgress and interrupt handling were repaired so an interrupted run stops cleanly. Parameter naming was also unified on n_studies.

    View source ↗