2026-11-04 Steven E. Pav <shabbychef@gmail.com>
* 0.2.6 on CRAN
* smaller, faster build: prune template instantiations which could never be called.
* compile with RCPP_NO_BOUNDS_CHECK: faster running computations and a smaller, faster build.
* integer and logical input to running moments, comoments and regressions is converted to double.
* running_sum, running_mean and t_ versions on integers: NA is now detected (it was summed as -2^31),
  NA in the window gives NA output when na_rm is false, and overflow gives NA with a warning.
* running_sum, running_mean and t_ versions on doubles with na_rm false: output returns from NA once
  the NA (or NaN value or weight) leaves the window, rather than at the next restart.
* running sums and means of integers with double weights no longer truncate to integer.
* t_running_mean of integers no longer uses integer division.
* t_running_sum and t_running_mean with NA restart_period no longer recompute every window.
* running moments of integers with NA now treat NA correctly.
* checked by LLM, which added unit tests and fixed some bugs.


2026-09-15 Steven E. Pav <shabbychef@gmail.com>
* 0.2.5 on CRAN
* now tests for positive restart_period, throw error otherwise.
* NA restart_period more explicitly means "infinite"
* remove unused compute_se from t_running_tstat
* catch size mismatch in running_co* and t_running_co*
* split out and fix cent_sums man page
* fix default order in centsums and centcosums
* fix documentation for t_running comoments
* adding various size checks.
* fix computation of comoments
* NA variance would have been standardized to 1, fixed.
* fix moments output size for edge cases of few moments.
* as.centsums default value of normalize_wts changed to true.
* (t_)running_cent_moments w/ max_order_only true, max_order=2 previously returned stdev; now returns variance.

2024-11-29 Steven E. Pav <shabbychef@gmail.com>
* 0.2.4 on CRAN
* adding running correlation, covariance, regression coefficients.
* adding t-running correlation, covariance, regression coefficients.

2024-11-07 Steven E. Pav <shabbychef@gmail.com>
* 0.2.3 on CRAN
* adding flag and checking for negative even moments due to roundoff.
* fixing divide by zero which would corrupt results when input was constant.

2024-11-03 Steven E. Pav <shabbychef@gmail.com>
* 0.2.2 on CRAN
* fix t_running_* to act as documented when variable_win is true.

2019-01-29 Steven E. Pav <shabbychef@gmail.com>
* 0.2.1 on CRAN
* fix memory leak for case where the mean only
  need be computed via a Welford object.

2019-01-12 Steven E. Pav <shabbychef@gmail.com>
* 0.2.0 on CRAN
* add std_cumulants.
* add running_sum and running_mean.
* Kahan compensated summation for these.
* Welford object under the hood.
* add weighted moments computation.
* add time-based running window computations.
* some speedups for obviously fast cases: no checking of NA, etc.
* move github figures to location CRAN understands

2016-04-04 Steven E. Pav <shabbychef@gmail.com>
* 0.1.3 on CRAN

2016-03-25 Steven E. Pav <shabbychef@gmail.com>
* start work

