If we do it: a producer can pass several series at once (stations, AOIs) through cd_baseline() → cd_anomaly() → cd_trend(), keeping its id and QA columns. If we never do: every caller splits by id, loops, and re-joins its own columns afterwards. That works, but each package writes it again.
Problem
After #92 (PR #94), the consumer functions reject more than one row per variable, period and year (series_check()), and cd_anomaly() returns a fixed set of columns. A table holding several stations therefore errors, and columns such as station_number, n_days or a data-quality fraction are dropped. The first multi-series producer (NewGraphEnvironment/wet#25, per-station flow in date windows) splits by station and re-joins its columns.
Proposed Solution
- A
by = argument (default NULL, today's behaviour) naming id columns. They are added to the grouping in cd_baseline(), cd_anomaly(), cd_trend() and cd_compare(), and to series_check()'s uniqueness key.
cd_anomaly() keeps the by columns in its output. Whether to pass other extra columns through is a separate call.- Tests on a two-series table matching two single-series runs.
If we do it: a producer can pass several series at once (stations, AOIs) through
cd_baseline()→cd_anomaly()→cd_trend(), keeping its id and QA columns. If we never do: every caller splits by id, loops, and re-joins its own columns afterwards. That works, but each package writes it again.Problem
After #92 (PR #94), the consumer functions reject more than one row per
variable,periodandyear(series_check()), andcd_anomaly()returns a fixed set of columns. A table holding several stations therefore errors, and columns such asstation_number,n_daysor a data-quality fraction are dropped. The first multi-series producer (NewGraphEnvironment/wet#25, per-station flow in date windows) splits by station and re-joins its columns.Proposed Solution
by =argument (defaultNULL, today's behaviour) naming id columns. They are added to the grouping incd_baseline(),cd_anomaly(),cd_trend()andcd_compare(), and toseries_check()'s uniqueness key.cd_anomaly()keeps thebycolumns in its output. Whether to pass other extra columns through is a separate call.