future.apply

• future_lapply() and friends now signal "Caught ... Canceling all iterations ..." as a message instead of a warning.

Bug Fixes

• future_lapply(), future_sapply(), future_vapply(), future_tapply(), future_eapply(), future_by(), and future_Filter() ignored ... %seed% seed. This led to random numbers depending on the future backend, the number of workers, and the chunk size. Per-element seeds were still generated, but silently dropped. Note that future_apply(), future_mapply(), future_.mapply(), and future_Map() were not affected.
• future_vapply() on empty input failed with an error if FUN.VALUE had length greater than one.
• future_vapply() did not always return an object of the same type as FUN.VALUE like vapply() does.
• future_vapply() with FUN.VALUE as a list would recursively flatten the results into an atomic vector instead of returning a list.
• future_lapply() and friends with %seed%, %packages%, ... would leak those options to the next call, if there were zero elements to process.
• future_lapply() and friends with a future.chunk.size less than one would launch futures that processed no elements. Now such chunk sizes result in one element per future.
• future_apply() with argument future.globals as a list could produce a false error on globals exceeding the maximum size allowed.
• future_.mapply() could fail with "formal argument matched by multiple actual arguments" if dots had names also used by future_mapply(), e.g. "MoreArgs".
• future_kernapply() ignored future.* arguments, and any additional arguments stats::kernapply() might take.

Miscellaneous

• Package **stats** is now declared as an imported package instead of just a suggested package.