Deprecated on 2019-05-16. Use drake_plan()
transformations instead. See
https://ropenscilabs.github.io/drake-manual/plans.html#large-plans
for the details.
reduce_by(plan, ..., prefix = "target", begin = "", op = " + ",
end = "", pairwise = TRUE, append = TRUE, filter = NULL,
sep = "_")Workflow plan data frame of prespecified targets.
Symbols, columns of plan to define target groupings.
A reduce_plan() call is applied for each grouping.
Groupings with all NAs in the selector variables are ignored.
Character, prefix for naming the new targets.
Suffixes are generated from the values of the columns
specified in ....
Character, code to place at the beginning of each step in the reduction.
Binary operator to apply in the reduction
Character, code to place at the end of each step in the reduction.
Logical, whether to create multiple
new targets, one for each pair/step in the reduction (TRUE),
or to do the reduction all in one command.
Logical. If TRUE, the output will include the
original rows in the plan argument.
If FALSE, the output will only include the new
targets and commands.
An expression like you would pass to dplyr::filter().
The rows for which filter evaluates to TRUE will be gathered,
and the rest will be excluded from gathering.
Why not just call dplyr::filter() before gather_by()?
Because gather_by(append = TRUE, filter = my_column == "my_value")
gathers on some targets while including all the original targets
in the output. See the examples for a demonstration.
Character scalar, delimiter for creating the names of new targets.
A workflow plan data frame.
Perform several calls to reduce_plan()
based on groupings from columns in the plan,
and then row-bind the new targets to the plan.