Row-wise caching of operations on data frame.
flow_dfr(..., fn = NULL, fn_id = NULL, flow_options = get_flow_options())
... | Named arguments to pass to |
---|---|
fn | The function to apply to the data frame. It must accept a data frame as the first argument. |
fn_id | Optional id to uniquely identify the function. By default,
rflow functions reuse the |
flow_options | List of options created using |
The flow object.
Function fn
operates on a data frame received as argument.
fn
will receive only the rows changed;
it may drop some of the rows, but will not add any new rows.
The function fn
may return fewer or more columns or modify
existing columns as long it always returns a consistent schema
(i.e., the same column data types and names) for all calls.
The data frame df
passed to fn
will include one
additional column ..row_hash..
that must be returned as is in
order to identify changes.
Arguments fn
, fn_id
and flow_options
, when provided,
must be named. Argument fn
must be always provided.
df_fn <- function(df, i = NULL) { if (is.null(i)) { dfi <- df dfi$rm <- rowMeans(dfi[1:10]) } else { dfi <- df[i, , drop = FALSE] } dfi } # the flow element can also become input for another flow_df function # in order to allow multiple, chained computations dfr_flow <- flow_dfr(mtcars, 1, fn = df_fn)#>