Skip to contents

Rolling grouping with time as the windowing unit

uby_rolling_group_time.Rd

Creates a rolling group UpdateByOp for each column in cols, using time as the windowing unit.

Arguments

ts_col

String denoting the column to use as the timestamp.

cols

String or list of strings denoting the column(s) to operate on. Can be renaming expressions, i.e. “new_col = col”. Default is to compute the rolling group for all non-grouping columns.

rev_time

ISO-8601-formatted string specifying the look-behind window size.

fwd_time

ISO-8601-formatted string specifying the look-ahead window size. Default is 0 seconds.

Value

UpdateByOp to be used in a call to update_by().

Details

This uses ISO-8601 time strings as the reverse and forward window parameters. Negative values are allowed and can be used to generate completely forward or completely reverse windows. A row containing a null value in the timestamp column belongs to no window and will not be considered in the windows of other rows; its output will be null. Here are some examples of window values:

  • rev_time = "PT00:10:00", fwd_time = "PT00:10:00" - contains rows from 10m before through 10m following the current row timestamp (inclusive)

  • rev_time = "PT00:10:00", fwd_time = "-PT00:05:00" - contains rows from 10m before through 5m before the current row timestamp (inclusive), this is a purely backwards looking window

  • rev_time = "-PT00:05:00", fwd_time = "PT00:10:00" - contains rows from 5m following through 10m following the current row timestamp (inclusive), this is a purely forwards looking window

This function acts on aggregation groups specified with the by parameter of the update_by() caller function. The aggregation groups are defined by the unique combinations of values in the by columns. For example, if by = c("A", "B"), then the aggregation groups are defined by the unique combinations of values in the A and B columns.

This function, like other Deephaven uby functions, is a generator function. That is, its output is another function called an UpdateByOp intended to be used in a call to update_by(). This detail is typically hidden from the user. However, it is important to understand this detail for debugging purposes, as the output of a uby function can otherwise seem unexpected.

For more information, see the vignette on uby functions by running vignette("update_by").

Examples

if (FALSE) { # \dontrun{
library(rdeephaven)

# connecting to Deephaven server
client <- Client$new("localhost:10000", auth_type = "psk", auth_token = "my_secret_token")

# create data frame, push to server, retrieve TableHandle
df <- data.frame(
  timeCol = seq.POSIXt(as.POSIXct(Sys.Date()), as.POSIXct(Sys.Date() + 0.01), by = "1 sec")[1:500],
  boolCol = sample(c(TRUE, FALSE), 500, TRUE),
  col1 = sample(10000, size = 500, replace = TRUE),
  col2 = sample(10000, size = 500, replace = TRUE),
  col3 = 1:500
)
th <- client$import_table(df)

# compute rolling group of col1 and col2, grouped by boolCol, using previous 5 seconds, and following 5 seconds
th1 <- th$
  update_by(uby_rolling_group_time(ts_col = "timeCol", cols = c("col1RollGroup = col1", "col2RollGroup = col2"), rev_time = "PT5s", fwd_ticks = "PT5s"), by = "boolCol")

# compute rolling group of col1 and col2, grouped by boolCol and parity of col3, using following 10 seconds
th2 <- th$
  update("col3Parity = col3 % 2")$
  update_by(uby_rolling_group_time(ts_col = "timeCol", cols = c("col1RollGroup = col1", "col2RollGroup = col2"), rev_time = "PT0s", fwd_time = "PT10s"), by = c("boolCol", "col3Parity"))

client$close()
} # }