Set action levels: failure thresholds and functions to invoke

The action_levels() function works with the actions argument that is present in the create_agent() function and in every validation step function (which also has an actions argument). With it, we can provide threshold fail levels for any combination of warn, stop, or notify states.

We can react to any entrance of a state by supplying corresponding functions to the fns argument. They will undergo evaluation at the time when the matching state is entered. If provided to create_agent() then the policies will be applied to every validation step, acting as a default for the validation as a whole.

Calls of action_levels() could also be applied directly to any validation step and this will act as an override if set also in create_agent(). Usage of action_levels() is required to have any useful side effects (i.e., warnings, throwing errors) in the case of validation functions operating directly on data (e.g., mtcars %>% col_vals_lt("mpg", 35)). There are two helper functions that are convenient when using validation functions directly on data (the agent-less workflow): warn_on_fail() and stop_on_fail(). These helpers either warn or stop (default failure threshold for each is set to 1), and, they do so with informative warning or error messages. The stop_on_fail() helper is applied by default when using validation functions directly on data (more information on this is provided in Details).

Usage

action_levels(warn_at = NULL, stop_at = NULL, notify_at = NULL, fns = NULL)

warn_on_fail(warn_at = 1)

stop_on_fail(stop_at = 1)

Arguments

warn_at, stop_at, notify_at: The threshold number or fraction of test units that can provide a fail result before entering the warn, stop, or notify failure states. If this a decimal value between 0 and 1 then it's a proportional failure threshold (e.g., 0.15 indicates that if 15% percent of the test units are found to fail, then the designated failure state is entered). Absolute values starting from 1 can be used instead, and this constitutes an absolute failure threshold (e.g., 10 means that if 10 of the test units are found to fail, the failure state is entered).
fns: A named list of functions that is to be paired with the appropriate failure states. The syntax for this list involves using failure state names from the set of warn, stop, and notify. The functions corresponding to the failure states are provided as formulas (e.g., list(warn = ~ warning("Too many failures.")). A series of expressions for each named state can be used by enclosing the set of statements with { }.

Details

The output of the action_levels() call in actions will be interpreted slightly differently if using an agent or using validation functions directly on a data table. For convenience, when working directly on data, any values supplied to warn_at or stop_at will be automatically given a stock warning() or stop() function. For example using small_table %>% col_is_integer("date") will provide a detailed stop message by default, indicating the reason for the failure. If you were to supply the fns for stop or warn manually then the stock functions would be overridden. Furthermore, if actions is NULL in this workflow (the default), pointblank will use a stop_at value of 1 (providing a detailed, context-specific error message if there are any fail units). We can absolutely suppress this automatic stopping behavior by at each validation step by setting active = FALSE. In this interactive data case, there is no stock function given for notify_at. The notify failure state is less commonly used in this workflow as it is in the agent-based one.

When using an agent, we often opt to not use any functions in fns as the warn, stop, and notify failure states will be reported on when using create_agent_report() (and, usually that's sufficient). Instead, using the end_fns argument is a better choice since that scheme provides useful data on the entire interrogation, allowing for finer control on side effects and reducing potential for duplicating any side effects.

Examples

For these examples, we will use the included small_table dataset.

small_table

## # A tibble: 13 × 8
##    date_time           date           a b             c      d e     f    
##    <dttm>              <date>     <int> <chr>     <dbl>  <dbl> <lgl> <chr>
##  1 2016-01-04 11:00:00 2016-01-04     2 1-bcd-345     3  3423. TRUE  high 
##  2 2016-01-04 00:32:00 2016-01-04     3 5-egh-163     8 10000. TRUE  low  
##  3 2016-01-05 13:32:00 2016-01-05     6 8-kdg-938     3  2343. TRUE  high 
##  4 2016-01-06 17:23:00 2016-01-06     2 5-jdo-903    NA  3892. FALSE mid  
##  5 2016-01-09 12:36:00 2016-01-09     8 3-ldm-038     7   284. TRUE  low  
##  6 2016-01-11 06:15:00 2016-01-11     4 2-dhe-923     4  3291. TRUE  mid  
##  7 2016-01-15 18:46:00 2016-01-15     7 1-knw-093     3   843. TRUE  high 
##  8 2016-01-17 11:27:00 2016-01-17     4 5-boe-639     2  1036. FALSE low  
##  9 2016-01-20 04:30:00 2016-01-20     3 5-bce-642     9   838. FALSE high 
## 10 2016-01-20 04:30:00 2016-01-20     3 5-bce-642     9   838. FALSE high 
## 11 2016-01-26 20:07:00 2016-01-26     4 2-dmx-010     7   834. TRUE  low  
## 12 2016-01-28 02:51:00 2016-01-28     2 7-dmx-010     8   108. FALSE low  
## 13 2016-01-30 11:23:00 2016-01-30     1 3-dka-303    NA  2230. TRUE  high

Create an action_levels object with fractional values for the warn, stop, and notify states.

al <- 
  action_levels(
    warn_at = 0.2,
    stop_at = 0.8,
    notify_at = 0.5
  )

A summary of settings for the al object is shown by printing it.

al

## -- The `action_levels` settings
## WARN failure threshold of 0.2 of all test units.
## STOP failure threshold of 0.8 of all test units.
## NOTIFY failure threshold of 0.5 of all test units.
## ----

Create a pointblank agent and apply the al object to actions. Add two validation steps and interrogate the small_table.

agent_1 <-
  create_agent(
    tbl = small_table,
    actions = al
  ) %>%
  col_vals_gt(
    columns = vars(a), value = 2
  ) %>%
  col_vals_lt(
    columns = vars(d), value = 20000
  ) %>%
  interrogate()

The report from the agent will show that the warn state has been entered for the first validation step but not the second one. We can confirm this in the console by inspecting the warn component in the agent's x-list.

x_list <- get_agent_x_list(agent = agent_1)

x_list$warn

## [1]  TRUE FALSE

Applying the action_levels object to the agent means that all validation steps will inherit these settings but we can override this by applying another such object to the validation step instead (this time using the warn_on_fail() shorthand).

agent_2 <-
  create_agent(
    tbl = small_table,
    actions = al
  ) %>%
  col_vals_gt(
    columns = vars(a), value = 2,
    actions = warn_on_fail(warn_at = 0.5)
  ) %>%
  col_vals_lt(
    columns = vars(d), value = 20000
  ) %>%
  interrogate()

In this case, the first validation step has a less stringent failure threshold for the warn state and it's high enough that the condition is not entered. This can be confirmed in the console through inspection of the x-list warn component.

x_list <- get_agent_x_list(agent = agent_2)

x_list$warn

## [1] FALSE FALSE

In the context of using validation functions directly on data (i.e., no involvement of an agent) we want to trigger warnings and raise errors. The following will yield a warning if it is executed (returning the small_table data).

small_table %>%
  col_vals_gt(
    columns = vars(a), value = 2,
    actions = warn_on_fail(warn_at = 2)
  )

## Warning: Exceedance of failed test units where values in `a` should have been > `2`.
## The `col_vals_gt()` validation failed beyond the absolute threshold level (2).
## * failure level (4) >= failure threshold (2)

## # A tibble: 13 × 8
##    date_time           date           a b             c      d e     f    
##    <dttm>              <date>     <int> <chr>     <dbl>  <dbl> <lgl> <chr>
##  1 2016-01-04 11:00:00 2016-01-04     2 1-bcd-345     3  3423. TRUE  high 
##  2 2016-01-04 00:32:00 2016-01-04     3 5-egh-163     8 10000. TRUE  low  
##  3 2016-01-05 13:32:00 2016-01-05     6 8-kdg-938     3  2343. TRUE  high 
##  4 2016-01-06 17:23:00 2016-01-06     2 5-jdo-903    NA  3892. FALSE mid  
##  5 2016-01-09 12:36:00 2016-01-09     8 3-ldm-038     7   284. TRUE  low  
##  6 2016-01-11 06:15:00 2016-01-11     4 2-dhe-923     4  3291. TRUE  mid  
##  7 2016-01-15 18:46:00 2016-01-15     7 1-knw-093     3   843. TRUE  high 
##  8 2016-01-17 11:27:00 2016-01-17     4 5-boe-639     2  1036. FALSE low  
##  9 2016-01-20 04:30:00 2016-01-20     3 5-bce-642     9   838. FALSE high 
## 10 2016-01-20 04:30:00 2016-01-20     3 5-bce-642     9   838. FALSE high 
## 11 2016-01-26 20:07:00 2016-01-26     4 2-dmx-010     7   834. TRUE  low  
## 12 2016-01-28 02:51:00 2016-01-28     2 7-dmx-010     8   108. FALSE low  
## 13 2016-01-30 11:23:00 2016-01-30     1 3-dka-303    NA  2230. TRUE  high

With the same pipeline, not supplying anything for actions (it's NULL by default) will have the same effect as using stop_on_fail(stop_at = 1).

small_table %>%
  col_vals_gt(columns = vars(a), value = 2)

## Error: Exceedance of failed test units where values in `a` should have been > `2`.
## The `col_vals_gt()` validation failed beyond the absolute threshold level (1).
## * failure level (4) >= failure threshold (1)

small_table %>%
  col_vals_gt(
    columns = vars(a), value = 2,
    actions = stop_on_fail(stop_at = 1)
  )

## Error: Exceedance of failed test units where values in `a` should have been > `2`.
## The `col_vals_gt()` validation failed beyond the absolute threshold level (1).
## * failure level (4) >= failure threshold (1)

This is because the stop_on_fail() call is auto-injected in the default case (when operating on data) for your convenience. Behind the scenes a 'secret agent' uses 'covert actions': all so you can type less.

Function ID

1-5