pointblank: Was I expecting this?

rainbowR 2026

Hannah Frick, Posit

• here to talk about expectations - on your data
• most of us don’t collect, but get access to it - and then build things with it, a dashboard or a prediction model
• sometimes, things change…
• if a column gets renamed, things break
• if a variable gets recorded in a different unit, things might now silently be garbage

• wouldn’t it be nice to get a heads-up?
• an email that says “these things are changing” rather than “the dashboard is broken”
• get a colleague to email you
• alternatively: run regular validation checks and have a system email you

• that’s what pointblank is for
• feature-rich, loads of documentation
• my goal: give you a sense of what’s possible and how to get started

• core output: html table
• stick it in a report, or have it emailed to you
• one row = one validation rule
• specification on the left, results on the right
• units: e.g. if you’re validating a column, the units are the rows
• pass/fail: how many test units passed/failed
• life is messy, your data might be too but in both cases, you typically can tolerate some amount of that
• specify your thresholds for when to warn, stop, and notify you
• minimum: it drives the color on the left but can trigger any action that you can express in R code!
• last column, for when things went wrong: access a csv with the units/rows that failed the condition

Core code pattern

library(pointblank)

# create
create_agent(table_to_validate) |>
  # specify
  col_vals_gt(column_d, value = 200) |>
  # ...
  # apply
  interrogate()

• core principle: sth you might have seen in httr(2) or recipes
• create an object, add specifications to it, and then apply those specifications to the data
• data: local or remote
• validation rules: many pre-made

Thresholds and actions

agent <- create_agent(
  tbl = table_to_validate,
  actions = action_levels(
    warn_at = 2, # absolute threshold
    stop_at = 0.2, # relative threshold
    notify_at = 0.3,
    fns = list(warn = ~ warning("Data (life?) is getting messy."))
  )
) |>
  col_vals_gt(column_d, value = 200) |>
  interrogate()

Trigger an email notification

create_agent(
  small_table,
  actions = action_levels(notify_at = 0.3),
  end_fns = list(
    ~ email_blast(
      x,
      to = "hello@rainbowr.org",
      from = "pb_notify@example.com",
      msg_subject = "I was not expecting this!",
      credentials = blastula::creds_key(id = "smtp2go"),
      send_condition = ~ TRUE %in% x$notify
    )
  )
)

Program against the results

• Validation results: get_agent_x_list()
• Data that failed any rule: get_sundered_data()
• Data that failed a specific rule: get_data_extracts(i = 1)

https://rstudio.github.io/pointblank/