include check_column_has_no_na in r2dii.utils

[jdhoffa]

I find myself using the following function frequently in r2dii.analysis to ensure that no crucial input columns contain any NA's:

check_column_has_no_na <- function(data, column) {
  if (anyNA(data[[column]])) {
    rlang::abort(
      class = "column_has_na",
      glue::glue("Column `{column}` must not contain any `NA`s.")
    )
  }

  invisible(data)
}

[maurolepore]

Before working on this I would like to learn more about existing packages for assertions and discuss when to use them versus when to write custom assertions. It seems a great topic for a ds-incubator meetup (2DegreesInvesting/ds-incubator#11 (comment)), and maybe @vintented and @looreen might help lead it?

[cjyetman]

I'm interested in this discussion, if it's still on the table. I've done a ton of this lately (custom assertions and error reporting).

[maurolepore]

Absolutely. You've been thinking a lot about 2DII's data and it'd be great to advertise your work and similar (e.g. https://cran.r-project.org/web/packages/assertr/vignettes/assertr.html).

What would be your preferred way to revive this conversation and engage our colleagues?

[cjyetman]

At the moment, I'd be more interested in a discussion with some at least minimally informed people about this topic. I think by now I'm particularly aware of the main options available, but in my use case none of them met all of my needs. I'd be interested in hearing about other available options and/or other needs that other people have had that were not met by current options. I'm also happy to discuss my self-made implementation of this, what it does, why I needed it, and possibly how it could be generalized to meet other needs.... but also would be happy to find an alternative that someone else maintains.

In the long run I'd be happy to do/help with a presentation of learnings/decisions about this topic, but at the moment I feel like I'm more at the discussion phase. I did do a DS-incubator early last summer about using assertr, but I probably know a lot more about this by now. 😉

[maurolepore]

Good. How about starting with a call with whoever you think might help at this stage? I thought of starting a conversation in the "Discussions" tab of some repo -- but what repo should that be?

[cjyetman]

Let's have a call and/or start a Slack channel early in the new year to start.