Bug/tidyselect not working

[Moohan]

This partially resolves #84 I'm pretty sure what I have is an improvement (with no regressions) but it's not a complete fix.

I don't have time to work on this anymore and to be honest, all the non-standard evaluation stuff is outside my comfort zone! It might be worth getting in touch with Nic - [email protected] - as he did the work on Public-Health-Scotland/phslookups#9 which is a simpler version of this.

I added tests which you'll probably want to expand, the only one that doesn't work is:

read_slf_episode(
      "1920",
      col_select = c("year", dplyr::contains("dd")),
      recids = "DD"
    )

i.e. when a user uses tidyselect helpers and uses the filter but (probably) didn't include the required filter variable. In the above example, it's filtering by recid but recid isn't in col_select.

[Jennit07]

Thanks @Moohan appreciate your help with this. Zihao is going to pick this up and look into developing it further.

[lizihao-anu]

Thank you for the work, @Moohan!

The current approach looks great but it fails in adding "recid" or "hscd2018" to columns to select. Honestly, I could not find a way to make tidy selection work outside a selecting function. The error looks like below.

Error:
! `ends_with()` must be used within a *selecting* function.
ℹ See ?tidyselect::faq-selection-context for details.

The workaround is to use read_parquet once more for recid or partnerships. Given the super-fast speed of parquet, although it doubles the read_slf execution time, it does not really matter. The benchmark shows that the running time increases from 250ms-ish to 500ms-ish.

Are you happy with the workaround and merging the pull request? @Moohan Thanks!

[Moohan]

Nice work, I think this solution is fine 👍 I can't see a way to clean up the code anymore than you have it already, and the final case is probably pretty niche (filtering without asking for the relevant filter variable), so probably fine if it's slightly slower in that case.