Use of is_testing() is a problem for folks who program against pins functions

[juliasilge]

Interesting - as I was writing more of my unit tests, I noticed that the pin_write function behaves differently when it is run within or outside unit tests (of the testthat variety). That's because the time stamp part of the version is fixed in a unit test:

pins-r/R/pin_versions.R

Line 146 in 9e8c081

version_name <- function(metadata) {

I didn't expect that, and it took me a good while to figure out why my unit tests were failing. (I was writing the same file twice into a versioned board. My project includes tracking the metadata in a SQL database - and that complained about duplicated version keys.)

There is nothing intrinsically wrong with the current approach, but it reminds me a bit of the Volkswagen scandal 😄 and the behavior definitely was surprising. (I solved it by temporarily unsetting the TESTTHAT environmental variable in my test.)

Originally posted by @tomsing1 in #637 (comment)

[juliasilge]

The current use of is_testing(), for example, in version_name():

pins-r/R/pin_versions.R

Lines 147 to 148 in 9e8c081

	if (is_testing()) {
	paste0("20120304T050607Z-", substr(metadata$pin_hash, 1, 5))

makes it hard for folks to develop against pins functions and write their own tests.

[juliasilge]

The new tests for write_board_manifest() (specifically the ones in test-board_folder.R that look at the structure of the manifest file) involve versions and could be improved in terms of how easy they are to grasp. Currently, we use is_testing() to handle versions in tests but we don't want to continue doing that.

One option for some cases might be to use snapshot testing with a function to redact versions.

[juliasilge]

We could switch to use some mocking, like this:

library(pins)
b <- board_temp()

## pin_write() calls version_name() internally
b %>% pin_write(1:10, name = "nice-numbers", type = "json")
#> Creating new version '20230301T003852Z-c3943'
#> Writing to pin 'nice-numbers'

library(mockery)

# call to version_name() is 2 layers down from pin_store()
stub(
  pin_store, 
  "version_name", 
  "20120304T050607Z-xxxxx",
  depth = 2
  )

## now with stub:
b %>% pin_write(1:10, name = "nice-numbers", type = "json")
#> Replacing version '20230301T003852Z-c3943' with '20120304T050607Z-xxxxx'
#> Writing to pin 'nice-numbers'

^{Created on 2023-02-28 with reprex v2.0.2}

Notice that the new version name has been stubbed out.

[github-actions]

This issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with a reprex: https://reprex.tidyverse.org) and link to this issue.