Attempt to use namespaces for stan_lmer and survreg

[rempsyc]

Attempt to use namespaces for stan_lmer and survreg. As showed in easystats/datawizard#401, works in reprex locally and console, but not in tests, so testing here to see if it fixes itself.

[IndrajeetPatil]

I think you will also need to use GitHub version of datawizard for this to work.

[codecov-commenter]

Codecov Report

Merging #363 (6267d57) into main (4f8a9d1) will increase coverage by 6.40%.
The diff coverage is 87.01%.

❗ Current head 6267d57 differs from pull request most recent head eee8c2b. Consider uploading reports for the commit eee8c2b to get more accurate results

📣 This organization is not using Codecov’s GitHub App Integration. We recommend you install it so Codecov can continue to function properly for your repositories. Learn more

@@            Coverage Diff             @@
##             main     #363      +/-   ##
==========================================
+ Coverage   60.92%   67.32%   +6.40%     
==========================================
  Files          46       46              
  Lines        3173     3244      +71     
==========================================
+ Hits         1933     2184     +251     
+ Misses       1240     1060     -180     

Impacted Files	Coverage Δ
R/report_table.R	53.84% <ø> (ø)
R/report_sample.R	88.38% <87.01%> (-0.60%)	⬇️

... and 12 files with indirect coverage changes

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[rempsyc]

This test works locally:

x1 <- lm(Sepal.Length ~ Petal.Length * Species, data = iris)
testthat::expect_identical(
  as.character(report::report_performance(x1)),
  paste(
    "The model explains a statistically significant and substantial proportion of",
    "variance (R2 = 0.84, F(5, 144) = 151.71, p < .001, adj. R2 = 0.83)"
  )
)

^{Created on 2023-04-13 with reprex v2.0.2}

But not using the random order test script in the console:

> purrr::walk(randomized_test_script_paths, test_path)
✔ All tests passing in `tests/testthat/test-report.data.frame.R`.
✔ All tests passing in `tests/testthat/test-report.lm.R`.
✔ All tests passing in `tests/testthat/test-format_model.R`.
✖ There was error while running tests in `tests/testthat/test-report_performance.R`.
Test `report_performance` has error:
expectation_success: as.character(report_performance(x1)) (`actual`) not identical to paste(...) (`expected`).

Even after renaming from x to x1 (there is no other object called x1). The tests are also fully contained within a testthat call. So no idea what the issue is?

[rempsyc]

I am not able to update snapshots for file test-report_performance.R. When running the tests locally in console with datawizard version ‘0.7.1.2’ I get no errors. But when using testthat::test_file("tests/testthat/test-report_performance.R"), I still get the following:

> testthat::test_file("tests/testthat/test-report_performance.R")
[ FAIL 2 | WARN 0 | SKIP 0 | PASS 12 ]

── Error (test-report_performance.R:109): report_performance Bayesian) ─────────
`report_performance(x7)` threw an unexpected error.
Message: Unable to refit the model with standardized data.
  Try instead to standardize the data (standardize(data)) and refit the
  model manually.
Class:   simpleError/error/condition

── Error (test-report_performance.R:113): report_performance Bayesian) ─────────
`summary(report_performance(x7))` threw an unexpected error.
Message: Unable to refit the model with standardized data.
  Try instead to standardize the data (standardize(data)) and refit the
  model manually.
Class:   simpleError/error/condition

[ FAIL 2 | WARN 0 | SKIP 0 | PASS 12 ]

Same behaviour when using RStudio's "Run Tests" or "Test" buttons.

@etiennebacher any idea? Namespace conflicts is your area of expertise now no? :P

[rempsyc]

Actually we can see the Unable to refit the model with standardized data. Try instead to standardize the data standardize(data)) and refit the model manually error in R CMD check here: https://github.com/easystats/report/actions/runs/4695121238/jobs/8323971085?pr=363#step:9:172

[etiennebacher]

It's hard to explain but I think when rstanarm::stan_lmer is run interactively, it finds stan_glmer <- rstanarm::stan_glmer in the environment but when it runs in a new env generated by testthat it doesn't find it anymore.

One solution is to use stan_glmer <<- rstanarm::stan_glmer instead so that it attaches it to the global env. To avoid side effects you can add remove(stan_glmer, envir = .GlobalEnv) just before unloadNamespace("rstanarm")

[rempsyc]

Interesting! I don't fully get it but it seems to work, thanks! :D

[IndrajeetPatil]

@rempsyc Can you please also have a look at the test failure in random order? Also, test coverage workflow is failing.

[rempsyc]

All tests are finally passing, yeah! :-)

[IndrajeetPatil]

Awesome work, Remi! Thanks 😊

[rempsyc]

@etiennebacher the stan_glmer <<- rstanarm::stan_glmer workaround was working when we merged this, but now this is throwing this error both locally and on GAs:

Error in `eval(code, test_env)`: cannot change value of locked binding for 'stan_glmer'

A brief online search seems to confirm that this error is generated when using super assignment. In fact, it seems that this error arises when rstanarm is loaded first.

suppressPackageStartupMessages(library(rstanarm))

stan_glmer <<- rstanarm::stan_glmer
#> Error in eval(expr, envir, enclos): cannot change value of locked binding for 'stan_glmer'

^{Created on 2023-05-07 with reprex v2.0.2}

Although we don't load it explicitly, sometimes, but not all the time (because this is rstanarm.....), it will load by itself just through using namespace.

Any idea for the best way to fix this? I also opened an issue at stan-dev/rstanarm#589

[etiennebacher]

[I posted this by mistake in the rstanarm issue, so I deleted it there and put it here]

This is very well explained in this SO answer. Paraphrasing what is said there, the problem is that there's already a stan_glmer() loaded by rstanarm. Our <<- assignment found that, tried to change its value, and refused (because the function stan_glmer() is "locked", i.e. immutable).

I think one solution would be to check whether stan_glmer() is available and if it's not, use <<-. In the last test of test-report_performance, you could replace the <<- by

is_stan_glmer_avail <- !inherits(try(stan_glmer, silent = TRUE), "try-error")
if (!is_stan_glmer_avail) {
  stan_glmer <<- rstanarm::stan_glmer
  on.exit(remove(stan_glmer, envir = .GlobalEnv))
}

(don't forget to remove the remove() line at the end of the test).

Can you check if this works? It works for me locally but I couldn't reproduce the error when running devtools::test() in the first hand so who knows

[rempsyc]

Thanks @etiennebacher that seems to have solved the issue!! :)