Equivalent environments are seen as different when named

[moodymudskipper]

e1 <- new.env()
e2 <- new.env()
attr(e1, "foo") <- attr(e2, "foo") <- "foo_"
waldo::compare(e1, e2)
#> ✔ No differences

attr(e1, "name") <- attr(e2, "name") <- "name_"
waldo::compare(e1, e2)
#> `old` is <env:name_>
#> `new` is <env:name_>

^{Created on 2023-03-30 with reprex v2.0.2}

[hadley]

Named environments are deliberately compared by reference as of #133. I made that change as part of #127 (comment) — waldo compares generally compares environments by values but it compare named environments by reference because these are usually special environments (i.e. empty/base/global and packages environments).

Can you please explain where you saw this problem?

[moodymudskipper]

Sure, it did happen in artificial cases as you expected.

The package {constructive} prints idiomatic building code from objects, this includes environments with all the caveats that this entails. It uses Waldo to confirm that the generated code is faithful to the object.

# devtools::install_github("cynkra/constructive")
library(constructive)
e1 <- new.env()
e1$x <- 1
attr(e1, "name") <- "my_env"

# identical but uninteresting
construct(e1)
#> constructive::env("0x11bd168a0", parents = "global")

# not identical but equivalent, and interesting 
construct(e1, opts_environment(constructor = "list2env"), check = TRUE)
#> list2env(list(x = 1), parent = .GlobalEnv) |>
#>   structure(name = "my_env")
#> Error in `check_round_trip()`:
#> ! {constructive} couldn't create code that reproduces perfectly the input
#> `original` is <env:my_env>
#> `recreated` is <env:my_env>

^{Created on 2023-04-11 with reprex v2.0.2}

It's important for {constructive} not to choke even on rare cases because we'd like the package to be a robust tool to investigate corrupt or weird objects.

You mentioned already in #152 that you expect regular objects in waldo so this issue might be out of scope. In this case I might be able to special case this with a compare_proxy.environment method defined locally with rlang::local_bindings(.env = .GlobalEnv)

[hadley]

I'm going to close this issue because I think it is out of scope, and it sounds like you have a work around.