Boottest looking for the data in the wrong place when running inside a function

[Yuval-OS]

See this reproducible code:
It works without a function, but when used inside a function it says it cannot find data_obj.
When I define data_obj in the global directory, it works fine.

requireNamespace("fwildclusterboot")
data(voters)

feols_fit <- feols(proposition_vote ~ treatment + ideology1 + log_income | Q1_immigration, data = voters)
boot1 <- boottest(feols_fit, B = 9999, param = "treatment", clustid = "group_id1")
summary(boot1)

boot_func <- function(data_obj) {
feols_fit1 <- feols(proposition_vote ~ treatment + ideology1 + log_income | Q1_immigration, data = data_obj)
boot1 <- boottest(feols_fit1,
B = 9999,
param = "treatment",
clustid = "group_id1")
return(summary(boot1))
}

boot_func(voters)
data_obj <- voters
boot_func(voters)

[ecrjak]

In principle, this should be solved by the new fixest option data.save. However, this still needs to be implemented in boottest, it seems. The following code still throws an error: Error in eval(call("model.frame", ff, data = cl$data, subset = cl$subset, : object 'DATA_SAVED' not found

boot_func <- function(data_obj) {
  feols_fit1 <- feols(proposition_vote ~ treatment + ideology1 + log_income | Q1_immigration, data = data_obj,
                      data.save = TRUE)
  boot1 <- boottest(feols_fit1,
                    B = 9999,
                    param = "treatment",
                    clustid = "group_id1")
  return(summary(boot1))
}

boot_func(voters)

[ecrjak]

Okay, it's a bit hacky but for me it seems to work if you explicitly define DATA_SAVED in the loop:

boot_func <- function(data_obj) {
  feols_fit1 <- feols(proposition_vote ~ treatment + ideology1 + log_income | Q1_immigration, data = data_obj,
                      data.save = TRUE)
  DATA_SAVED <- feols_fit1$data
  boot1 <- boottest(feols_fit1,
                    B = 9999,
                    param = "treatment",
                    clustid = "group_id1")
  return(summary(boot1))
}

boot_func(voters)

Note that you will also need to set the same seeds to ensure results are identical to the non-loop version.