Support survdiff() on a survfit object

[DanChaltiel]

Dear Pr Therneau,

In my analyses, I almost never calculate a survdiff test without plotting the KM curves.
This results in the formula being written twice in my code, which is tedious and error-prone in case I want to modify it.

Here is a small improvement suggestion where you can use the survfit object directly instead:

library(survival)

survdiff(Surv(time, status) ~ inst, data=lung, rho=.5)$pvalue
#> [1] 0.5903835

km <- survfit(Surv(time, status) ~ inst, lung)
survdiff(km, rho=.5)$pvalue
#> [1] 0.5903835

^{Created on 2025-07-06 with reprex v2.1.1}

Many thanks once more for your work on this exceptionally useful package.

Best regards
Dan