ard_survival_survfit fails when stratification variable level includes a "="

[dchiu911]

What happened?

In L384 we see that the pivoting of the survfit result on the strata levels matches all "=" delimiters, but sometimes, we include "=" itself in the strata levels, e.g. comparing age "<60" vs. ">=60". The reprex below demonstrates the error thrown.

library(survival)
library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
library(cardx)
lung2 <- lung %>% 
  mutate(age_bin = factor(ifelse(age < 60, "<60", ">=60")))
survfit(Surv(time, status) ~ age_bin, data = lung2) %>% 
  ard_survival_survfit(times = 100)
#> Error in `tidyr::separate_wider_delim()`:
#> ! Expected 2 pieces in each element of `strata`.
#> ! 5 values were too long.
#> ℹ Use `too_many = "debug"` to diagnose the problem.
#> ℹ Use `too_many = "drop"/"merge"` to silence this message.

^{Created on 2025-01-06 with reprex v2.1.1}

A more conservative approach to splitting the stratification variable name and its levels could be to separate by the first "=" sign:

ret %>% tidyr::separate_wider_regex("strata",
                                    patterns = c(
                                      group1 = ".*",
                                      "(?<=[[:alnum:]])\\=(?=.*)",
                                      group1_level = ".*"
                                    ))

There's probably better ways to capture the regex.

[ddsjoberg]

Thanks @dchiu911 for the report! I can't recall the exact details, but I thought we were using stats::terms() to parse the variable names. @edelarua when you have a moment, can you investigate?