Use of format() in removeFormulaTerms() produces warning when long formulas are used

[amanda-hi]

I'm receiving warnings when providing a formula >60 characters (give or take) to predict.psm(). These warnings seem to be stemming from removeFormulaTerms(), specifically the use of format() that was added in the rms 6.8.1 release. When long formulas are provided to predict(), format() breaks them apart into vectors of length >1. This throws the following warning:

Using formula(x) is deprecated when x is a character vector of length > 1.
  Consider formula(paste(x, collapse = " ")) instead.

See below for a reprex.

reprex::reprex({
  # Setup --------------------------------------------------
  suppressPackageStartupMessages({
    library(rms) # Must be version 6.8.1 or greater
    library(withr)
  })

  n <- 1000
  withr::with_seed(17, {
    df <- data.frame(age  = rnorm(n, 50, 10),
                     blood.pressure = rnorm(n, 120, 15),
                     cholesterol    = rnorm(n, 200, 25),
                     sex            = factor(sample(c('female','male'), n, TRUE)),
                     treat          = factor(sample(c('a','b','c'), n, TRUE))
    )
    df$L <- .4*(df$sex=='male') + .045*(df$age-50) +
      (log(df$cholesterol - 10)-5.2)*(-2*(df$sex=='female') + 2*(df$sex=='male')) +
      .3*sqrt(df$blood.pressure-60)-2.3 + 1*(df$treat=='b')
    df$y <- ifelse(runif(n) < plogis(df$L), 1, 0)
    df$h <- .02*exp(.06*(df$age-50)+.8*(df$sex=='Female'))
    df$d.time <- -log(runif(n))/df$h
    df$cens <- 15*runif(n)
    df$death <- ifelse(df$d.time <= df$cens,1,0)
    df$d.time <- pmin(df$d.time, df$cens)
    
    feats <- replicate(5, stats::rnorm(n, mean = 2500, sd = 500))
    colnames(feats) <- paste0("feature", 1:5)
    df <- cbind(df, feats)
  })
  
  # Testing --------------------------------------------
  form_short   <- Surv(d.time, death) ~ feature1 + age
  form_long    <- Surv(d.time, death) ~ feature1 + feature2 + feature3 + feature4 + age
  model_short  <- rms::psm(form_short, df)
  model_long   <- rms::psm(form_long, df)
  
  invisible(predict(model_short, type = "lp"))
  invisible(predict(model_long, type = "lp")) # Throws warning
  
  invisible(
    rms:::removeFormulaTerms(model_short$sformula,
                                  which = "offset", 
                                  delete.response = TRUE))
  
  invisible(
    rms:::removeFormulaTerms(model_long$sformula, 
                                  which = "offset", 
                                  delete.response = TRUE)) # Throws warning
  
  # Dissecting relevant section of the updated removeFormulaTerms()
  # https://github.com/harrelfe/rms/blob/ef8982e237fbbf561fd5b7fe675e786dd1118549/R/rmsMisc.s#L1396
  which <- "offset"
  if('offset' %in% which) {
    test_form_long <- format(model_long$sformula)
    print(length(test_form_long))
    which[which == 'offset'] <- '.off.'
    mod_form_long <- gsub('offset(', '.off.(', test_form_long, fixed=TRUE)
    form <- as.formula(mod_form_long)
  }

  format(model_short$sformula)
  format(model_long$sformula)
})

[amanda-hi]

I think this is the same problem that's being reported in #148