The base::rle function might offer an alternative approach that does not require string manipulation gymnastics. The function below finds the longest stretch of 0s in a vector, x, but could be modified to find values above/below a threshold.

# Create a function that will find maximum length of specific values
rle_max_lengths <- function(x, value = 0) {
  # Do run length calculations
  run_length <- rle(x)
  # Find which of the runs are the value of interest
  is_value <- run_length$values == value
  max_length <- 0
  # Only try extracting values if there was at least one
  if (any(is_value)) {
    # Pull out all lengths of value of interest
    value_lengths <- run_length$lengths[is_value]
    # Find the maximum length of all runs of value of interest
    max_length <- max(value_lengths)
  }
  return(max_length)
}

In the returned data frame / tibble, the values that are calculated for a site over all seasons should be named in a consistent manner. One way to do this would be to ensure the string period is in the corresponding column name.

Clarify that the caveats at the very end of the vignette only apply to the slow, serial approach. Perhaps best to add code that will convert wide-formatted output to long format following calculations of summary statistics.

When period = "mid", and all days are below rain_cutoff, dry_interval returns 0 (according to documentation). Should this instead return the actual number of days, since it was, technically, the longest stretch of dry days?

• instructions for installing vignette aren't complete
• include a minimal example

Brief explanation of how "seasons" are determined from data, especially in cases where the season includes the new year (i.e. start_month > end_month). Use language in the documentation for season_year(), which is in enumerate_seasons.R.