Alternative solution for nflverse#389:

Start with the latest season. This season has column "weather" filled. So BigQuery will choose the right column type STRING.

R/helper_database_functions.R

@@ -137,7 +137,8 @@ build_db <- function(tblname = "nflfastR_pbp", db_conn, rebuild = FALSE, show_me
     dplyr::filter(.data$season >= 1999 & !is.na(.data$result)) %>%
     dplyr::group_by(.data$season) %>%
     dplyr::summarise() %>%
-    dplyr::ungroup()
+    dplyr::ungroup() %>%
+    dplyr::arrange(dplyr::desc(season))
 
   if (all(rebuild == TRUE)) {
     cli::cli_ul("{my_time()} | Purging the complete data table {.val {tblname}}

R/utils.R

@@ -173,7 +173,6 @@ write_pbp <- function(seasons, dbConnection, tablename){
   purrr::walk(seasons, function(x, p){
     pbp <- nflreadr::load_pbp(x)
     if (!DBI::dbExistsTable(dbConnection, tablename)){
-      default_play$weather <- "some other string"
       pbp <- dplyr::bind_rows(default_play, pbp)
     }
     if (DBI::dbExistsTable(dbConnection, tablename)) {