Correctly setting zero voltages to NaN in DB cleaning

BDInterface/data_cleaning_functions.R

@@ -200,6 +200,7 @@ voltages_in_bounds <- function(v) {
 remove_outlying_voltages <- function(time_series) {
   voltage_cols <- c("v", "vmin", "vmax", "vmean")
   time_series[, voltage_cols] <- sapply(time_series[, voltage_cols], as.numeric)
+  old_time_series <- time_series
 
   voltage_extremes <- time_series %>%
     select(c("c_id"), voltage_cols) %>%
@@ -213,5 +214,8 @@ remove_outlying_voltages <- function(time_series) {
     time_series[, voltage_cols] *
     voltage_extremes[match(time_series$c_id, voltage_extremes$c_id), voltage_cols]
   )
+  time_series['v_changed'] <- rowSums(
+    is.na(time_series[, voltage_cols]) & !is.na(old_time_series[, voltage_cols])
+  ) > 0
   return(time_series)
 }

BDInterface/interface.R

@@ -428,9 +428,13 @@ DBInterface <- R6::R6Class("DBInterface",
 
       while (length(circuits$c_id) > 0){
         time_series <- self$get_time_series_data_by_c_id_full_row(circuits)
+        time_series <- remove_outlying_voltages(time_series)
+        records_to_update <- filter(time_series, v_changed)
+        records_to_update <- within(time_series, rm(v_changed))
+        self$update_timeseries_table_in_database(records_to_update)
+
         time_series <- mutate(time_series, d = as.numeric(d))
         time_series <- mutate(time_series, time = fastPOSIXct(ts, tz="Australia/Brisbane"))
-        time_series <- remove_outlying_voltages(time_series)
         time_series_5s_and_unknown_durations <- filter(time_series, (!d %in% c(30, 60)))
         time_series_30s_and_60s_data <- filter(time_series, (d %in% c(30, 60)))
         time_series_5s_and_unknown_durations <- self$clean_duration_values(time_series_5s_and_unknown_durations)