duration by bin

tidy.r

@@ -9,13 +9,42 @@ load <- function() {
 
 # data.frame -> data.frame
 set.types <- function(df) {
-  df$ts <- as.POSIXct(df$ts, origin = '1970-01-01')
+  df$date <- as.POSIXct(df$ts, origin = '1970-01-01')
+  df$ts <- NULL
   df$uid <- as.factor(df$uid)
   df$nick <- as.factor(df$nick)
   df$status <- factor(df$status, levels = c('avail', 'notavail'))
   df
 }
 
+# Total duration online based on the date column
+# If you want to group by uid or date, do that first.
+# data.frame -> data.frame
+duration <- function(df, start.date, end.date) {
+  # Skip empty data frames.
+  if (nrow(df) == 0) {
+    return(0)
+  }
+
+  # Deal with sessions spanning date cuts
+  if (df$status[1] == 'notavail') {
+    head.duration <- df$date[1] - start.date
+    df <- df[-1,]
+  } else {
+    head.duration <- 0
+  }
+  if (df$status[nrow(df)] == 'avail') {
+    tail.duration <- end.date - df$date[nrow(df)]
+    df <- df[-nrow(df),]
+  } else {
+    tail.duration <- 0
+  }
+
+  # Now we can assume that the data frame starts with 'avail', ends with 'notavail' and alternates.
+  avail <- ((1:nrow(df)) %% 2) == 1
+  cumsum(df[-avail,'date'] - df[avail,'date']) + head.duration + tail.duration
+}
+
 # IO ()
 main <- function () {
   df <- load()