Nothing
#' ParseOvertime
#'
#' Internal Function - Parses an overtime report returned from the API
#'
#' @param report.data jsonlite formatted data frame of report data returned from the API
#'
#' @importFrom plyr ldply
#'
#' @return Formatted data frame
#'
#' @family internal
#' @keywords internal
ParseOvertime <- function(report.data) {
# jsonlite already makes this into a nice data frame for us
data <- report.data$report$data
# Create a column of R datetimes
if('hour' %in% colnames(data)){
datetime <- strptime(paste(data$year,data$month,data$day,data$hour,sep="-"), "%Y-%m-%d-%H")
} else {
datetime <- strptime(paste(data$year,data$month,data$day,sep="-"), "%Y-%m-%d")
}
counts.df <- ldply(data$counts)
metrics <- report.data$report$metrics$id
names(counts.df) <- metrics #assign names to counts.df
drops <- c("counts")
rows.df <- data[,!(names(data) %in% drops)]
# check if we have anomaly detection
if("forecasts" %in% colnames(data)) {
forecasts.df <- ldply(data$forecasts)
names(forecasts.df) <- paste("forecast.",metrics,sep="")
counts.df <- cbind(counts.df,forecasts.df)
drops <- c("forecasts")
rows.df <- rows.df[,!(names(rows.df) %in% drops)]
}
if("upperBounds" %in% colnames(data)) {
upperBounds.df <- ldply(data$upperBounds)
names(upperBounds.df) <- paste("upperBound.",metrics,sep="")
counts.df <- cbind(counts.df,upperBounds.df)
drops <- c("upperBounds")
rows.df <- rows.df[,!(names(rows.df) %in% drops)]
}
if("lowerBounds" %in% colnames(data)) {
lowerBounds.df <- ldply(data$lowerBounds)
names(lowerBounds.df) <- paste("lowerBound.",metrics,sep="")
counts.df <- cbind(counts.df,lowerBounds.df)
drops <- c("lowerBounds")
rows.df <- rows.df[,!(names(rows.df) %in% drops)]
}
# convert all count columns to numeric
for(i in 1:ncol(counts.df)) {
counts.df[,i] <- as.numeric(counts.df[,i])
}
#Get segment
seg <- report.data$report$segments
#If segment null, make a dummy data frame
if(is.null(seg)){
seg <- data.frame(list("", ""))
names(seg) <- c("segment.id", "segment.name")
}
#If segment has values, concatenate all values with "AND". R puts the
#concatenated values in every single row, so I dedupe the dataframe
else{
names(seg) <- c("segment.id", "segment.name")
seg$segment.name<-(paste(as.list(seg$segment.name),collapse=" AND "))
seg$segment.id<-(paste(as.list(seg$segment.id),collapse=" AND "))
seg<-subset(seg,!duplicated(seg$segment.name))}
#Put segment after dates
formatted.df <- cbind(datetime,rows.df, seg, counts.df, row.names = NULL)
return(formatted.df)
}
Any scripts or data that you put into this service are public.
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.