R/GBAT.process_addresses.R

In gmculp/rGBAT16AB: GeoSupport for R

#' The \code{GBAT.process_addresses} function is a more efficient way to geocode a large data frame of NYC addresses where there are duplicates.  It aggregates, geocodes and then merges the geocoded results back to the original data frame.

#function to aggregate addresses, geocode them
#and merge the results back to the original dataset
#this is a more efficient way to handle large datasets with a lot
#of duplicate addresses
GBAT.process_addresses <- function(in_df, add_col_name, third_col_name, source_cols, geocode_cols, add_type="boro_code", return_type="all") {

	ret_cols <- c(source_cols, geocode_cols)

	#aggregate duplicate addresses
	df_unq <- unique(in_df[,c(add_col_name,third_col_name)])

	#add unique id
	df_unq$u_id <- 1:nrow(df_unq)

	#record process start time
	p1 <- proc.time()

	#call geocoder based on address type
	if(grepl("^Z.*$",add_type, ignore.case = TRUE)){
		GC_df <- GBAT.zip_code(df_unq,"u_id",add_col_name,third_col_name)
	} else {
		GC_df <- GBAT.borough_code(df_unq,"u_id",add_col_name,third_col_name)
	}

	#record process end time
	p2 <- proc.time() - p1

	#processing statistics (records per second)
	cat(paste0(round((nrow(GC_df)/as.numeric(p2[3])), digits = 0)," records per second", "\n"))

	#in order to lighten load on merge, discard extraneous columns
	keep_col <- unique(c(c(add_col_name,third_col_name),geocode_cols,"F1E_GRC"))
	GC_df <- GC_df[,keep_col]

	#join back to original
	output_df <- merge(in_df, GC_df, by=c(add_col_name,third_col_name))

	#extracted successfully geocoded records
	success_df <- unique(output_df[output_df[["F1E_GRC"]]=="00" | output_df[["F1E_GRC"]]=="01",ret_cols])

	#percentage of successfully geocoded records
	cat(paste0(round((nrow(success_df)/nrow(in_df))*100, digits = 0),"% records successfully geocoded"))

	#option to return only successfully geocoded records
	if (return_type != "all") {
		return(success_df[,ret_cols])
	} else {
		return(output_df[,ret_cols])
	}
}