data_dir <- "data"

download_bl_data <- function(directory) {
	csv_urls = c(
		"http://www.barrolee.com/data/BL_v1.2/BL(2010)_MF2599_v1.2.csv",
		"http://www.barrolee.com/data/BL_v1.2/BL(2010)_MF1599_v1.2.csv"
	)
	for (url in csv_urls) {
		download.file(url, paste(directory, basename(url), sep="/"), method="auto")
	}
}

import_bl15_data <- function(directory) {
	read.csv(paste(directory, "BL(2010)_MF1599_v1.2.csv", sep="/"))
}

import_bl25_data <- function(directory) {
	read.csv(paste(directory, "BL(2010)_MF2599_v1.2.csv", sep="/"))
}

# Fill in missing values by linear interpolation
interp_nas <- function(x) {
	pos=which(!is.na(x))
	if (length(pos) < 2) {
		warning("Not enough data to interpolate.")
		return(x)
	}
	for (i in 1:(length(pos) - 1)) {
		x[pos[i]:pos[i+1]] <- approx(
			c(x[pos[i]], x[pos[i+1]]),
			n=(pos[i+1] - pos[i] + 1)
		)$y
	}
	return(x)
}

# Adds the specified column from edudata to pwtdata and fills in missing values
add_education_column <- function(pwtdata, edudata, colname="yr_sch") {
	# subset to year, country code, and the column we want to merge
	edudata <- subset(edudata, select = c("year", colname, "WBcode"))

	# merge data frames
	combined <- merge(
		pwtdata,
		edudata,
		by.x = c("year", "isocode"),
		by.y = c("year", "WBcode"),
		all.x = TRUE # include all years in the result
	)

	# interpolate missing values
	combined <- do.call(
		rbind,
		lapply(
			split(combined, combined$isocode),
			function(cdata) {
				print(cdata$yr_sch)
				cdata$yr_sch <- interp_nas(cdata$yr_sch)
				cdata
			}
		)
	)
	return(combined)
}