###############################################################################
# File: 
#    readarff.r
#
# Description:
#    Provides a function for reading ARFF files into the R statistical
#    package. Currently supports dense file formats without support
#    for timestamps.
#
# Author:
#    Craig A. Struble, Ph.D.
#
# Created:
#    Mon Aug 18 15:16:19 CDT 2003
#
# Last Modified:
#    Wed Sep 10 14:39:44 CDT 2003 - modified lastChar variable to attempt to
#        support S-PLUS.
###############################################################################

read.arff <- function (file, header = TRUE, sep = ",", quote="\"", dec=".",
                       fill = TRUE, ...)
{
    comment.char <- "%"                          # ARFF file comment char
    enum.open <- "{"
    enum.close <- "}"
    sep <- ","                                   # value separator

    ###########################################################################
    # Get the next token from the ARFF file
    # file - an open file or connection
    ###########################################################################
    nextArffToken <- function(file, lastChar) {
        aToken <- FALSE # something implying end of file
        if (lastChar == "") {
            aChar <- readChar(file, 1)
        } else {
            aChar <- lastChar
        }

        while (aToken == FALSE && aChar != "") {
            # skip whitespace
            while (regexpr("[[:space:]]", aChar) != -1 || aChar == sep) {
                aChar <- readChar(file, 1)
            }
            # Check for comment
            if (aChar == comment.char) {
                # skip until the end of the line
                readLines(file, n=1,ok=TRUE)
                # Get the next character
                aChar <- readChar(file,1)
            } else {
                aToken <- aChar
            }
        }

        if (aToken == "\"" || aToken == "'") { # Quoted string
            quote <- aToken
            aToken <- ""
            aChar <- readChar(file, 1)
            while(aChar != "" && aChar != quote) {
                aToken <- paste(aToken,aChar,sep="")
                if (aChar == "\\") { # Escaped character in string
                    aChar <- readChar(file, 1)
                    aToken <- paste(aToken,aChar,sep="")
                }
                aChar <- readChar(file, 1)
            }
            if (aChar == quote) {
                aChar <- readChar(file, 1)
            }
        } else { # Unquoted string
            aChar <- readChar(file, 1)
            # Until a brace, whitespace, comma, or EOF
            while (aChar != "" 
                   && regexpr("[[:space:]]", aChar) == -1
                   && !aChar %in% c(enum.open, enum.close, "\"", "'", sep)
                   && aToken != enum.open && aToken != enum.close)
            {
                aToken <- paste(aToken,aChar,sep="")
                aChar <- readChar(file, 1)
            }
        }

        return(list(token=aToken, lastChar=aChar))
    }

    # Check function parameters and open the file if necessary
    if(is.character(file)) {
        file <- file(file,"r")
        on.exit(close(file))
    }
    if (!inherits(file,"connection"))
        stop("argument `file' must be a character string or connection")

    if (!isOpen(file)) {
        open(file,"r")
        on.exit(close(file))
    }

    # Read in the header information
    # Read the @relation line, first token should be @relation,
    # otherwise an error
    res <- nextArffToken(file, "")
    if (tolower(res$token) != "@relation") {
        stop("@relation is the first token expected, got: ", aToken)
    }
    res <- nextArffToken(file, res$lastChar)
    relation <- res$token

    # Read in attributes
    res <- nextArffToken(file, res$lastChar)
    colClasses <- c()
    col.names <- c()
    while(tolower(res$token) == "@attribute") {
        res <- nextArffToken(file, res$lastChar)
        aName <- res$token
        col.names <- append(col.names, aName)
        res <- nextArffToken(file, res$lastChar)
        aType <- tolower(res$token)
        if (aType %in% c("numeric", "real", "integer")) {
            # The value is a numeric type
            colClasses <- append(colClasses, "numeric")
        } else if (aType %in% c("string")) {
            # The value is a character string type
            colClasses <- append(colClasses, "character")
        } else if (aType %in% c("date")) {
            stop("ARFF date type is not supported")
        } else if (aType == enum.open) {
            # nominal type read until enum.close
            colClasses <- append(colClasses, NA)
            res <- nextArffToken(file, res$lastChar)
            while (res$token != enum.close && res$token != FALSE) {
                res <- nextArffToken(file, res$lastChar)
            }
        } else {
            stop("invalid data type")
        }
        res <- nextArffToken(file, res$lastChar)
    }

    # Read the data
    if (tolower(res$token) != "@data") {
        stop("@data expected, got: ", aToken)
    }

    table <- read.table(file, na.strings="?", sep=",", 
                        comment.char=comment.char, col.names=col.names,
                        colClasses=colClasses)

    return(table)
}