eb-html-min-KQNEWEJ6.wok

#!/usr/bin/wok -f

/** eb-html-min.wok
  * 
  * Usage:
  *   java -jar wok-0.1.0.jar -f eb-html-min.wok -v src=0 -v dst=1 -v update=true -v@rawstr dic=path_to_epwing_dir -v@rawstr ebmap=path_to_ebmap_file -v@rawstr dir=path_to_media_dir input > output
  * 
  * Option:
  *   -v src:           The index number of the source column.
  *   -v dst:           The index number of the destination column.
  *   -v@rawstr dic:    Path to a directory containing Epwing's CATALOGS file.
  *   -v@rawstr ebmap:  Path to a EBWin3/4 map file corresponding to the EPWING dictionary specified by `dic`.
  *   -v@rawstr media:  Path to a directory to be used as the output destination of image files.
  *   -v update:        Whether or not update all cells of the column specified by `dst`.
  *
  * Note: This script has a dependency for a funcion to decode BASE64, only exists in Sun's Java Runtime Environment.
  */

FS = '\t'
FQ = Quote.Min
OFS = '\t'
OFQ = Quote.All

val eb = "ebquery-0.3.1.jar"
val tag = """<img alt="([0-9a-zA-Z]+)" class="ebec" src="data:image/png;base64,([^"]+)">""".r
val b64 = new sun.misc.BASE64Decoder()
val md = java.security.MessageDigest.getInstance("md5")

if (eb nonExistent) 
  eb write Resource.fromURL("https://bitbucket.org/rubyu/ebquery/downloads/ebquery-0.3.1.jar").bytes

if (media nonExistent)
  media.createDirectory()

def query(s: String) = {
  val html = Seq("java", "-Dfile.encoding=UTF-8", "-jar", eb, "-d", dic, "-f", "html", "-m", "en,kw,in,em,dc,tx,sb,sp,ec,ls", "--ebmap", ebmap, s).!>.string
  tag.replaceAllIn(html, { m =>
    val (alt, data) = (m.group(1), m.group(2))
    val bytes = b64 decodeBuffer data
    md.reset()
    md update bytes
    val hash = md.digest().map("%02X" format _).mkString
    val file = media / s"$hash.png"
    if (file nonExistent)
      file write bytes
    s"""<img alt="$alt" class="ebec" src="${file.name}">"""
  })
}

In { _ 
  .filter (_ isDefinedAt src)
  .map (_.padTo(dst+1, ""))
  .map (row => if (update || row(dst).isEmpty) row.updated(dst, query(row(src))) else row)
  .foreach (row => println(row: _*))
}