/src/main/scala/com/minzc/triple/exp/StrTools.scala

package com.minzc.triple.exp

import org.apache.commons.cli.{Options, CommandLine}
import org.apache.hadoop.conf.Configuration
import _root_.util.{AdvFile, TripleConsts}
import scala.io.Source
import java.util
import com.yeezhao.commons.util.{CliRunner, StringUtil, FreqDist, AdvCli}
import scala.collection.JavaConversions._
import java.io.{FileInputStream, FileWriter}
import org.ansj.recognition.NatureRecognition
import org.ansj.splitWord.analysis.ToAnalysis

/**
 * @author congzicun
 * @since 17/6/14 11:56 AM
 */
object StrToolsRunner {
  def main(args: Array[String]) = {
    AdvCli.initRunner(args, "StringTools", new StrTools())
  }
}

class StrTools extends CliRunner {
  val PARAM_FUNC = "func"
  val SEGMENT = "seg"
  val SEGANSJ = "ansj"
  val conf = new Configuration()
  conf.addResource(TripleConsts.CONFIG_FILE)

  def start(cmdLine: CommandLine) = {
    cmdLine.getOptionValue(PARAM_FUNC) match {
      case SEGMENT => segFile(cmdLine.getOptionValue(AdvCli.CLI_PARAM_I), cmdLine.getOptionValue(AdvCli.CLI_PARAM_O))
      case SEGANSJ => ansjSegFile(cmdLine.getOptionValue(AdvCli.CLI_PARAM_I), cmdLine.getOptionValue(AdvCli.CLI_PARAM_O))
    }
  }

  def initOptions(): Options = {
    val options = new Options()
    options.addOption(AdvCli.CLI_PARAM_I, true, "input file")
    options.addOption(AdvCli.CLI_PARAM_O, true, "output file")
    options.addOption(PARAM_FUNC, true, "function")
  }

  def validateOptions(cmdLine: CommandLine): Boolean = cmdLine.hasOption(AdvCli.CLI_PARAM_I) && cmdLine.hasOption(AdvCli.CLI_PARAM_O)

  def ansjSegFile(inputFile: String, outputFile: String) = {
    val counter = new FreqDist[String]
    AdvFile.loadFileInDelimitLine(new FileInputStream(inputFile), (ln) => {
      val posTags = ToAnalysis.parse(ln)
      new NatureRecognition(posTags).recognition()
      posTags.filter(_.getNatrue.natureStr.contains("n")).foreach(term => counter.incr(term.getName + "$" + term.getNatrue.natureStr))
    })

    val fileWriter = new FileWriter(outputFile)
    counter.foreach(pair => fileWriter.write("%s\t%s\n".format(pair._1, pair._2)))
    fileWriter.close()
    println("output file is " + outputFile)
  }

  def segFile(inputFile: String, outputFile: String) = {
    val dic = new util.HashMap[String, String]
    AdvFile.loadFileInDelimitLine(conf.getConfResourceAsInputStream(conf.get(TripleConsts.CONF_DIC_FILE)), (ln: String) => dic.put(ln, ln))

    val fileWriter = new FileWriter(outputFile)
    var counter = 0
    for (line <- Source.fromFile(inputFile).getLines()) {
      val wordsPos = StringUtil.backwardMaxMatch(line, dic, 100, 1)
      val segrst = wordsPos.map(position => {
        val word = line.substring(position.first, position.second)
        word
      }).mkString(" ")
      fileWriter.write(segrst + '\n')
      counter += 1
      println("Finishing processing %s\r".format(counter))
    }
    fileWriter.close()
    println("Output file is " + outputFile)
  }
}