gistfile1.scala

import org.allenai.common.Config._

object ClassificationReader {                                                                       
  /** Reads classifications from disk, using the given config to find the classification file(s).   
    * This requires that the config have a key {classificationType}.file for each classification
    * type.
    * @throws IllegalArgumentException if any of the required keys are missing                      
    */
  def readClassifications(config: Config): Seq[PaperClassification] = {                             
    // Read a classification per file. This will generate PaperClassification instances with a      
    // single value.
    val paperClassifications = for {                                                                
      classification <- ClassificationTypes.Values                                                  
      filePath = config.get[String](s"$classification.file") getOrElse {                            
        throw new IllegalArgumentException(s"config is missing key $classification.file")           
      } 
      (paperId, value) <- readFromFile(filePath)                                                    
    } yield PaperClassification(paperId, Map(classification -> value))                              
    
    // Collapse duplicate entries into a single sequence.                                           
    for {
      (paperId, classifications) <- paperClassifications.groupBy(_.paperId)                         
      allClassifications = (classifications flatMap { _.classifications }).toMap                    
    } yield PaperClassification(paperId, allClassifications)
  } 
  
  /** Reads a classification from a file.                                                           
    * @return the mapping of paper ID to classification value                                       
    * @throws IllegalArgumentException if any line doesn't contain two tab-separated values         
    * @throws NumberFormatException if the second value on any line isn't a Double
    */
  def readFromFile(file: File): Map[String, Double] = {                                             
    (Resource.using(Source.fromFile(file)) { source =>
      for {
        line <- source.getLine                                                                      
        values = line.split('\t')                                                                   
      } yield values match {
        case Array(paperId, value) => (paperId -> value.toDouble)                                   
        case _ => throw new IllegalArgumentException("Malformed line " + line)                      
      } 
    }).toMap                                                                                        
  } 
}