spark-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From rxin <...@git.apache.org>
Subject [GitHub] spark pull request: [SPARK-8307] [SQL] improve timestamp from parq...
Date Thu, 11 Jun 2015 07:47:56 GMT
Github user rxin commented on a diff in the pull request:

    https://github.com/apache/spark/pull/6759#discussion_r32196749
  
    --- Diff: sql/core/src/main/scala/org/apache/spark/sql/parquet/ParquetConverter.scala
---
    @@ -498,69 +493,21 @@ private[parquet] object CatalystArrayConverter {
     }
     
     private[parquet] object CatalystTimestampConverter {
    -  // TODO most part of this comes from Hive-0.14
    -  // Hive code might have some issues, so we need to keep an eye on it.
    -  // Also we use NanoTime and Int96Values from parquet-examples.
    -  // We utilize jodd to convert between NanoTime and Timestamp
    -  val parquetTsCalendar = new ThreadLocal[Calendar]
    -  def getCalendar: Calendar = {
    -    // this is a cache for the calendar instance.
    -    if (parquetTsCalendar.get == null) {
    -      parquetTsCalendar.set(Calendar.getInstance(TimeZone.getTimeZone("GMT")))
    -    }
    -    parquetTsCalendar.get
    -  }
    -  val NANOS_PER_SECOND: Long = 1000000000
    -  val SECONDS_PER_MINUTE: Long = 60
    -  val MINUTES_PER_HOUR: Long = 60
    -  val NANOS_PER_MILLI: Long = 1000000
    +  // see http://stackoverflow.com/questions/466321/convert-unix-timestamp-to-julian
    +  val JULIAN_DAY_OF_EPOCH = 2440587.5
    +  val SECONDS_PER_DAY = 60 * 60 * 24
     
    -  def convertToTimestamp(value: Binary): Timestamp = {
    +  def convertToTimestamp(value: Binary): Long = {
         val nt = NanoTime.fromBinary(value)
    -    val timeOfDayNanos = nt.getTimeOfDayNanos
    -    val julianDay = nt.getJulianDay
    -    val jDateTime = new JDateTime(julianDay.toDouble)
    -    val calendar = getCalendar
    -    calendar.set(Calendar.YEAR, jDateTime.getYear)
    -    calendar.set(Calendar.MONTH, jDateTime.getMonth - 1)
    -    calendar.set(Calendar.DAY_OF_MONTH, jDateTime.getDay)
    -
    -    // written in command style
    -    var remainder = timeOfDayNanos
    -    calendar.set(
    -      Calendar.HOUR_OF_DAY,
    -      (remainder / (NANOS_PER_SECOND * SECONDS_PER_MINUTE * MINUTES_PER_HOUR)).toInt)
    -    remainder = remainder % (NANOS_PER_SECOND * SECONDS_PER_MINUTE * MINUTES_PER_HOUR)
    -    calendar.set(
    -      Calendar.MINUTE, (remainder / (NANOS_PER_SECOND * SECONDS_PER_MINUTE)).toInt)
    -    remainder = remainder % (NANOS_PER_SECOND * SECONDS_PER_MINUTE)
    -    calendar.set(Calendar.SECOND, (remainder / NANOS_PER_SECOND).toInt)
    -    val nanos = remainder % NANOS_PER_SECOND
    -    val ts = new Timestamp(calendar.getTimeInMillis)
    -    ts.setNanos(nanos.toInt)
    -    ts
    -  }
    -
    -  def convertFromTimestamp(ts: Timestamp): Binary = {
    -    val calendar = getCalendar
    -    calendar.setTime(ts)
    -    val jDateTime = new JDateTime(calendar.get(Calendar.YEAR),
    -      calendar.get(Calendar.MONTH) + 1, calendar.get(Calendar.DAY_OF_MONTH))
    -    // Hive-0.14 didn't set hour before get day number, while the day number should
    -    // has something to do with hour, since julian day number grows at 12h GMT
    -    // here we just follow what hive does.
    -    val julianDay = jDateTime.getJulianDayNumber
    -
    -    val hour = calendar.get(Calendar.HOUR_OF_DAY)
    -    val minute = calendar.get(Calendar.MINUTE)
    -    val second = calendar.get(Calendar.SECOND)
    -    val nanos = ts.getNanos
    -    // Hive-0.14 would use hours directly, that might be wrong, since the day starts
    -    // from 12h in Julian. here we just follow what hive does.
    -    val nanosOfDay = nanos + second * NANOS_PER_SECOND +
    -      minute * NANOS_PER_SECOND * SECONDS_PER_MINUTE +
    -      hour * NANOS_PER_SECOND * SECONDS_PER_MINUTE * MINUTES_PER_HOUR
    -    NanoTime(julianDay, nanosOfDay).toBinary
    +    ((nt.getJulianDay - JULIAN_DAY_OF_EPOCH) * SECONDS_PER_DAY * 1e7
    --- End diff --
    
    similarly with convertFromTimestamp


---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.
---

---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org


Mime
View raw message