Github user jkbradley commented on a diff in the pull request:
https://github.com/apache/spark/pull/4233#discussion_r23909199
--- Diff:
mllib/src/main/scala/org/apache/spark/mllib/classification/LogisticRegression.scala
---
@@ -68,6 +79,65 @@ class LogisticRegressionModel (
case None => score
}
}
+
+ override def save(sc: SparkContext, path: String): Unit = {
+ val sqlContext = new SQLContext(sc)
+ import sqlContext._
+
+ // Create JSON metadata.
+ val metadata = LogisticRegressionModel.Metadata(
+ clazz = this.getClass.getName, version = Exportable.latestVersion)
+ val metadataRDD: DataFrame = sc.parallelize(Seq(metadata))
+ metadataRDD.toJSON.saveAsTextFile(path + "/metadata")
+ // Create Parquet data.
+ val data = LogisticRegressionModel.Data(weights, intercept, threshold)
+ val dataRDD: DataFrame = sc.parallelize(Seq(data))
+ dataRDD.saveAsParquetFile(path + "/data")
+ }
+}
+
+object LogisticRegressionModel extends Importable[LogisticRegressionModel]
{
+
+ private case class Metadata(clazz: String, version: String)
+
+ private case class Data(weights: Vector, intercept: Double, threshold:
Option[Double])
+
+ override def load(sc: SparkContext, path: String):
LogisticRegressionModel = {
+ val sqlContext = new SQLContext(sc)
+ import sqlContext._
+
+ // Load JSON metadata.
+ val metadataRDD = sqlContext.jsonFile(path + "/metadata")
--- End diff --
That's not quite my question: I think the confusion is mixing "row" (line)
in a text file vs "row" (or record) in an RDD. How about we store the metadata
in a single record in an RDD, but we print that RDD as multi-line JSON to a
single text file? It will be easier for humans to read and will be easy to
load as a single record as well.
---
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 infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
