[GitHub] spark pull request #20929: [SPARK-23772][SQL][WIP] Provide an option to igno...

[mengxr <gi...@git.apache.org>]

Github user mengxr commented on a diff in the pull request:

    https://github.com/apache/spark/pull/20929#discussion_r180291486
  
    --- Diff: sql/catalyst/src/main/scala/org/apache/spark/sql/types/TypePlaceholder.scala ---
    @@ -0,0 +1,23 @@
    +/*
    + * Licensed to the Apache Software Foundation (ASF) under one or more
    + * contributor license agreements.  See the NOTICE file distributed with
    + * this work for additional information regarding copyright ownership.
    + * The ASF licenses this file to You under the Apache License, Version 2.0
    + * (the "License"); you may not use this file except in compliance with
    + * the License.  You may obtain a copy of the License at
    + *
    + *    http://www.apache.org/licenses/LICENSE-2.0
    + *
    + * Unless required by applicable law or agreed to in writing, software
    + * distributed under the License is distributed on an "AS IS" BASIS,
    + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
    + * See the License for the specific language governing permissions and
    + * limitations under the License.
    + */
    +
    +package org.apache.spark.sql.types
    +
    +/**
    + * An internal type that is a not yet available and will be replaced by an actual type later.
    + */
    +case object TypePlaceholder extends StringType
    --- End diff --
    
    Is it necessary to introduce a new DataType? Would it be the same if we use `NullType`? With the flag on, at the end of schema inference, `NullType`, `ArrayType(NullType)`, etc should be dropped instead of using StringType as fallback. Basically, during schema inference, we keep the one that reveals more details, for example:
    
    ```
    (NullType, ArrayType(NullType)) => ArrayType(NullType)
    (ArrayType(NullType), ArrayType(StructType(Field("a", NullType)))) => ArrayType(StructType(Field("a", NullType))))
    ```
    
    At the end, we implement a util method that determine whether a field is all null and drop them if true. It should be done recursively. I have an internal implementation that implements a similar logic, but on the JSON record itself. You might want to apply it to data types.
    
    ```scala
      /**
       * Removes null fields recursively from the input JSON record.
       * An array is null if all its elements are null.
       * An object is null if all its values are null.
       */
      def removeNullRecursively(jsonStr: String): String = {
        val json = parse(jsonStr)
        val cleaned = doRemoveNullRecursively(json)
        compact(render(cleaned)) // should handle null correctly
      }
    
      private def doRemoveNullRecursively(value: JValue): JValue = {
        value match {
          case null =>
            null
    
          case JNull =>
            null
    
          case JArray(values) =>
            val cleaned = values.map(doRemoveNullRecursively)
            if (cleaned.exists(_ != null)) {
              JArray(cleaned)
            } else {
              null
            }
    
          case JObject(pairs) =>
            val cleaned = pairs.flatMap { case (k, v) =>
              val cv = doRemoveNullRecursively(v)
              if (cv != null) {
                Some((k, cv))
              } else {
                None
              }
            }
            if (cleaned.nonEmpty) {
              JObject(cleaned)
            } else {
              null
            }
    
          // all other types are non-null
          case _ =>
            value
        }
      }
    ```


---

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