SandishKumarHN commented on code in PR #39550:
URL: https://github.com/apache/spark/pull/39550#discussion_r1170615200
##########
connector/protobuf/src/main/scala/org/apache/spark/sql/protobuf/functions.scala:
##########
@@ -72,51 +69,129 @@ object functions {
}
/**
- * Converts a binary column of Protobuf format into its corresponding
catalyst value. The
- * specified Protobuf class must match the data, otherwise the behavior is
- * undefined: it may fail or return arbitrary result. The jar containing
Java class should be
+ * Converts a binary column of Protobuf format into its corresponding
catalyst value.
+ * `messageClassName` points to Protobuf Java class. The jar containing Java
class should be
* shaded. Specifically, `com.google.protobuf.*` should be shaded to
* `org.sparkproject.spark-protobuf.protobuf.*`.
+ * https://github.com/rangadi/shaded-protobuf-classes is useful to create
shaded jar from
+ * Protobuf files.
*
* @param data
* the binary column.
- * @param shadedMessageClassName
- * The Protobuf class name. E.g.
<code>org.spark.examples.protobuf.ExampleEvent</code>.
+ * @param messageClassName
+ * The full name for Protobuf Java class. E.g.
<code>com.example.protos.ExampleEvent</code>.
* The jar with these classes needs to be shaded as described above.
* @since 3.4.0
*/
@Experimental
- def from_protobuf(data: Column, shadedMessageClassName: String): Column = {
- new Column(ProtobufDataToCatalyst(data.expr, shadedMessageClassName))
+ def from_protobuf(data: Column, messageClassName: String): Column = {
+ new Column(ProtobufDataToCatalyst(data.expr, messageClassName))
}
/**
- * Converts a column into binary of protobuf format.
+ * Converts a binary column of Protobuf format into its corresponding
catalyst value.
+ * `messageClassName` points to Protobuf Java class. The jar containing Java
class should be
+ * shaded. Specifically, `com.google.protobuf.*` should be shaded to
+ * `org.sparkproject.spark-protobuf.protobuf.*`.
+ * https://github.com/rangadi/shaded-protobuf-classes is useful to create
shaded jar from
+ * Protobuf files.
+ *
+ * @param data
+ * the binary column.
+ * @param messageClassName
+ * The full name for Protobuf Java class. E.g.
<code>com.example.protos.ExampleEvent</code>.
+ * The jar with these classes needs to be shaded as described above.
+ * @param options
+ * @since 3.4.0
+ */
+ @Experimental
Review Comment:
@rangadi After doing some initial reading on the .pbf file format, I
understand that it is a binary file format that uses zlib compression to
compress protobuf format data. This means that in order to read a .pbf file, we
need to first decompress the data.
- If we want to break a big .pdf file into N chunks, we can use the .pdf
BlobHeader to determine the size of the file and then split it into N
equal-sized chunks. However, this will not work if the whole .pdf file is a
single large protobuf value. In this case, we will need to use a different
approach to break the file into chunks.
- One approach is to use the protobuf schema to break the file into chunks.
We can use the protobuf schema to identify the different fields in the protobuf
data and then break the file into chunks based on these fields.
- Another approach is to use a streaming parser to break the file into
chunks.
- Once we have broken the file into chunks, we can distribute the chunks to
different Spark executors. We can then parse the chunks into Spark DataFrames.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
