saurfang / sparksql-protobuf Goto Github PK

View Code? Open in Web Editor NEW

96.0 10.0 37.0 63 KB

Read SparkSQL parquet file as RDD[Protobuf]

Home Page: http://spark-packages.org/package/saurfang/sparksql-protobuf

License: Apache License 2.0

Scala 53.32% Java 45.47% Shell 1.21%

protobuf sparksql parquet

sparksql-protobuf's Introduction

sparksql-protobuf

This library provides utilities to work with Protobuf objects in SparkSQL. It provides a way to read parquet file written by SparkSQL back as an RDD of compatible protobuf object. It can also converts RDD of protobuf objects into DataFrame.

For sbt 0.13.6+

resolvers += Resolver.jcenterRepo

libraryDependencies ++= Seq(
    "com.github.saurfang" %% "sparksql-protobuf" % "0.1.3",
    "org.apache.parquet" % "parquet-protobuf" % "1.8.3"
)

Motivation

SparkSQL is very powerful and easy to use. However it has a few limitations and schema is only detected during runtime makes developers a lot less confident that they will get things right at first time. Static typing helps a lot! This is where protobuf comes in:

Protobuf defines nested data structure easily
It doesn't constraint you to the 22 fields limit in case class (no longer true once we upgrade to 2.11+)
It is language agnostic and generates code that gives you native objects hence you get all the benefit of type checking and code completion unlike operating Row in Spark/SparkSQL

Features

Read Parquet file as `RDD[Protobuf]`

val personsPB = new ProtoParquetRDD(sc, "persons.parquet", classOf[Person])

where we need SparkContext, parquet path and protobuf class.

This converts the existing workflow:

Ingest raw data as DataFrame with nested data structure
Create awkward runtime type checking udfs
Transform raw DataFrame using above udfs into a tabular DataFrame for data analytics

Ingest raw data as DataFrame with nested data structure and persist as Parquet file
Read Parquet file back as RDD[Protobuf]
Perform any data transformation and extraction by working with compile typesafe Protobuf getters
Create a DataFrame out of the above transformation and perform additional downstream data analytics on the tabular DataFrame

Infer SparkSQL Schema from Protobuf Definition

val personSchema = ProtoReflection.schemaFor[Person].dataType.asInstanceOf[StructType]

Convert `RDD[Protobuf]` to `DataFrame`

import com.github.saurfang.parquet.proto.spark.sql._
val personsDF = sqlContext.createDataFrame(protoPersons)

For more information, please see test cases.

Under the hood

ProtoMessageConverter has been improved to read from LIST specification according to latest parquet documentation. This implementation should be backwards compatible and is able to read repeated fields generated by writers like SparkSQL.
ProtoMessageParquetInputFormat helps the above process by correctly returning the built protobuf object as value.
ProtoParquetRDD abstract the Hadoop input format and returns an RDD of your protobuf objects from parquet files directly.
ProtoReflection infers SparkSQL schema from any Protobuf message class.
ProtoRDDConversions converts Protobuf objects into SparkSQL rows.

Related Work

Elephant Bird

Spark-9999

sparksql-protobuf's People

Contributors

Stargazers

Watchers

sparksql-protobuf's Issues

Support for Protobuf DynamicMessages

Hi,
We tested sparksql-protobuf here, works pretty well,
but we need to generate the protobuf classes with Protoc .

We have created a feature to pull the latest proto descriptors version from S3 , so that we don't have to keep compiling and including new versions of the compiled ProtoBuf classes. The problem is that the com.google.protobuf.message doesn't contains the method getDescriptor(), it contains getDescriptorForType() instead.

is it possible to create a Support for Protobuf DynamicMessages like this parquet-proto issue ?
apache/parquet-mr#414

support scala 2.11

as the title, can sparksql-protobuf support scala version 2.11?

[Question] Is there a release compatible with Scala 2.11 available?

Looking at https://bintray.com/saurfang/maven/sparksql-protobuf I can see that the release is compatible to Scala 2.10. Is it hosted somewhere else also with Scala 2.11 compatibility?

[Question] How to build the project properly?

Hi saurfang,

Awesome project. I am trying to build it locally. This is my very first time exposing to sbt build tool. I am trying to run sbt compile, but it failed due to not be able to to pick up protobuf and @transient.

~/internship/sparksql-protobuf on  master! ⌚ 9:06:21
$ protoc --version
libprotoc 2.6.1

$ tree -L 2 -d
.
├── lib .     // output of my compiled protobuf 2.6.1
│   ├── bin
│   ├── include
│   └── lib
├── project
│   ├── project
│   └── target
├── src
│   ├── main
│   └── test
└── target
    ├── protobuf_external
    ├── resolution-cache
    ├── scala-2.11
    └── streams

When I am trying to compile the project, sbt throws the following errors:

$ sbt compile -Xlint
[info] Loading global plugins from /Users/datng2/.sbt/0.13/plugins
[info] Loading project definition from /Users/datng2/internship/sparksql-protobuf/project
Missing bintray credentials /Users/datng2/.bintray/.credentials. Some bintray features depend on this.
[info] Set current project to sparksql-protobuf (in build file:/Users/datng2/internship/sparksql-protobuf/)
[warn] Credentials file /Users/datng2/.bintray/.credentials does not exist
[warn] Credentials file /Users/datng2/.ivy2/.sbtcredentials does not exist
[info] Compiling 5 Scala sources and 6 Java sources to /Users/datng2/internship/sparksql-protobuf/target/scala-2.11/classes...
[warn] /Users/datng2/internship/sparksql-protobuf/src/main/scala/com/github/saurfang/parquet/proto/spark/ProtoParquetRDD.scala:18: no valid targets for annotation on value conf - it is discarded unused. You may specify targets with meta-annotations, e.g. @(transient @param)
[warn]                                                         @transient conf: Configuration
[warn]                                                          ^
[warn] one warning found
[error] /Users/datng2/internship/sparksql-protobuf/src/test/generated/com/github/saurfang/parquet/proto/AddressBook.java:79: error: cannot find symbol
[error]     com.google.protobuf.ProtocolStringList
[error]                        ^
[error]   symbol:   class ProtocolStringList
[error]   location: package com.google.protobuf
[error] /Users/datng2/internship/sparksql-protobuf/src/test/generated/com/github/saurfang/parquet/proto/AddressBook.java:1367: error: cannot find symbol
[error]     public com.google.protobuf.ProtocolStringList
[error]                               ^
[error]   symbol:   class ProtocolStringList
[error]   location: package com.google.protobuf
[error] /Users/datng2/internship/sparksql-protobuf/src/test/generated/com/github/saurfang/parquet/proto/AddressBook.java:2247: error: cannot find symbol
[error]       public com.google.protobuf.ProtocolStringList
[error]                                 ^
[error]   symbol:   class ProtocolStringList
[error]   location: package com.google.protobuf
[error] /Users/datng2/internship/sparksql-protobuf/src/test/generated/com/github/saurfang/parquet/proto/AddressBook.java:219: error: cannot find symbol
[error]           address_ = address_.getUnmodifiableView();
[error]                              ^
[error]   symbol:   method getUnmodifiableView()
[error]   location: variable address_ of type LazyStringList
[error] /Users/datng2/internship/sparksql-protobuf/src/test/generated/com/github/saurfang/parquet/proto/AddressBook.java:1692: error: cannot find symbol
[error]           address_ = address_.getUnmodifiableView();
[error]                              ^
[error]   symbol:   method getUnmodifiableView()
[error]   location: variable address_ of type LazyStringList
[error] /Users/datng2/internship/sparksql-protobuf/src/test/generated/com/github/saurfang/parquet/proto/AddressBook.java:2249: error: cannot find symbol
[error]         return address_.getUnmodifiableView();
[error]                        ^
[error]   symbol:   method getUnmodifiableView()
[error]   location: variable address_ of type LazyStringList
[error] Note: /Users/datng2/internship/sparksql-protobuf/src/main/java/com/github/saurfang/parquet/proto/ProtoLISTReadSupport.java uses or overrides a deprecated API.
[error] Note: Recompile with -Xlint:deprecation for details.
[error] Note: Some input files use unchecked or unsafe operations.
[error] Note: Recompile with -Xlint:unchecked for details.
[error] 6 errors
[error] (compile:compile) javac returned nonzero exit code
[error] Total time: 6 s, completed Feb 8, 2019 9:14:21 AM

Any pointer or help would be greatly appreciated. Thanks :)

Protobuf 3

How can I help make sure this project can support proto3?

"Loose" Schema Mode

Hi, this is a great, useful library.

I tested this libraries compatibility with various 'schema evolution' scenarios where I changed the protobuf (added fields, renamed fields, changed optional->repeated, etc) pass it to the ProtoParquetWriter and try to read it back using ProtoParquetRDD. I found that many 'legal' protobuf evolution rules, like field renames or type changes were not compatible with this library.

Now, I don't want to conflate parquet's own schema evolution rules, or spark 'schema merging' capabilities (http://spark.apache.org/docs/latest/sql-programming-guide.html#schema-merging), but overall it seems technically feasible to have a 'loose' mode where fields that exist in parquet and which have a compatible protobuf type will hydrate the protobuf, regardless of any schema mismatches which makes the process fail hard and fast.

Any thoughts on this?

Examples:
Adding new Enum:
Caused by: org.apache.parquet.io.InvalidRecordException: Illegal enum value

Adding new field:
org.apache.parquet.schema.IncompatibleSchemaModificationException: Cant find "timeout" Scheme mismatch

Support protobuf GeneratedMessage3 for ProtoParqeutRDD

Current code just support GeneratedMessage
that is for first version of protobuf but now many users like me use proto 3
please update ProtoParquetRDD for supporting GenerateMessage3

how to query on SPARKSQL

Hi Team,

I am trying following SQL:-

var work__store_level_vend_pack_loc_final_data =

  sparksession.read.format("csv")
 .option("header", "true")
 .option("delimiter", "|")
 .option("inferSchema", "true")
 .load("C:\\Users\\jabin\\Desktop\\project_files\\work__store_level_vend_pack_loc_final_data.txt");
 
  work__store_level_vend_pack_loc_final_data.registerTempTable("work__store_level_vend_pack_loc_final_data_table");

var r1 = sparksession.sqlContext.sql( "SELECT shc_item_id ,'K' as source_owner_cd,item_purchase_status_cd, vendor_package_id,vendor_package_purchase_status_cd,flow_type_cd as vendor_package_flow_type_cd,vendor_carton_qty,vendor_stock_nbr,ksn_package_id,ksn_purchase_status_cd,import_ind,sears_divission_nbr,sears_item_nbr,sears_sku_nbr,scan_based_trading_ind,cross_merchandising_cd,retail_carton_vendor_package_id,vendor_package_owner_cd,can_carry_model_id,'' AS days_to_check_begin_day_qty,'' AS days_to_check_end_day_qty ,dotcom_allocation_ind ,retail_carton_internal_package_qty,allocation_replenishment_cd,shc_item_type_cd,idrp_order_method_cd,source_package_qty as store_source_package_qty,order_duns_nbr FROM work__store_level_vend_pack_loc_final_data_table WHERE flow_type_cd = 'JIT' OR servicing_dc_nbr > '0' ")
// .collect.foreach(println)

now i want to distinct all column of the r1 using sparksession.sqlContext.sql("")
how to do above thing?