How to read Parquet file from S3 without spark? Java

user_12830699 0
Viewed 6161

Currently, I am using the Apache ParquetReader for reading local parquet files, which looks something like this:

ParquetReader<GenericData.Record> reader = null;
    Path path = new Path("userdata1.parquet");
    try {
        reader = AvroParquetReader.<GenericData.Record>builder(path).withConf(new Configuration()).build();
        GenericData.Record record;
        while ((record = reader.read()) != null) {
            System.out.println(record);

However, I am trying to access a parquet file through S3 without downloading it. Is there a way to parse Inputstream directly with parquet reader?

parquet apache-spark hadoop amazon-s3 java
3 Answers

Yes, the latest versions of hadoop include support for S3 filesystem. Use the s3a client from hadoop-aws library to directly access the S3 filesystem.

The HadoopInputFile Path should be constructed as s3a://bucket-name/prefix/key along with the authentication credentials access_key and secret_key configured using the properties

  • fs.s3a.access.key
  • fs.s3a.secret.key

Additionally, you would require these dependant libraries

  • hadoop-common JAR
  • aws-java-sdk-bundle JAR

Read more: Relevant configuration properties

user_73034470

I got it working with this following dependencies

compile 'org.slf4j:slf4j-api:1.7.5'
compile 'org.slf4j:slf4j-log4j12:1.7.5'
compile 'org.apache.parquet:parquet-avro:1.12.0'
compile 'org.apache.avro:avro:1.10.2'
compile 'com.google.guava:guava:11.0.2'
compile 'org.apache.hadoop:hadoop-client:2.4.0'
compile 'org.apache.hadoop:hadoop-aws:3.3.0'   
compile 'org.apache.hadoop:hadoop-common:3.3.0'      
compile 'com.amazonaws:aws-java-sdk-core:1.11.563'
compile 'com.amazonaws:aws-java-sdk-s3:1.11.563'

Example

Path path = new Path("s3a://yours3path");
Configuration conf = new Configuration();
conf.set("fs.s3a.access.key", "KEY");
conf.set("fs.s3a.secret.key", "SECRET");
conf.set("fs.s3a.impl", "org.apache.hadoop.fs.s3a.S3AFileSystem");
conf.setBoolean("fs.s3a.path.style.access", true);
conf.setBoolean(org.apache.parquet.avro.AvroReadSupport.READ_INT96_AS_FIXED, true);

InputFile file = HadoopInputFile.fromPath(path, conf);
ParquetReader<GenericRecord> reader = AvroParquetReader.<GenericRecord>builder(file).build();
GenericRecord record;
while ((record = reader.read()) != null) {
  System.out.println(record);
}
user_122604420

Just adding on top of @franklinsijo , for freshers starting S3, Please note that access key and secret key is set for Hadoop Configuration: Here is a snippet of code that might be useful:

public static void main(String[] args) throws IOException {
String PATH_SCHEMA = "s3a://xxx/xxxx/userdata1.parquet";
Path path = new Path(PATH_SCHEMA);
Configuration conf = new Configuration();
conf.set("fs.s3a.access.key", "xxxxx");
conf.set("fs.s3a.secret.key", "xxxxx");
InputFile file = HadoopInputFile.fromPath(path, conf);
ParquetReader<GenericRecord> reader = AvroParquetReader.<GenericRecord>builder(file).build();
GenericRecord record;
while ((record = reader.read()) != null) {
  System.out.println(record.toString());
}

My imports:

import org.apache.avro.generic.GenericRecord;
import org.apache.hadoop.conf.Configuration;
import org.apache.parquet.hadoop.ParquetReader;
import org.apache.parquet.avro.AvroParquetReader;
import org.apache.parquet.hadoop.util.HadoopInputFile;
import org.apache.parquet.io.InputFile;
import java.io.IOException;
import org.apache.hadoop.fs.Path;
user_136861390
Related