read footer using 1 call readFully(byte[8]) instead of 5 calls ( 4 x read() for footer length + 1 x read(byte[4]) for magic marker )

En parquet-hadoop/src/main/java/org/apache/parquet/hadoop/ParquetFileReader.java, inspecciona primero readFooter() y las lecturas actuales de la longitud del footer y de los valores magic. Verifica que una sola llamada a readFully(byte[8]) conserve el manejo tanto del magic del footer estándar como del footer cifrado y, después, ejecuta las pruebas relevantes del proyecto; se considera completado cuando el análisis del footer siga siendo correcto con menos llamadas de lectura.

Describe the enhancement requested

This is a minor performance improvement, but worth when reading many files.
read footer using 1 call readFully(byte[8]) instead of 5 calls ( 4 x read() for footer length + 1 x read(byte[4]) for magic marker )

in summary the patch is for file ParquetFileReader.java, method "readFooter()" :

--- a/parquet-hadoop/src/main/java/org/apache/parquet/hadoop/ParquetFileReader.java
+++ b/parquet-hadoop/src/main/java/org/apache/parquet/hadoop/ParquetFileReader.java
@@ -585,14 +585,18 @@ public class ParquetFileReader implements Closeable {
     }

     // Read footer length and magic string - with a single seek
-    byte[] magic = new byte[MAGIC.length];
-    long fileMetadataLengthIndex = fileLen - magic.length - FOOTER_LENGTH_SIZE;
+    long fileMetadataLengthIndex = fileLen - MAGIC.length - FOOTER_LENGTH_SIZE;
     LOG.debug("reading footer index at {}", fileMetadataLengthIndex);
     f.seek(fileMetadataLengthIndex);
-    int fileMetadataLength = readIntLittleEndian(f);
-    f.readFully(magic);
+    byte[] magicAndLengthBytes = new byte[FOOTER_LENGTH_SIZE + MAGIC.length];
+    f.readFully(magicAndLengthBytes);
+    int fileMetadataLength = readIntLittleEndian(magicAndLengthBytes, 0);

     boolean encryptedFooterMode;
+    // using JDK >= 9: if (Arrays.equals(MAGIC, 0, MAGIC.length, magicAndLengthBytes, FOOTER_LENGTH_SIZE, FOOTER_LENGTH_SIZE + MAGIC.length)) {
+    // using JDK <= 8: need extract sub array then compare
+    byte[] magic = new byte[MAGIC.length];
+    System.arraycopy(magicAndLengthBytes, FOOTER_LENGTH_SIZE, magic, 0, MAGIC.length);
     if (Arrays.equals(MAGIC, magic)) {
       encryptedFooterMode = false;
     } else if (Arrays.equals(EFMAGIC, magic)) {

Component(s)

Core