Buffering the input stream for metadata read

twitter-forks · Jul 5, 2018 · 01812d5 · 01812d5
1 parent c2428ab
commit 01812d5
Showing 1 changed file with 5 additions and 3 deletions.
diff --git a/presto-hive/src/main/java/com/facebook/presto/hive/parquet/reader/ParquetMetadataReader.java b/presto-hive/src/main/java/com/facebook/presto/hive/parquet/reader/ParquetMetadataReader.java
@@ -37,6 +37,7 @@
 import parquet.schema.Type.Repetition;
 import parquet.schema.Types;
 
+import java.io.BufferedInputStream;
 import java.io.EOFException;
 import java.io.IOException;
 import java.io.InputStream;
@@ -85,10 +86,11 @@ public static ParquetMetadata readFooter(FSDataInputStream inputStream, Path fil
         long metadataLengthIndex = fileSize - PARQUET_METADATA_LENGTH - MAGIC.length;
 
         inputStream.seek(metadataLengthIndex);
-        int metadataLength = readIntLittleEndian(inputStream);
+        InputStream in = new BufferedInputStream(inputStream, PARQUET_METADATA_LENGTH + MAGIC.length);
+        int metadataLength = readIntLittleEndian(in);
 
         byte[] magic = new byte[MAGIC.length];
-        inputStream.readFully(magic);
+        validateParquet(in.read(magic) == magic.length, "No enough data for MAGIC");
         validateParquet(Arrays.equals(MAGIC, magic), "Not valid Parquet file: %s expected magic number: %s got: %s", file, Arrays.toString(MAGIC), Arrays.toString(magic));
 
         long metadataIndex = metadataLengthIndex - metadataLength;
@@ -98,7 +100,7 @@ public static ParquetMetadata readFooter(FSDataInputStream inputStream, Path fil
                 file,
                 metadataIndex);
         inputStream.seek(metadataIndex);
-        FileMetaData fileMetaData = readFileMetaData(inputStream);
+        FileMetaData fileMetaData = readFileMetaData(new BufferedInputStream(inputStream, metadataLength));
         List<SchemaElement> schema = fileMetaData.getSchema();
         validateParquet(!schema.isEmpty(), "Empty Parquet schema in file: %s", file);