Github user mridulm commented on a diff in the pull request:

    https://github.com/apache/spark/pull/15408#discussion_r82816160
  
    --- Diff: 
core/src/main/java/org/apache/spark/io/NioBufferedFileInputStream.java ---
    @@ -0,0 +1,129 @@
    +/*
    + * Licensed under the Apache License, Version 2.0 (the "License");
    + * you may not use this file except in compliance with the License.
    + * You may obtain a copy of the License at
    + *
    + *     http://www.apache.org/licenses/LICENSE-2.0
    + *
    + * Unless required by applicable law or agreed to in writing, software
    + * distributed under the License is distributed on an "AS IS" BASIS,
    + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
    + * See the License for the specific language governing permissions and
    + * limitations under the License.
    + */
    +package org.apache.spark.io;
    +
    +import org.apache.spark.storage.StorageUtils;
    +
    +import java.io.File;
    +import java.io.IOException;
    +import java.io.InputStream;
    +import java.nio.ByteBuffer;
    +import java.nio.channels.FileChannel;
    +import java.nio.file.StandardOpenOption;
    +
    +/**
    + * {@link InputStream} implementation which uses direct buffer
    + * to read a file to avoid extra copy of data between Java and
    + * native memory which happens when using {@link 
java.io.BufferedInputStream}.
    + * Unfortunately, this is not something already available in JDK,
    + * {@link sun.nio.ch.ChannelInputStream} supports reading a file using nio,
    + * but does not support buffering.
    + *
    + * TODO: support {@link #mark(int)}/{@link #reset()}
    + *
    + */
    +public final class NioBufferedFileInputStream extends InputStream {
    +
    +  private static int DEFAULT_BUFFER_SIZE_BYTES = 8192;
    +
    +  private final ByteBuffer byteBuffer;
    +
    +  private final FileChannel fileChannel;
    +
    +  public NioBufferedFileInputStream(File file, int bufferSizeInBytes) 
throws IOException {
    +    byteBuffer = ByteBuffer.allocateDirect(bufferSizeInBytes);
    +    fileChannel = FileChannel.open(file.toPath(), StandardOpenOption.READ);
    +    byteBuffer.flip();
    +  }
    +
    +  public NioBufferedFileInputStream(File file) throws IOException {
    +    this(file, DEFAULT_BUFFER_SIZE_BYTES);
    +  }
    +
    +  /**
    +   * Checks weather data is left to be read from the input stream.
    +   * @return true if data is left, false otherwise
    +   * @throws IOException
    +   */
    +  private boolean refill() throws IOException {
    +    if (!byteBuffer.hasRemaining()) {
    +      byteBuffer.clear();
    +      int nRead = fileChannel.read(byteBuffer);
    +      if (nRead <= 0) {
    --- End diff --
    
    There arearly a few concerns here:
    * immediate was to prevent underflow exception in get ()
    * file channels don't return 0 when reading off local disk unless 
configured as non blocking.
    Though this might be implementation detail we don't want to rely on ? (If 
yes, we got to fix other places in our code as @srowen mentioned )
    * if we do want to handle 0 as retry, we will need to ensure we don't go 
into infinite loop (since this should be an anamoly to begin with).
    Typical readFully () does infinite loop since it is working off generic 
InputStream where this behavior is possible, in this context it would be the 
exception.
    
    
    
    Having said that, I can see a nas or nfs mount repeatedly returning 0 even 
though file is not done.
    We could special case 0 and retry for 'n' (3 ?) times whenever we see a 
zero and file not done ?
    



---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---

---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org

Reply via email to