This code implements class FileBuffer that can be written to with a java.io.OutputStream and read via a java.io.InputStream. Writes go into the tail of the buffer and reads take place at the head. The implementation ensures that reading doesn't overtake writing which preserves the blocking semantics of java.io.InputStream.
+--------------+
| |
OutputStream.write -> | FileBuffer | -> InputStream.read
| |
+--------------+
FileBuffer was meant to allow concurrently downloading a large amount of data and reading the data -- without using too much memory. FileBuffer takes a threshold parameter that specificies the maximum # of bytes to buffer in memory before starting to buffer to a file. Furthermore, the overall in-memory portion of the buffer can be split into segments. When the reader of the FileBuffer reads beyond a segment, the memory for that segment can be freed.
+--------------+ +--------------+ +------------+
| MemorySegment| -> | MemorySegment| -> .. -> | FileSegment|
+--------------+ +--------------+ | |
| |
+------------+
The concurrency model of this code assumes a single reader and a single writer. However, the read and write can be on separate threads, of course.
Copyright © 2015 Zensight, Inc.
Distributed under the MIT License.