package com.bigdata.bfs;
import com.bigdata.sparse.KeyType;
import com.bigdata.sparse.Schema;
import com.bigdata.sparse.SparseRowStore;
/**
* The schema for metadata about file versions stored in the
* {@link BigdataFileSystem}. Some well known properties are always defined,
* but any property may be stored - ideally within their own namespace!
* <p>
* Note: File version creation time and update times are available using the
* {@link SparseRowStore}, which stores and reports the timestamp for each
* property value. Convenience methods are available on
* {@link RepositoryDocumentImpl} to report those timestamps. Timestamps for
* file blocks can NOT be obtained.
* <p>
* Note: A content length property was deliberately NOT defined. The design is
* geared towards very large file and asynchronous read/write of file blocks.
* The length of short files may be readily computed by the expediency of
* sucking their contents into a buffer. Large files should always be processed
* using a stream-oriented technique or distributed to concurrent clients in
* block sized pieces.
*
* @todo other obvious metadata would include the user identifier associated
* with each update request.
*/
public class FileMetadataSchema extends Schema {
/**
*
*/
private static final long serialVersionUID = 2908749650061841935L;
/**
* The content identifer is an arbitrary Unicode {@link String} whose value
* is defined by the client. This is generally a pathname, similar to a path
* in a file system.
*/
public static transient final String ID = FileMetadataSchema.class
.getPackage().getName()
+ ".Id";
/**
* The MIME type associated with the content (the same semantics as the
* HTTP <code>Content-Type</code> header).
*/
public static transient final String CONTENT_TYPE = FileMetadataSchema.class
.getPackage().getName()
+ ".ContentType";
/**
* The encoding, if any, used to convert the byte[] content to
* characters.
* <p>
* Note: This is typically deduced from an analysis of the MIME Type in
* <code>Content-Type</code> header and at times the leading bytes of
* the response body itself.
*/
public static transient final String CONTENT_ENCODING = FileMetadataSchema.class
.getPackage().getName()
+ ".ContentEncoding";
/**
* The file version number. Together the file {@link #ID} and the file
* {@link #VERSION} form the primary key for the data index. A
* <code>null</code> value is stored in this field when the file version
* is deleted.
*/
public static transient final String VERSION = FileMetadataSchema.class
.getPackage().getName()
+ ".Version";
/**
* Optional boolean property. When present, the text in the file version
* will be tokenized and indexed for free text search. Non-text files should
* not set this property. Instead, a workflow or map/reduce job should be
* written to generate a text-only variant for the indexer.
*
* @todo need a property that is the "URI" of the indexed document?
*/
public static transient final String INDEX_TEXT = FileMetadataSchema.class
.getPackage().getName()
+ ".IndexText";
public FileMetadataSchema() {
super( "__bfs"/*FileMetadataSchema.class.getName()*/, ID, KeyType.Unicode);
}
}