|
Proposal
A component "HadoopXmlFiles" which takes a list of filenames/FileIntf/resources. These are loaded in order. Alternatively, run through all attributes with the name fileXXX or resourceXXX in the order they are declared. This makes inheritance easier and avoids complex file-or-resource logic. fileXXX: FileIntf or string filename. if "" or NULL then skip resourceXXX: string resource name, sfClass classloader. If "" or NULL then skip failOnError true/false: should failure to read a file/resource be an error? urlXXX: read in a URL (java.net?). even better, an anubis reference :)
Started this; using Hadoop's own logic to load files as that will do the inline variable expansion the default files require.
Also consider: 1. adding a validate option that does some validation of required attributes 2. add a list of required attributes which are resolved; if not local are set in place here. |
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
|
Sub-task
Closed
Major
This would let us
-keep up to date with hadoop default.xml changes
-import site specific settings, without having to re-enter stuff.