I still think XML is a better format. But if you do allow property files, consider first an XSD file that converts XML to properties. Because if you can accomplish that, you will have proven to yourself that the property file can represent everything an XML file can. On Jun 8, 2014 2:00 PM, "Matt Sicker" <boa...@gmail.com> wrote:
> I'm only working on this because it sounds interesting and has been > requested by several people. I personally never use this file format in > Log4j 1, so I'm not entirely sure on how to best maintain compatibility or > similarity to the old format. > > The technical side of parsing a flat properties map into a tree of Nodes > isn't that difficult. I'm sure we all took data structures at some point in > our lives ;) > > Due to the limitations of properties files, the format has to be slightly > different than the usual hierarchy used in all the other formats. The key > difference I'd say is that instead of the "name" attribute used on all the > appenders and loggers, the name would be the child "node" of the appenders > element. For instance: > > appenders.Name.attribute = ... > appenders.Name.anotherAttribute = ... > > Of course, the keys would be converted to lower case for case > insensitivity (which makes me think we could really use a > CaseInsensitiveHashMap or something). > > The old format uses something more like: > > log4j.appender.Name.attribute = ... > > For consistency, I think this should be appenders, and we could use > "log4j2" as the prefix (or even "configuration" for ultimate consistency). > Thoughts? > > -- > Matt Sicker <boa...@gmail.com> >
