I agree that if it were changed there may be some compatibility issues. But, if it's doable, then introducing a new property could bridge the change. Not saying it's doable, because I haven't looked, but a new property and a deprecation warning (in docs, I expect) would allow the change to happen. Very preliminary data showed me that parsing 1000 events slowed my parser from < 500 ms (w/o contextMap) to 2000 ms when each event contained 2 contextMap entries, requiring the list of maps to be converted to a single map. Not sure what the time would be to parse a multi-valued map, though, so I can't be sure of the overhead of walking the list wrapper.
On Tue, Jan 12, 2016 at 6:05 AM, Mikael Ståldal <[email protected]> wrote: > I think that the current JSONLayout format is unfortunate, and I would > prefer to have it as you propose. But we cannot change it now since that > will break backwards compatibility. > > See: https://issues.apache.org/jira/browse/LOG4J2-623 > > Perhaps GELFLayout would work better for you. > > On Mon, Jan 4, 2016 at 10:00 PM, Gary Gregory <[email protected]> > wrote: > >> The point I was trying to make is that you cannot describe what you are >> asking for with a generic XML schema, not sure about JSON schema, but the >> idea is the same. Since we use Jackson, that also means we use the same >> code to emit JSON and XML. >> >> Gary >> On Jan 4, 2016 12:25 PM, "Robin Coe" <[email protected]> wrote: >> >>> I can see that XML entities requires conforming to a schema but isn't >>> the writer implementation capable of wrapping the map entries when >>> required? Seems like it's making the JSON representation more complex (and >>> less performant) at the cost of some wrapper code for the xml writer. >>> >>> On Mon, Jan 4, 2016 at 3:19 PM, Gary Gregory <[email protected]> >>> wrote: >>> >>>> Yes, that is because we can define this kind of structure with XML/JSON >>>> schema with ease. >>>> >>>> Gary >>>> On Jan 4, 2016 11:55 AM, "Robin Coe" <[email protected]> wrote: >>>> >>>>> I was trying to deserialize a log event written by the JSONLayout >>>>> appender, which uses Jackson. I therefore also am using Jackson but with >>>>> the MrBeanModule, which is a POJO materializer. After much difficulty >>>>> with >>>>> Jackson throwing deserialization exceptions with the "contextMap" field, I >>>>> learned that the map is actually written out as a List of Maps (i.e. >>>>> List<Map<String,String>>. I've included one such event here, with >>>>> unnecessary fields shortened: >>>>> >>>>> >>>>> {"timeMillis":...,"thread":"...","level":"OFF","loggerName":"...","message":"...","endOfBatch":false,"loggerFqcn":"...","contextMap":[{"key":"LOGROLL","value":"com.xxx.xxx.handler.event.FailoverHandler"},{"key":"ROUTINGKEY","value":"elasticsearch-rollover"}]} >>>>> >>>>> I'm curious why the contextMap is represented as the more complex List >>>>> of single entry Maps, as opposed to a single multi-valued Map? So, >>>>> instead >>>>> of something that looks like: >>>>> >>>>> >>>>> {"contextMap":[{"key":"key1"},{"value":"value1"},{"key":"key2"},{"value":"value2"},...] >>>>> >>>>> I would expect the much simpler (and easily parseable): >>>>> {"contextMap":{"key1":"value1","key2":"value2",...}. >>>>> >>>>> Is this intended? >>>>> >>>>> Thanks, >>>>> Robin. >>>>> >>>> >>> > > > -- > [image: MagineTV] > > *Mikael Ståldal* > Senior software developer > > *Magine TV* > [email protected] > Grev Turegatan 3 | 114 46 Stockholm, Sweden | www.magine.com > > Privileged and/or Confidential Information may be contained in this > message. If you are not the addressee indicated in this message > (or responsible for delivery of the message to such a person), you may not > copy or deliver this message to anyone. In such case, > you should destroy this message and kindly notify the sender by reply > email. >
