the job repository is used indirectly for passing data between steps (Jean-Philippe is right that the best way to do that is to put data into the StepExecutionContext
and then use the verbosely named ExecutionContextPromotionListener
to promote the step execution context keys to the JobExecutionContext
.
It's helpful to note that there is a listener for promoting JobParameter
keys to a StepExecutionContext
as well (the even more verbosely named JobParameterExecutionContextCopyListener
); you will find that you use these a lot if your job steps aren't completely independent of one another.
Otherwise you're left passing data between steps using even more elaborate schemes, like JMS queues or (heaven forbid) hard-coded file locations.
As to the size of data that is passed in the context, I would also suggest that you keep it small (but I haven't any specifics on the
与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…