We can document the new API from the beginning, should we still document the old API?
Good question... There is some documentation on the wiki at present, which is more than has ever existed. I am for thoroughly documenting the new api and leaving the old as is.
Should we include the user tutorials right inside the source control ?
We can use document generators such as AsciiDoc (http://www.methods.co.nz/asciidoc/index.html)
Its already being used in other Hadoop subproject ( Sqoop) You can see the link for document generated for it
I'd prefer we use the wiki:
as it's easy to update and we have other docs there. However, I could be persuaded otherwise if there was a compelling reason to move.
It might be beneficial to include some portion of the documentation to the release artifacts as well. I'm expecting that wiki will probably contain only one documentation for most newer version. But there might be users out there that will be bound to some specific release and it will be difficult for them to find correct source of information.
Created wiki page at https://cwiki.apache.org/confluence/display/MRUNIT/MRUnit+Tutorial
Added tutorial till test of counters
Please give review comments as i write the tutorial
Great work! Two comments:
1) MapReduceDriver doesn't seem to be used
2) The tabs are messing up the formatting a little. Hadoop related projects typically use two spaces as opposed to a tab.
Thank you very much!
Jim created an awesome new website: http://incubator.apache.org/mrunit/ I wonder if we should move the Tutorial there?
Jagat, since this issue is still open, do you mind if I work on it? I did a presentation at the Chicago Hadoop Users Group last month and have a lot of example code I could use for this tutorial.
I just wanted to post a comment to say that I intend to finish this soon (hopefully within a couple of weeks), and now that MRUnit 1.0.0 has been released, I will make sure that I update it to describe the new version.