On Wed, Sep 13, 2017 at 1:56 PM, Eugene Kirpichov
<[EMAIL PROTECTED]lid> wrote:

Sure, one could implement this in Python, but why? It has few, if any,
advantages in Python, but does have several disadvantages to the way
side inputs are currently handled. The only reason to do so would be
because it works better for the Java api.

I was imagining something more like a TupleTag.output(...) which would
affect that ParDo only rather than appending to an existing
PCollection from anywhere (which is a really big change, but could
also be useful for dead-letter stuff). Let's table that for now.

Shouldn't block indefinitely, but I think it's quite relevant and
should be taken into consideration.
Too bad these issues didn't come up when NewDoFn was being proposed...
It could also be an argument for keeping ProcessContext around long
term. Maybe that's the answer to these "how do I pass this pile of
stuff to a user's lambda" issues as well.

NEW: Monitor These Apps!
elasticsearch, apache solr, apache hbase, hadoop, redis, casssandra, amazon cloudwatch, mysql, memcached, apache kafka, apache zookeeper, apache storm, ubuntu, centOS, red hat, debian, puppet labs, java, senseiDB