YahooArchive
Switch to Maven
Before Howl development progresses too far, can we consider switching to Maven? Hadoop is trying to make the switch, as are other projects. Perhaps we can save ourselves pain down the line by making the switch as soon as possible?
+1 to move to Maven. Though I myself have used Maven on a very limited occasion. I have heard people praising it elaborately over ant based system. Having said that it might be a complicated project because we heavily rely on Hive build system (which in itself is pretty extensive) and is based on ant. Deal here will be to make ant (used in hive) and maven (used for howl build) happy in presence of each other.
I imagine it will depend quite a bit on how Howl's directories are organized. My understanding is that the goal is to make Hive (7) a direct dependency, which would correct the very odd Howl = Hive source + new directories layout we currently have. Is this correct/
That is the "long-term" goal of just having howl code and using hive's jars as dependencies. Major reason that we have not done that is whenever Howl requires some code changes from Hive (usually metastore/ dir) it takes time to get that committed into official hive repo. Moreover, it may so happen that hive doesnt want our patch. In all these cases we need to have a separate hive codebase which contains the changes Howl wants. So, having hive sources with us is an advantage in that regard. Agreed, it is a burden if we need to maintain that code separately. Till now, we have tried very hard to get everything we need coming in through official hive repo. But depending on just jars we will give up on the flexibility of changing sources if we absolutely have to.
Perhaps we can take the opportunity to do the maven conversion during the time of separating out the Hive code to prep Howl for incubation. It's probably just a bit more work. Speaking of which, is this work ongoing?
Sorry, I meant is the work to disentangle Howl from Hive in order to prep for incubator ongoing?