Github user JamesRTaylor commented on the issue:

    https://github.com/apache/phoenix/pull/295
  
    Thinking more, I think it’s fine to go through the CREATE TABLE code path 
and move the logic in the try block of init down into ensureTableCreated. 
It’s only called once on the first connection made to a cluster from a 
client. Keeping the code simple and doing a single RPC to get the metadata and 
another to get the version will outweigh the minor overhead of compiling CREATE 
TABLE (plus like you mentioned before, we need to do that to build up the args 
for ensureTableCreated anyway).


---

Reply via email to