Pinot bugs and making it work with the latest trunk pinot

[agrawaldevesh]

Three changes here: 1 medium and 2 tiny:

• The nameToIndex used for the pinot column ordering map should not assume that the variable reference name is the same as the pinot lowercase name. It should be keyed off the keys in the selections map (in the pinot query generator context) and not the value. It was throwing an error with the query below because the output variables were named like player_initial_33, g_old_51 etc instead of just player_initial, g_old. This is because these columns are referenced twice in the query.
  This again smells like a core bug and it is totally my bad that I screwed this up the last time.

select * from (select yearid, substr(playerid, 1, 3) as player_initial, sum(homeruns) from pinot.baseballStats.baseballStats group by 1, 2) as m INNER join (select g_old, substr(playername, 1, 3) as player_name_initial, sum(numberofgames) from pinot.baseballStats.baseballStats group by 1, 2) as w ON m.player_initial = w.player_name_initial;

• [tiny] Make this work with the new pinot trunk code which no longer likes "Content-Type: Application/json" for just GET requests and non body POSTS :-). As it shouldn't.
• [tiny] renamed the config prefer broker queries to ! forbid broker queries to be true to the name

== RELEASE NOTES ==

Pinot Changes
*  Replace config `pinot.prefer-broker-queries` with the inverse config `pinot.forbid-broker-queries`.

[highker]

First commit: could you remove the "[Pinot]" flag from the commit title?

[highker]

second commit: could you update the title to be:
Rename preferBrokerQueries to its negate format forbidBrokerQueries

[highker]

3rd commit: could you update the title to be
Avoid passing Content-Type header when Pinot request body is empty

[highker]

Last commit: could you update the title to be

Fail with informative error message if no Pinot expression is selected

[agrawaldevesh]

James, thanks for the comments and careful review. As always !.

[highker]

lgtm