Query impala directly from Python. Note that doesn't do things like query cancellation that impala-shell actually does.

gistfile1.txt

#!/usr/bin/python
# Apache License
# 
# $python impala.py 'select "hello", "there"'
# ['hello\tthere']


import sys
import os

BATCH=1000

def query(query, host='localhost', port=21000):
  from beeswaxd import BeeswaxService
  from beeswaxd.BeeswaxService import QueryState
  from ImpalaService import ImpalaService
  from ImpalaService.ImpalaService import TImpalaQueryOptions
  from ImpalaService.constants import DEFAULT_QUERY_OPTIONS
  from Status.ttypes import TStatus, TStatusCode
  from thrift.transport.TSocket import TSocket
  from thrift.transport.TTransport import TBufferedTransport, TTransportException
  from thrift.protocol import TBinaryProtocol
  from thrift.Thrift import TApplicationException

  sock = TSocket(host, port)
  transport = TBufferedTransport(sock)
  transport.open()
  protocol = TBinaryProtocol.TBinaryProtocol(transport)
  service = ImpalaService.Client(protocol)
  service.PingImpalaService()
  bq = BeeswaxService.Query()
  bq.query = query
  bq.options = []

  handle = service.query(bq)
  while True:
    state = service.get_state(handle)
    if state == QueryState._NAMES_TO_VALUES["FINISHED"]:
      break
    elif state == QueryState._NAMES_TO_VALUES["EXCEPTION"]:
      raise Exception()
    else:
      time.sleep(0.5)
  ret = []
  while True:
    results = service.fetch(handle, False, BATCH)
    ret.extend(results.data)
    if not results.has_more:
      return ret

def adjust_sys_path():
  real = os.path.realpath("/usr/bin/impala-shell")
  comps = real.split("/")
  comps = comps[0:-2]
  comps.append("lib")
  comps.append("impala-shell")
  sys.path.append(os.path.join(*(["/"] + comps + ["gen-py"])))
  sys.path.append(os.path.join(*(["/"] + comps + ["lib"])))

if __name__ == "__main__":
  adjust_sys_path()
  print query(sys.argv[1])