Hi all, I'm working with Cloudera Hive for the first time, seen that my company chose to use this database. I received the JDBC connectionString and I created a little method in java. Here the script public String[] getEmployee(){
String[] idcod ={"ABC123", "FLS163","XYZ001","PLE456", "ERV021", and so on......};
String[] names= new String[idcod.length];
int ncount=0;
long start1, elapsedTimequery;
try{
Statement s = conn.createStatement();
System.out.println("Fetch size: "+s.getFetchSize());
ResultSet r=null;
for(int i=0; i<idcod.length; i++) {
String query="";
query+=" SELECT code, surname, name FROM mytable where code='"+idcod[i]+"'";
ncount=0;
start1 = System.nanoTime();
r = s.executeQuery(query);
elapsedTimequery = System.nanoTime() - start1;
System.out.println((double)(elapsedTimequery/1000/1000)/1000+" seconds");
while(r.next()){
names[i]=r.getString("surname") + " " +r.getString("name");
ncount++;
}
System.out.println(ncount+"--"+idcod[i]);
}
r.close();
s.close();
}
catch(SQLException sqlex){
sqlex.printStackTrace();
}
catch(Exception ex){
ex.printStackTrace();
}
return names;
} For each query r = s.executeQuery(query) it takes more of 4 seconds to search data and got it, meanwhile with Oracle a query takes about 2 milliseconds with Fetch "10". With technical referent tried ICEBERG Table but it's no change. I tried to add "hive.fetch.task.conversion" to my connectionString with all possible values (none, minimal, more) with no change. I can't use clause IN because my main program manage the codes row by row. How can I avoid all this time for an hive query and to go near to Oracle timing? Thank you to all.
... View more
