This is mostly only useful, of course, where you are using VIW many times. If you are simply encoding a single integer, as for a key, the difference is overwhelmed by the general Hadoop overhead.
On Thu, Jun 20, 2013 at 3:48 PM, Sean Owen <[email protected]> wrote: > "Var" uses a variable length encoding. It will be much smaller for > small integer values. If your values were roughly uniformly > distributed over all ints/longs, then it would actually result in a > longer encoding on average. It's useful for things like indices and > offsets but not hash values. > > On Thu, Jun 20, 2013 at 3:24 PM, Dan Filimon > <[email protected]> wrote: > > Hi! > > > > Can anyone explain what the difference between IntWritable and > > VarIntWritable is? > > When should I use each? >
