Much as it pains me -- what happens if you ungenerify (just use Map instead
of Map<String, Integer>)?

On Fri, Jan 14, 2011 at 3:06 PM, Jonathan Coveney <[email protected]>wrote:

> I did a search online, and while someone had the same error, I don't think
> it was related. From the error log, I see this...
>
> Caused by: java.lang.RuntimeException: Final function of
> squeal.map.makecountalgq is not of the expected type.
>        at org.apache.pig.EvalFunc.<init>(EvalFunc.java:146)
>        at squeal.map.makecountalgq.<init>(makecountalgq.java:32)
>        at sun.reflect.NativeConstructorAccessorImpl.newInstance0(Native
> Method)
>        at
>
> sun.reflect.NativeConstructorAccessorImpl.newInstance(NativeConstructorAccessorImpl.java:39)
>        at
>
> sun.reflect.DelegatingConstructorAccessorImpl.newInstance(DelegatingConstructorAccessorImpl.java:27)
>        at java.lang.reflect.Constructor.newInstance(Constructor.java:513)
>        at java.lang.Class.newInstance0(Class.java:355)
>        at java.lang.Class.newInstance(Class.java:308)
>        at
> org.apache.pig.impl.PigContext.instantiateFuncFromSpec(PigContext.java:472)
>
> But here is my udf (I've truncated a bit, but it works if it isn't
> algebraic)
>
> public class makecountalgq extends EvalFunc<Map<String,Integer>> implements
> Algebraic, Accumulator<Map<String,Integer>> {
>        static TupleFactory mTupleFactory = TupleFactory.getInstance();
>        static BagFactory mBagFactory = BagFactory.getInstance();
>        private MapWrapper ourWrap = new MapWrapper();
>
>        public String getInitial() { return Initial.class.getName(); }
>
>        public String getIntermed() { return Intermed.class.getName(); }
>
>        public String getFinal() { return Final.class.getName(); }
>
>        //An initial tuple, crates partial results
>        public static class Initial extends EvalFunc<Tuple> {
>                public Tuple exec(Tuple input) throws IOException {
>                        MapWrapper aWrap = new MapWrapper(new
> HashMap<String,MutableInt>());
>                        DataBag bag = (DataBag)input.get(0);
>                        aWrap.addSBag(bag);
>                        return mTupleFactory.newTuple(aWrap);
>                }
>        }
>        //Called on the result of the Initial, or on another intermediate
>        public static class Intermed extends EvalFunc<Tuple> {
>                public Tuple exec(Tuple input) throws IOException {
>                        return
> mTupleFactory.newTuple(mergeWrappers((DataBag)input.get(0)));
>                }
>        }
>        //Can take the tuple that initial or intermed creates
>        public static class Final extends EvalFunc<Map<String,Integer>> {
>                public Map<String,Integer> exec(Tuple input) throws
> IOException {
>                        return
> mergeWrappers((DataBag)input.get(0)).convertMap();
>                }
>        }
>
>        //takes a bag of MapWrappers and merges them
>        public static MapWrapper mergeWrappers(DataBag bag) throws
> IOException {
>                Iterator<Tuple> bagIt = bag.iterator();
>                MapWrapper aWrap = (MapWrapper)bagIt.next().get(0);
>                while (bagIt.hasNext())
>                        aWrap.merge((MapWrapper)bagIt.next().get(0));
>                return aWrap;
>        }
> }
>
> Any idea why I would get this error? It compiles fine, and if I even just
> take out Algebraic, I can use the UDF (it just uses the accumulator). I
> don't see how the type is off...
>
> Any ideas? (in case the context matters: this is a script that takes a list
> of (a, b:chararray), and gives you, for a given key, a map from b-># of
> times b occured.
>
> Here is how I invoke it:
>
> register squeal.jar;
> A = LOAD 'test.txt' AS (a:chararray, b:chararray);
> B = GROUP A BY a;
> C = FOREACH B GENERATE group, squeal.map.makecountalgq(A.b) as mapz;
>
> And I get the error.
>
> Thanks in advance. I guess I could content myself with an Accumulator but
> the Algebraic solution is so close... a MapWrapper is a class I created
> that
> lets me pass around Maps and whatnot.
>

Reply via email to