Minor Description
(Bug filed on 0.8. I do not have 0.9 to test.)
This applies to org.apache.pig.impl.builtin.ReadScalars.java:83
http://search-hadoop.com/c/Pig:/src/org/apache/pig/impl/builtin/ReadScalars.java
I have bitten myself with the same programming error several times, and each time I spent too long diagnosing my error.
The error message "scalar has more than one row in the output" is a bit misleading, considering the underlying programming mistake.
Consider this Pig script:
A = LOAD 'a' as (key, a1, a_junk);
B = LOAD 'b' as (key, b1, b_junk);
C = join A by key, B by key;
– Now, we want to project (key, a1, b1)
– CORRECT:
D_GOOD = foreach C generate A::key, a1, b1; – Disambiguate 'key' correctly.
– Now, consider some common programmer errors:
– INCORRECT:
– This fails, because 'key' is ambiguous. The error message is clear enough.
D_BAD_1 = foreach C generate key, a1, b1;
– This fails whenever A has multiple rows.
D_BAD_2 = foreach C generate A.key, a1, b1
– Error: "Scalar has more than one row in the output 1st : t1, 2nd : t2"
That's non-illuminating, for the following reason:
The error message is assuming that the programmer is making a semantic error, trying to use a value from the original A, which is impossible if A has more than one row.
In actuality, the programmer wants A::key, but he made a syntax error by typing "A.key", and it resulted in "scalar has more than one row" message that has nothing to do with what he intended.
Since he has confused "." and "::", he has no context for interpreting the message properly.
Ideally, the error message would say something like this:
"A.key cannot be used as scalar here, because A has more than one row. Did you mean A::key?"
If the identifiers are not available at error-logging time, something like this would be helpful:
"Relation cannot be used as scalar here, because A has more than one row. Did you mean to use '::' instead of '.'? "
