Minor Description
Here is a correct script:
user_hashes = LOAD 'users.tsv' AS (info:map[], fullref:chararray, shortref:chararray); -- [userid#123,username#jerry,followers#192] name username -- [userid#987,username#george,followers#31] id userid -- [userid#568,username#elaine,followers#40] name followers users = FOREACH user_hashes GENERATE info#'userid' AS userid:chararray, info#'username' AS username:chararray; DUMP users; -- (123,jerry) -- (987,george) -- (568,elaine)
Omitting the quotes on the key dereference gives a very unhelpful error message.
users = FOREACH user_hashes GENERATE info#userid AS userid:chararray; -- 400 ERROR: ERROR 1200: <file ./foo.pig, line 8, column 42> [...] mismatched input 'userid' expecting set null
It may be that the user forgot the quotes, or may instead be assuming that Pig allows dereferencing a map by the value of an alias or expression:
users = FOREACH user_hashes GENERATE info#'username', -- works info#username, -- need quotes around literal info#fullref, -- no, can't use an alias' value to deref info#(CONCAT('user',shortref)) -- and can't use an expression to deref ;
The error would be better off reading
Values may only be retrieved from a map by using a literal chararray key. Did you mean << info#'userid' >>? See http://pig.apache.org/docs/r0.12.0/basic.html#map
----------------------
Forgetting to attach the dummy square brackets on a map schema gives another confusing error message:
user_hashes = LOAD 'users.tsv' AS (id:chararray, profile:map); -- 390 ERROR: ERROR 1200: <file ./foo.pig, line 1, column 60> [...] mismatched input ')' expecting LEFT_BRACKET
This message should read something like
A map type must supply (within square brackets) the schema for its values; or supply empty square brackets if it holds generic values. Did you mean << profile:map[chararray] >> or << profile:map[] >>? See http://pig.apache.org/docs/r0.12.0/basic.html#map-schema