Python – Singles with python transformation: “cannot recognize input near ‘transform'” error

Singles with python transformation: “cannot recognize input near ‘transform'” error… here is a solution to the problem.

Singles with python transformation: “cannot recognize input near ‘transform'” error

I have a Hive table that tracks the state of objects as they move through various stages of the process. The table is as follows:

hive> desc journeys;
object_id           string                                      
journey_statuses    array<string>

Here is a typical logging example:

12345678    ["A","A","A","B","B","B","C","C","C","C","D"]

The

records in the table are generated using collect_list of Hive 0.13, and the states are ordered (I’ll use collect_set if the order isn’t important object_id).

I wrote a quick Python script that reads from standard input:

#!/usr/bin/env python
import sys
import itertools

for line in sys.stdin:
    inputList = eval(line.strip())
    readahead = iter(inputList)
    next(readahead)
    result = []
    for id, (a, b) in enumerate(itertools.izip(inputList, readahead)):
        if id == 0:
          result.append(a)
        if a != b:
          result.append(b)
    print result

I plan to use it in the Hive transform call. It seems to work when running locally :

$ echo '["A","A","A","B","B","B","C","C","C","C","D"]' | python abbreviate_list.py
['A', 'B', 'C', 'D']

However, when I add a file and try to execute in Hive, an error is returned:

hive> add file abbreviateList.py;                                                                           
Added resource: abbreviateList.py

hive> select
    >   object_id,
    >   transform(journey_statuses) using 'python abbreviateList.py' as journey_statuses_abbreviated
    > from journeys;
NoViableAltException( ... wall of Java error messages ... )
FAILED: ParseException line 3:2 cannot recognize input near 'transform' '(' 'journey_statuses' in select expression

Can you see what I’m doing wrong?

Solution

Obviously, you can’t select other fields that aren’t in the transformation (object_id in your example). This other SO question seems to indirectly solve this problem:

How can select a column and do a TRANSFORM in Hive?

Theoretically, you can modify Python to accept object_id as an input parameter and make it a passthrough to another output field when you need to include it in the output.

Related Problems and Solutions