I'm trying to load files to pig while use python udf, i've tried two ways:
• (myudf1, sample1.pig): try to read the file from python, the file is located on my client server.
• (myudf2, sample2.pig): load file from hdfs to grunt shell first, then pass it as a parameter to python udf.
myudf1.py
from __future__ import with_statement
def get_words(dir):
stopwords=set()
with open(dir) as f1:
for line1 in f1:
stopwords.update([line1.decode('ascii','ignore').split("\n")[0]])
return stopwords
stopwords=get_words("/home/zhge/uwc/mappings/english_stop.txt")
@outputSchema("findit: int")
def findit(stp):
stp=str(stp)
if stp in stopwords:
return 1
else:
return 0
sample1.pig:
REGISTER '/home/zhge/uwc/scripts/myudf1.py' USING jython as pyudf;
item_title = load '/user/zhge/data/item_title_sample/000000_0' USING PigStorage(',') AS (title:chararray);
T = limit item_title 1;
S = FOREACH T GENERATE pyudf.findit(title);
DUMP S
I get: IOError: (2, 'No such file or directory', '/home/zhge/uwc/mappings/english_stop.txt')
For solution 2:
myudf2:
def get_wordlists(wordbag):
stopwords=set()
for t in wordbag:
stopwords.update(t.decode('ascii','ignore'))
return stopwords
@outputSchema("findit: int")
def findit(stopwordbag, stp):
stopwords=get_wordlists(stopwordbag)
stp=str(stp)
if stp in stopwords:
return 1
else:
return 0
Sample2.pig
REGISTER '/home/zhge/uwc/scripts/myudf2.py' USING jython as pyudf;
stops = load '/user/zhge/uwc/mappings/stopwords.txt' AS (stop_w:chararray);
-- this step works fine and i can see the "stops" obejct is loaded to pig
item_title = load '/user/zhge/data/item_title_sample/000000_0' USING PigStorage(',') AS (title:chararray);
T = limit item_title 1;
S = FOREACH T GENERATE pyudf.findit(stops.stop_w, title);
DUMP S;
Then I got: ERROR org.apache.pig.tools.grunt.Grunt -ERROR 1066: Unable to open iterator for alias S. Backend error : Scalar has more than one row in the output. 1st : (a), 2nd :(as