I think I made a small mistake, iot should be result = FOREACH ad_group GENERATE udf.name(Actual_data, cd_group.Calculation-data);
Basically the group all make a bag of one tuple, whgich is a null key and a bag of all elements. I think you still have to explicitly refer to the bag of elements, but a describe cd_group; would tell you Sent via BlackBerry -----Original Message----- From: <deepak....@wipro.com> Date: Thu, 13 Jan 2011 18:56:16 To: <user@pig.apache.org> Reply-To: user@pig.apache.org Subject: RE: Pass an Outer Bag into UDF Hi Jonathan, Thanks for your response. How would the PIG Statement look? Building upon my previous example, ------------ Calculation-data = LOAD /path/to/data AS (field:int); Actual_data = LOAD /path/to/data AS (field1:int, field2:int); -- I want to do something like this: Result = FOREACH Actual_data GENERATE udf_file_name.fun_name(Calculation_data, field1, field2); -- But it's not legal. ------------- There is an example in the UDFScriptingwithPython page on the PIG site, where an internal bag is being converted to an outer bag. But that doesn't seem to be very relevant here, except for the point that if I somehow pass an outer or inner bag into the UDF, I can work upon it like any other python list. Thanks, Deepak -----Original Message----- From: Jonathan Coveney [mailto:jcove...@gmail.com] Sent: Thursday, January 13, 2011 6:31 PM To: user@pig.apache.org Subject: Re: Pass an Outer Bag into UDF Woops, look at me, I didn't realize you were passing it to python. The same should work, just...in python. The bags will be lists, the input a tuple of those two bag lists. 2011/1/13 Jonathan Coveney <jcove...@gmail.com> > You absolutely can do this, I'm just not sure if you can do it using > the accumulator interface (I THINK you can, as I think it ratchets > only the first tuple input, and passes the entire second one, but am > not sure, someone else can weigh in). If you CAN do it with the > accumulator interface though I highly recommend it, as it's more memory > efficient. > > Basicaly, you'll have this as your pig script: > > public class doublebag { > public <OutputType> exec(Tuple input) throws IOException { > DataBag innerBag = (DataBag)input.get(0); > DataBag outerBag = (DataBag)input.get(1); > Iterator<Tuple> ibit = innerBag.iterator(); > while (ibit.hasNext()) { > Tuple ibelem = ibit.next(); > Iterator<Tuple> obit = outerBag.iterator(); > while (obit.hasNext()) { > Tuple obelem = obit.next(); > } > } > } > } > > Obviously, you need to do things like checking for empty input etc > etc, this is just a rough rough example of the code. The point is, if > you just do exec, you'll simply have a bag as your first input and a bag as > the second. > > And then in your code if you want to do one entire bag against > another, you'd just do a group thing all; and pass it that thing. > > Hope that explanation made sense, if it didn't just ask again. It's > worth going over the bag->bag example in the UDF manual, and really, > you just have > 2 bag inputs instead of one. > > 2011/1/13 <deepak....@wipro.com> > > Hi, >> >> I wish to pass an outer bag into a Python UDF. >> >> Something like: >> >> Calculation-data = LOAD /path/to/data AS (field;int); Actual_data = >> LOAD /path/to/data AS (field1:int, field2:int) >> >> >> Calculation_data is not a very big bag. Maybe about 500 tuples in all >> - a single file. >> Actual_data is the real data source lying on HDFS. >> >> For each tuple in Actual_data, I wish to have the entire >> Calculation_data (whole bag) for some calculation that I wish to do. >> >> So, with every call to a UDF, I need to pass this bag along with >> tuple from Actual_data. >> >> Simply, is there any way of passing an outer bag into a Python UDF? >> >> Regards, >> Deepak >> Please do not print this email unless it is absolutely necessary. Please do not print this email unless it is absolutely necessary. The information contained in this electronic message and any attachments to this message are intended for the exclusive use of the addressee(s) and may contain proprietary, confidential or privileged information. If you are not the intended recipient, you should not disseminate, distribute or copy this e-mail. Please notify the sender immediately and destroy all copies of this message and any attachments. WARNING: Computer viruses can be transmitted via email. The recipient should check this email and any attachments for the presence of viruses. The company accepts no liability for any damage caused by any virus transmitted by this email. www.wipro.com
