[
https://issues.apache.org/jira/browse/PIG-349?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12620922#action_12620922
]
sms edited comment on PIG-349 at 8/8/08 5:25 AM:
-----------------------------------------------------------------
After thinking for a while, I realized that I had implemented this in the
parser and the patch is probably not required. The unit test case for the bag
schema as it exists today is at the end of this section.
The bag schema can be expressed as shown below. The idiosyncrasy here is the
naming of the tuple inside the bag which is not required as the tuple can never
be accesses directly.
{code}
c = load 'a' as (name, details: bag{mytuple: tuple(age: int, gpa)});
c = load 'a' as (name, details: bag{mytuple (age: int, gpa)});
{code}
Going back to the bag schema that failed due to the patch, it should work as
mentioned in comment 2:
{code}
A = load 'foo' as (B: bag{T: tuple(I: int)});
{code}
{code}
@Test
public void testQuery64() {
buildPlan("a = load 'a' as (name: chararray, details: tuple(age, gpa),
mymap: map[]);");
buildPlan("c = load 'a' as (name, details: bag{mytuple: tuple(age: int,
gpa)});");
buildPlan("b = group a by details;");
String query = "d = foreach b generate group.age;";
buildPlan(query);
buildPlan("e = foreach a generate name, details;");
buildPlan("f = LOAD 'myfile' AS (garage: bag{tuple1:
tuple(num_tools: int)}, links: bag{tuple2: tuple(websites: chararray)}, page:
bag{something_stupid: tuple(yeah_double: double)}, coordinates:
bag{another_tuple: tuple(ok_float: float, bite_the_array: bytearray,
bag_of_unknown: bag{})});");
}
@Test
public void testQueryFail64() {
String query = "foreach (load 'myfile' as (col1, col2 : bag{age: int}))
generate col1 ;";
try {
buildPlan(query);
} catch (AssertionFailedError e) {
assertTrue(e.getMessage().contains("Exception"));
}
}
{code}
was (Author: sms):
After thinking for a while, I realized that I had implemented this in the
parser and the patch is probably not required. The unit test case for the bag
schema as it exists today is at the end of this section.
The bag schema can be expressed as shown below. The idiosyncrasy here is the
naming of the tuple inside the bag which is not required as the tuple can never
directly.
{code}
c = load 'a' as (name, details: bag{mytuple: tuple(age: int, gpa)});
c = load 'a' as (name, details: bag{mytuple (age: int, gpa)});
{code}
Going back to the bag schema that failed due to the patch, it should work as
mentioned in comment 2:
{code}
A = load 'foo' as (B: bag{T: tuple(I: int)});
{code}
{code}
@Test
public void testQuery64() {
buildPlan("a = load 'a' as (name: chararray, details: tuple(age, gpa),
mymap: map[]);");
buildPlan("c = load 'a' as (name, details: bag{mytuple: tuple(age: int,
gpa)});");
buildPlan("b = group a by details;");
String query = "d = foreach b generate group.age;";
buildPlan(query);
buildPlan("e = foreach a generate name, details;");
buildPlan("f = LOAD 'myfile' AS (garage: bag{tuple1:
tuple(num_tools: int)}, links: bag{tuple2: tuple(websites: chararray)}, page:
bag{something_stupid: tuple(yeah_double: double)}, coordinates:
bag{another_tuple: tuple(ok_float: float, bite_the_array: bytearray,
bag_of_unknown: bag{})});");
}
@Test
public void testQueryFail64() {
String query = "foreach (load 'myfile' as (col1, col2 : bag{age: int}))
generate col1 ;";
try {
buildPlan(query);
} catch (AssertionFailedError e) {
assertTrue(e.getMessage().contains("Exception"));
}
}
{code}
> discrepancy in bags representation
> ----------------------------------
>
> Key: PIG-349
> URL: https://issues.apache.org/jira/browse/PIG-349
> Project: Pig
> Issue Type: Bug
> Affects Versions: types_branch
> Reporter: Olga Natkovich
> Assignee: Santhosh Srinivasan
> Priority: Critical
> Fix For: types_branch
>
> Attachments: bag_schema.patch
>
>
> Currently, when I describe a bag in AS clause of the load statement, I can
> place bag of integers there. However, when describing constant bags, I can
> only create a bag of tuples that contain integers.
> My understanding is that at this time we only support bags of tuples. If
> that's the case, AS clause needs to match that
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
