Ben Roling created CRUNCH-204:
---------------------------------
Summary: MemPipeline.write() is inconsistent with
MemPipeline.read()
Key: CRUNCH-204
URL: https://issues.apache.org/jira/browse/CRUNCH-204
Project: Crunch
Issue Type: Bug
Affects Versions: 0.5.0
Reporter: Ben Roling
MemPipeline.read(AvroFileSource) of a file written with
MemPipeline.write(collection, AvroFileTarget) fails with java.io.IOException:
Not a data file.
I seems the way the file is written is inconsistent with the way it is read.
The appears to not actually be written out in Avro format. It seems
MemPipeline.write() simply does a toString() on each of the elements in the
collection and spits that out to the target's path with each element separated
by a newline.
Here is a simple test that demonstrates the issue:
{code}
final Pipeline memPipeline = MemPipeline.getInstance();
final String path = "persons";
final PCollection<Person> persons =
MemPipeline.collectionOf(Collections.singleton(new Person("John Doe")));
memPipeline.write(persons, new AvroFileTarget(path));
// throws IOException!
memPipeline.read(new AvroFileSource<Person>(new Path(path),
Avros.records(Person.class)));
{code}
The Person class in the example is is based on this simple Avro schema:
{code}
@namespace("org.foo.model")
protocol PersonProtocol {
record Person {
string name;
}
}
{code}
This is pretty confusing behavior. I ran into it trying to do some simple
testing and it took me longer than I'd like to admit to figure out what was
going on. I imagine others will run into it and be similarly confused.
I've left the priority as the default of Major although I suppose that point
could be argued. Reset it as you like.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
