[
https://issues.apache.org/jira/browse/PIG-20?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12545942
]
Alan Gates commented on PIG-20:
-------------------------------
Mostly looks good. There are a few issues that need addressed:
1) Data in the provided unit tests is already sorted before it is fed to the
sort tests. It should randomized to assure we're really sorting the data and
not merely passing it through.
2) In EvalSpec, the comparator can be instantiated either via a call to
instantiateFunc or getComparator. But in getComparator there is no way to
instantiate the user defined comparator. It depends instead on instantiateFunc
having been called, otherwise it returns the default comparator. This doesn't
seem right. This means there is a possible code path where the user would
provide a comparator but the default one would still get used. If we cannot
guarantee that instantiateFunc has always been called before getComparator,
then getComparator needs to be changed to take a FunctionInstantiator so it can
create the user defined function if necessary. If we can guarantee that
instantiateFunc _should_ always be called before getComparator, then
getComparator should throw a RuntimeException if the member variable comparator
is null.
3) Nitpicky code formatting issues:
a) The pig coding standard is now to use spaces instead of tabs. In the
existing files you were correct to stay with standard of the surrounding code,
but in the new files that were created spaces should be used.
b) Lines shouldn't exceed 80 characters in length when possible.
> Sorting using custom comparison functions
> ------------------------------------------
>
> Key: PIG-20
> URL: https://issues.apache.org/jira/browse/PIG-20
> Project: Pig
> Issue Type: New Feature
> Components: impl
> Reporter: Olga Natkovich
> Attachments: usercompare.patch
>
>
> Currently, onlu string based sorting is supported. Once we have types,
> numeric sort will be supported as well. However, soem users express need for
> custome comparison functions for sort.
> Alan put together a design document for this:
> http://wiki.apache.org/pig/UserDefinedOrdering
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
