Github user mengxr commented on the pull request:
https://github.com/apache/spark/pull/7258#issuecomment-119761323
@zhangjiajin The issue with method 2 is projection before filtering. It may
increase the shuffle size. After we generate possible prefixes, we should count
their frequencies (that does not require shuffling the suffixes), filter out
invalid candidates, and then project. Basically, I'm suggesting the following:
1. generate prefix candidates
2. count their frequencies
3. filter and collect frequent ones
4. Do we have enough candidates to distribute the work? If no, go to 1 and
generate candidates with length + 1.
5. project and group by prefixes
6. generate frequent patterns for each prefix
7. union with local frequent sequences
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
