Github user decibel commented on the issue:
https://github.com/apache/incubator-madlib/pull/78
Yeah, the 1GB limit is certainly a consideration. Using int4's for
everything, a composite of (srt, dest, weight) would be 13-16 bytes (3 * int4 =
12 + varlena and maybe alignment). With the worst case of 16 bytes, that would
be 65k edges, so that's the maximum that could be updated at once.
Before coding a full implementation, could you produce pseudocode of the
ideal minimum set-based implementation would be? In particular, I'm thinking
that {messages} is completely redundant, and if that's the case I suspect this
entire algorithm could be done in a recursive CTE.
---
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.
---
