[
https://issues.apache.org/jira/browse/CALCITE-4770?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17414583#comment-17414583
]
Julian Hyde commented on CALCITE-4770:
--------------------------------------
I don't see this as a 'service'. I'd like to see this as an API or data
structure, and how it fits within the existing planning process.
How does this relate to Lattice?
"TableScan operator size" was confusing. I think you mean "The number of
TableScan operators in the plan".
To me, this looks like a Comparator or Ordering, described thus: "When
comparing two materialized views, prefer the one that has fewer TableScan
operators; if that is equal, prefer the one with fewer Calc operators; ...".
And given that ordering, maybe you have in mind a data structure (an "index")
that can quickly find the best MV.
But the other important part of MV matching -- the most important part -- is
that the MV needs to be able to satisfy the query. How do you plan to include
that among other criteria?
> Design of SearchMvService
> -------------------------
>
> Key: CALCITE-4770
> URL: https://issues.apache.org/jira/browse/CALCITE-4770
> Project: Calcite
> Issue Type: New Feature
> Reporter: xzh_dz
> Priority: Major
>
> In the data system, materialized view can speed up SQL query. Materialized
> view represents the pre-calculated SQL logic of the target table and contains
> real data. In the project, we use materialized views to speed up SQL queries,
> use the materialized view recognition technology to identify materialized
> views, and enhance the ability of materialized recognition. However, with
> more and more materialized views , materialized view recognition takes a lot
> of time, which seriously affects the performance of SQL queries. With this
> problem, we propose a design of accurate search materialized view based on
> operator features.
> DOC：[Design of
> SearchMvService|[https://docs.google.com/document/d/1mmAsK_uW-fBs893JERP1gspMurX2lNXcVGeesP4XUqQ/edit]]
> Welcome to discuss and expect more feedback,thank you.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
