[jira] [Created] (CALCITE-4039) Missing LATERAL keyword after sql validate

2020-06-02 Thread YufeiLiu (Jira)
YufeiLiu created CALCITE-4039: - Summary: Missing LATERAL keyword after sql validate Key: CALCITE-4039 URL: https://issues.apache.org/jira/browse/CALCITE-4039 Project: Calcite Issue Type: Bug

Re: Using indexes rather than table scans with Calcite

2020-06-02 Thread Haisheng Yuan
> using this approach, bitmap indexes may be represented as set operations > (union, intersect) over idx_a and idx_b followed by joining with scan_A > using 'rowId'. If I didn't get it wrong, joining with scan_A means it has to read all the tuples from table A, this will make index meaningless,

Re: [DISCUSS] Towards Calcite-Avatica 1.17.0

2020-06-02 Thread Francis Chuang
Hey everyone, Just a reminder that I'd like to make Avatica 1.17.0 rc0 available for voting soon. However, I am hoping we can review or merge a few open PRs for this release. If anyone has any spare cycles to review or merge any of the open PRs on Github, it'd be highly appreciated!

[jira] [Created] (CALCITE-4038) Refactor RexVisitor, RexBiVisitor, RelOptUtil.InputFinder

2020-06-02 Thread Julian Hyde (Jira)
Julian Hyde created CALCITE-4038: Summary: Refactor RexVisitor, RexBiVisitor, RelOptUtil.InputFinder Key: CALCITE-4038 URL: https://issues.apache.org/jira/browse/CALCITE-4038 Project: Calcite

Re: Using indexes rather than table scans with Calcite

2020-06-02 Thread Roman Kondakov
Vladimir, Haisheng, thank you for sharing your thoughts. And special thanks to Haisheng for awesome examples. I found Julian's reasoning about representing bitmap indexes as joins very deep and interesting. As I understand this idea right, for table A(a,b,c) indexes over columns, say, 'a' and

Re: Using indexes rather than table scans with Calcite

2020-06-02 Thread Julian Hyde
Vladimir, I feel the same way. MVs are more powerful and general, and with some effort they could be just as efficient as other approaches. One problem that needs to be solved is the “registration problem”. If you have a lot of MVs they all have to be registered in the planner’s search space,

Re: Regarding CALCITE-3997

2020-06-02 Thread Ruben Q L
Hello Haisheng, thanks for bringing up this discussion. I was also part of the people who expressed concerns about CALCITE-3997 due to a backwards compatibility issue in a very specific scenario in a downstream project. This issue was satisfactory handled with a temporary workaround for the RC,

Re: Using indexes rather than table scans with Calcite

2020-06-02 Thread Vladimir Ozerov
Hi Roman, To me, there is no principal difference between materialized views and rule-based approaches - they are both "rules". In the materialized views approach the rule is to create all alternative access paths unconditionally, this rule is always fired first during the optimization process. A