Re: [PR] Draft: Make Clickbench Q29 5x faster for datafusion [datafusion]

2025-04-05 Thread via GitHub
zhuqi-lucas commented on PR #15532: URL: https://github.com/apache/datafusion/pull/15532#issuecomment-2771316983 More than 5X faster for clickbench Q29 with this PR: ```rust cargo run --profile release-nonlto --target aarch64-apple-darwin --bin dfbench -- clickbench -p benchmark

Re: [PR] Draft: Make Clickbench Q29 5x faster for datafusion [datafusion]

2025-04-04 Thread via GitHub
zhuqi-lucas commented on code in PR #15532: URL: https://github.com/apache/datafusion/pull/15532#discussion_r2024355571 ## datafusion/sqllogictest/test_files/explain.slt: ## @@ -244,6 +244,159 @@ physical_plan DataSourceExec: file_groups={1 group: [[WORKSPACE_ROOT/datafusion/

Re: [PR] Draft: Make Clickbench Q29 5x faster for datafusion [datafusion]

2025-04-04 Thread via GitHub
Dandandan commented on code in PR #15532: URL: https://github.com/apache/datafusion/pull/15532#discussion_r2024211429 ## datafusion/sqllogictest/test_files/explain.slt: ## @@ -244,6 +244,159 @@ physical_plan DataSourceExec: file_groups={1 group: [[WORKSPACE_ROOT/datafusion/ ph

Re: [PR] Draft: Make Clickbench Q29 5x faster for datafusion [datafusion]

2025-04-04 Thread via GitHub
andygrove commented on PR #15532: URL: https://github.com/apache/datafusion/pull/15532#issuecomment-2772755908 This seems like it changes the semantics of the query and does not guarantee the correct results. I am thinking specifically about order-of-operations when applying this to decimal

Re: [PR] Draft: Make Clickbench Q29 5x faster for datafusion [datafusion]

2025-04-04 Thread via GitHub
alamb commented on code in PR #15532: URL: https://github.com/apache/datafusion/pull/15532#discussion_r2029009800 ## datafusion/optimizer/src/simplify_expressions/expr_simplifier.rs: ## @@ -1447,15 +1449,240 @@ impl TreeNodeRewriter for Simplifier<'_, S> { }

Re: [PR] Draft: Make Clickbench Q29 5x faster for datafusion [datafusion]

2025-04-03 Thread via GitHub
zhuqi-lucas commented on PR #15532: URL: https://github.com/apache/datafusion/pull/15532#issuecomment-2775156329 Thank you @andygrove @berkaysynnada , i believe we need more effort if we want to support such as sum(a + b) => sum(a) + sum(b). **Need help, i am not sure if i do the

Re: [PR] Draft: Make Clickbench Q29 5x faster for datafusion [datafusion]

2025-04-02 Thread via GitHub
berkaysynnada commented on PR #15532: URL: https://github.com/apache/datafusion/pull/15532#issuecomment-2772920116 > This seems like it changes the semantics of the query and does not guarantee the correct results. I am thinking specifically about order-of-operations when applying this to d

Re: [PR] Draft: Make Clickbench Q29 5x faster for datafusion [datafusion]

2025-04-02 Thread via GitHub
zhuqi-lucas commented on PR #15532: URL: https://github.com/apache/datafusion/pull/15532#issuecomment-2772619544 Addressed the comments for sum(3*a + 2*b) => 3*sum(a) + 2*sum(b), and but more corner cases and testing failed, i am looking into and try to fix it. -- This is an automated mes

Re: [PR] Draft: Make Clickbench Q29 5x faster for datafusion [datafusion]

2025-04-01 Thread via GitHub
Dandandan commented on code in PR #15532: URL: https://github.com/apache/datafusion/pull/15532#discussion_r2024173764 ## datafusion/sqllogictest/test_files/explain.slt: ## @@ -244,6 +244,159 @@ physical_plan DataSourceExec: file_groups={1 group: [[WORKSPACE_ROOT/datafusion/ ph

[PR] Draft: Make Clickbench Q29 5x faster for datafusion [datafusion]

2025-04-01 Thread via GitHub
zhuqi-lucas opened a new pull request, #15532: URL: https://github.com/apache/datafusion/pull/15532 ## Which issue does this PR close? - Closes #. ## Rationale for this change ## What changes are included in this PR? ## Are these changes tes