DataFusion weekly project plan (Andrew Lamb) - March 18, 2024

[alamb]

Follow on to #9555

It would be great for other contributors to DataFusion who plan non trivial work could try to make them visible somehow as well 🙏 (feel free to copy / modify the format)

My Highlights from Last Week

• Add StringViewArray implementation and layout and basic construction + tests arrow-rs#5469 (first step towards [EPIC] Implement StringViewArray and BinaryViewArray arrow-rs#5374 was merged -- thanks @ariesdevil and @XiangpengHao)
• @backkem and @devinjdangelo began implementing Support LogicalPlan --> SQL String #8661
• @jayzhan211 started working on improving planning speed with [EPIC] Stop copying LogicalPlan during OptimizerPasses #9637 (and found a nice improvement already Minor: remove clone in exprlist_to_fields #9657)
• Recursive CTE's enabled by default: feat: track memory usage for recursive CTE, enable recursive CTEs by default #9619 (thanks @jonahgao !)
• Large scale "extract scalar functions from the core" continues at good pace way [Epic] Port BuiltInFunctons to datafusion-functions-* crates #9285 (thanks @jayzhan211 and @erenavsarogullari for being a machine here)

My (personal) goals for this project

1000 projects are built using DataFusion 📈

My (personal) plans for this week

• SIGMOD paper feedback : Finalize SIGMOD 2024 paper ~(if accepted)~ #8373 (10 days left)
• Prepare content for UPDATE: VENUE IS FULL Invitation: Mar 25, 2024 DataFusion meetup before the DataCouncil March 2024 in Austin TX #8522
• Fix Regression: Wrong result when there are 2 count(distinct) #9586

Project Queue (list of future projects)

Projects I plan to help actively help review / plan

• [EPIC] Implement StringViewArray and BinaryViewArray arrow-rs#5374
• [Epic] Port BuiltInFunctons to datafusion-functions-* crates #9285
• Support "A column is known to be entirely NULL" in PruningPredicate #9171 with @appletreeisyellow (we need this for InfluxDB)
• Supporting using parallel parquet writer outside of Datafusion query execution #9493 with @wiedld

Algorithm for (my) prioritizing PR reviews

Note there are many committers who can and do review and merge PRs, so this is not the priorities of the project as a whole, just the approximate algorithm I use to prioritize my own time

Priority:

1. Bug fixes (where something is just incorrect), especially regressions (where it used to work and now does not)
2. Improvements directly related to features needed for InfluxDB (my employer)
3. Documentation and test improvements (I view these as very strategically important)
4. PRs that I think are strategically important
5. Other new features / additions to functionality (note this is the lowest)

The top strategically important projects in my head are:

• Anything that makes it easier to use DataFusion as a user (docs, examples, better APIs, etc), including Getting started guide for new users (who want to use DataFusion in their project)
• Anything that improves DataFusion's quality such as bug fixes, and improved / expanded tests, etc
• Making function packages modular and easier to mix/match: [EPIC] Unify Function Interface (remove BuiltInScalarFunction) #8045
• Improving performance of aggregation - especially high cardinality groupings, where we still lag state of the art by 2x

Thus, if you are interested in contributing to DataFusion and are interested in a fast turn around time I would recommend looking into bug fixes / test improvements / documentation or the projects named above

If you propose adding new functionality, the review cycle will likely be longer. You can make it a shorter cycle by looking at the comments on other recent PRs and following the same model (e.g. ensure there are tests in sqllogictest for example, the CI passes, includes documentation, etc)

Background

The idea of this ticket is make my plans for DataFusion visible, largely for my own personal organizational needs, but also to:

1. Communicate / coordinate in the community
2. Help provide an interesting summary of what is happening in DataFusion this week

[alamb]

Reviews:

DataFusion:

• Update Arrow/Parquet to 51.0.0, tonic to 0.11 #9613 (review)
• parquet: Add row_groups_matched_{statistics,bloom_filter} statistics #9640

[alamb]

DataFusion:

• Support Union types in ScalarValue  #9683 (review)
• core/tests/parquet/row_group_pruning.rs: Add tests for strings #9642
• fix: recursive cte hangs on joins #9687
• parquet: Add support for Bloom filters on binary columns #9644
• Use object_store:BufWriter to replace put_multipart #9648
• improve null handling for to_char #9689 (review)
• Move inlist rule to expr_simplifier #9692
• Projection Pushdown through user defined LogicalPlan nodes. #9690 (review)
• refactor: make dfschema wrap schemaref #9595
• Add GroupValuesFullyOrdered mode to GroupValues trait for aggregate grouping. #9662
• parquet: Add support for row group pruning on FixedSizeBinary  #9646
• feat: Use file cache to list partitions if available #9655
• feat: Determine ordering of file groups #9593 (review)

Arrow

• Replace AsyncWrite with Upload trait and rename MultiPartStore to MultipartStore (#5458) arrow-rs#5500
• IPC format support for StringViewArray and BinaryViewArray arrow-rs#5525

[alamb]

DataFusion

• refactor: make dfschema wrap schemaref #9595
• build: modify code to comply with latest clippy requirement #9725
• make format prefix optional for format options in COPY #9723
• Move trim functions (btrim, ltrim, rtrim) to datafusion_functions, make expr_fn API consistent #9730
• feat: Determine ordering of file groups #9593
• Projection Pushdown through user defined LogicalPlan nodes. #9690
• Optimization: concat function #9732
• fix: parallel parquet can underflow when max_record_batch_rows < execution.batch_size #9737
• Move ascii function to datafusion_functions #9740
• adding expr to string for IsNotNull IsTrue IsFalse and IsUnkown #9739

Arrow

• IPC format support for StringViewArray and BinaryViewArray arrow-rs#5525
• Add IPC StreamDecoder arrow-rs#5531
• feat: implemented with_field() for FixedSizeListBuilder arrow-rs#5541

[alamb]

DataFusion

• Move levenshtein, uuid, overlay to datafusion-functions #9760
• Add Expr->String for ScalarFunction and InList #9759
• fix: duplicate output for HashJoinExec in CollectLeft mode #9757
• Support custom struct field names with new scalar function named_struct #9743
• Allow declaring partition columns in PARTITION BY clause, backwards compatible #9599
• Support tencent cloud COS storage in datafusion-cli #9734
• feat: Use file cache to list partitions if available #9655
• Add support for Bloom filters on unsigned integer columns in Parquet tables #9770

[alamb]

#9796 next week