Change mapping of SQL `VARCHAR` from `Utf8` to `Utf8View` #15096

alamb · 2025-03-08T11:36:59Z

Is your feature request related to a problem or challenge?

DataFusion uses Arrow types internally. Thus when planning SQL queries there is a mapping from SQL types to Arrow Types. The current mapping for character types is shown in the docs https://datafusion.apache.org/user-guide/sql/data_types.html#character-types

SQL DataType	Arrow DataType
`CHAR`	`Utf8`
`VARCHAR`	`Utf8`
`TEXT`	`Utf8`
`STRING`	`Utf8`

So this means that when you do something like create table foo(x varchar); the x column is Utf8

DataFusion CLI v46.0.0
> create table foo(x varchar);

0 row(s) fetched.
Elapsed 0.019 seconds.

> describe foo;
+-------------+-----------+-------------+
| column_name | data_type | is_nullable |
+-------------+-----------+-------------+
| x           | Utf8      | YES         |
+-------------+-----------+-------------+
1 row(s) fetched.
Elapsed 0.008 seconds.

When reading parquet files however, a different type, Utf8View is used as it is faster in most cases.

This can be seen in this example:

DataFusion CLI v46.0.0
> describe 'hits.parquet';
+-----------------------+-----------+-------------+
| column_name           | data_type | is_nullable |
+-----------------------+-----------+-------------+
| WatchID               | Int64     | NO          |
| JavaEnable            | Int16     | NO          |
| Title                 | Utf8View  | NO          |
...
+-----------------------+-----------+-------------+
105 row(s) fetched.
Elapsed 0.032 seconds.

Thus there is a discrepancy when creating external tables with a schema (VARCHAR) as that will use Utf8 rather than UTF8View

I believe this is the root cause of the issue @zhuqi-lucas filed:

schema_force_view_type configuration not working for CREATE EXTERNAL TABLE #14909

Describe the solution you'd like

I think we should consider changing the default SQL mapping from VARCHAR --> Utf8View

Describe alternatives you've considered

@zhuqi-lucas has a PR that does this: BUG: schema_force_view_type configuration not working for CREATE EXTERNAL TABLE #14922

There are a few subtasks required before we can merge it:

feat: Support Utf8View in JSON reader arrow-rs#7244
Support datatype cast for insert api same as insert into sql #15015
Update the SQL documentation to refer to the changed mapping
Support approx_distinct for Utf8View
feat: topk functionality for aggregates should support utf8view and largeutf8 #15152
Improve performance sort TPCH q3 with Utf8Vew ( Sort-preserving merging on a single Utf8View ) #15403

Additional context

You can see some of the history related to using string view / Utf8View here:

[Epic] Complete Initial StringView in DataFusion #11752

The text was updated successfully, but these errors were encountered:

alamb · 2025-03-08T11:39:34Z

Please add comments if you find other needed items / issues

alamb · 2025-03-08T11:45:58Z

To begin this project so that we can implement it incrementally, I suggest we create a new config option like default_sql_to_view_types or something so we can test / work through issues before turning it on by default

zhuqi-lucas · 2025-03-08T11:48:34Z

To begin this project so that we can implement it incrementally, I suggest we create a new config option like default_sql_to_view_types or something so we can test / work through issues before turning it on by default

Thank you @alamb , this is a great suggestion! And we finally can make it default to true when we finish all tasks!

zhuqi-lucas · 2025-03-08T11:51:30Z

I will try to create more sub-tasks related to this effort!

zhuqi-lucas · 2025-03-09T06:57:11Z

Config: Add support default sql varchar to view types

Submitted a PR for the first step of our incremental work:
#15104

cc @alamb Thanks!

zhuqi-lucas · 2025-03-09T10:15:39Z

I also testing the tcph when it use the utf8view default, here is the result:

./benchmarks/bench.sh compare main issue_14909
Comparing main and issue_14909
--------------------
Benchmark tpch_sf1.json
--------------------
┏━━━━━━━━━━━━━━┳━━━━━━━━━┳━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query        ┃    main ┃ issue_14909 ┃        Change ┃
┡━━━━━━━━━━━━━━╇━━━━━━━━━╇━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1     │ 72.78ms │     70.32ms │     no change │
│ QQuery 2     │ 27.64ms │     26.30ms │     no change │
│ QQuery 3     │ 37.75ms │     38.22ms │     no change │
│ QQuery 4     │ 27.97ms │     28.27ms │     no change │
│ QQuery 5     │ 52.59ms │     49.36ms │ +1.07x faster │
│ QQuery 6     │ 20.81ms │     20.66ms │     no change │
│ QQuery 7     │ 70.44ms │     75.06ms │  1.07x slower │
│ QQuery 8     │ 48.32ms │     49.02ms │     no change │
│ QQuery 9     │ 62.60ms │     63.14ms │     no change │
│ QQuery 10    │ 55.94ms │     58.75ms │  1.05x slower │
│ QQuery 11    │ 19.44ms │     21.21ms │  1.09x slower │
│ QQuery 12    │ 36.59ms │     37.42ms │     no change │
│ QQuery 13    │ 34.05ms │     34.88ms │     no change │
│ QQuery 14    │ 26.50ms │     26.77ms │     no change │
│ QQuery 15    │ 42.97ms │     45.06ms │     no change │
│ QQuery 16    │ 19.25ms │     20.02ms │     no change │
│ QQuery 17    │ 73.64ms │     68.81ms │ +1.07x faster │
│ QQuery 18    │ 96.62ms │     95.08ms │     no change │
│ QQuery 19    │ 46.77ms │     45.75ms │     no change │
│ QQuery 20    │ 45.54ms │     40.98ms │ +1.11x faster │
│ QQuery 21    │ 95.29ms │     95.19ms │     no change │
│ QQuery 22    │ 18.34ms │     17.99ms │     no change │
└──────────────┴─────────┴─────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary          ┃           ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (main)          │ 1031.87ms │
│ Total Time (issue_14909)   │ 1028.26ms │
│ Average Time (main)        │   46.90ms │
│ Average Time (issue_14909) │   46.74ms │
│ Queries Faster             │         3 │
│ Queries Slower             │         3 │
│ Queries with No Change     │        16 │
└────────────────────────────┴───────────┘

zhuqi-lucas · 2025-03-11T06:17:32Z

Create the ticket for avro:

Support Utf8View for avro #7262

zhuqi-lucas · 2025-03-11T09:50:37Z

New sub_task:

topk functionality for aggregates should support utf8view

PR: #15152

zhuqi-lucas · 2025-03-14T12:04:32Z

New sub_task:

Support logic optimize rule to pass the case that Utf8view datatype combined with Utf8 datatype

zhuqi-lucas · 2025-03-16T06:43:24Z

New sub_task:

Support Utf8View datatype for range queries

Submitted a PR:
#15257

zhuqi-lucas · 2025-03-16T15:57:04Z

New sub_task:

Performance improvement: Binary operators native support for Utf8View

alamb · 2025-03-17T14:29:41Z

erformance improvement: Binary operators native support for Utf8View

Yes, 100%

zhuqi-lucas · 2025-03-17T15:05:17Z

erformance improvement: Binary operators native support for Utf8View

Yes, 100%

Submitted the PR for review:
#15275

zhuqi-lucas · 2025-03-21T09:33:40Z

New sub_task:

Perf: Support Utf8View datatype single column comparisons for SortPreservingMergeStream

zhuqi-lucas · 2025-03-21T09:36:29Z

Updated:

Most of the tasks are resolved, i am trying to do more performance investigation and testing if we default change to Utf8View for all varchar.

zhuqi-lucas · 2025-03-28T03:50:27Z

Also updated the latest clickbench for the current main compare the default mapping varchar to utf8view:

Small improvement, i think becasue it's parquet format, mostly we already load it as the Utf8View for benchmark:

Using --profile release-nonlto result:

┏━━━━━━━━━━━━━━┳━━━━━━━━━━━┳━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query        ┃       tmp ┃       tmp ┃        Change ┃
┡━━━━━━━━━━━━━━╇━━━━━━━━━━━╇━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0     │    0.36ms │    0.32ms │ +1.13x faster │
│ QQuery 1     │   27.55ms │   28.53ms │     no change │
│ QQuery 2     │   54.73ms │   58.73ms │  1.07x slower │
│ QQuery 3     │   49.67ms │   51.85ms │     no change │
│ QQuery 4     │  321.47ms │  333.88ms │     no change │
│ QQuery 5     │  374.77ms │  370.26ms │     no change │
│ QQuery 6     │   26.06ms │   27.14ms │     no change │
│ QQuery 7     │   29.89ms │   28.19ms │ +1.06x faster │
│ QQuery 8     │  397.90ms │  372.24ms │ +1.07x faster │
│ QQuery 9     │  583.40ms │  599.98ms │     no change │
│ QQuery 10    │  148.50ms │  147.43ms │     no change │
│ QQuery 11    │  163.20ms │  165.56ms │     no change │
│ QQuery 12    │  399.31ms │  407.08ms │     no change │
│ QQuery 13    │  568.61ms │  576.26ms │     no change │
│ QQuery 14    │  389.05ms │  374.46ms │     no change │
│ QQuery 15    │  375.66ms │  370.85ms │     no change │
│ QQuery 16    │  720.94ms │  719.03ms │     no change │
│ QQuery 17    │  662.21ms │  638.33ms │     no change │
│ QQuery 18    │ 1694.34ms │ 1507.92ms │ +1.12x faster │
│ QQuery 19    │   41.26ms │   42.08ms │     no change │
│ QQuery 20    │  619.60ms │  549.74ms │ +1.13x faster │
│ QQuery 21    │  779.77ms │  691.91ms │ +1.13x faster │
│ QQuery 22    │ 1411.33ms │ 1375.59ms │     no change │
│ QQuery 23    │ 3891.02ms │ 3946.51ms │     no change │
│ QQuery 24    │  252.52ms │  247.12ms │     no change │
│ QQuery 25    │  252.81ms │  248.90ms │     no change │
│ QQuery 26    │  264.57ms │  276.89ms │     no change │
│ QQuery 27    │  842.86ms │  854.72ms │     no change │
│ QQuery 28    │ 6461.67ms │ 6410.47ms │     no change │
│ QQuery 29    │  379.88ms │  359.71ms │ +1.06x faster │
│ QQuery 30    │  352.77ms │  332.15ms │ +1.06x faster │
│ QQuery 31    │  366.75ms │  371.79ms │     no change │
│ QQuery 32    │ 1273.77ms │ 1427.04ms │  1.12x slower │
│ QQuery 33    │ 1601.21ms │ 1599.55ms │     no change │
│ QQuery 34    │ 1605.00ms │ 1701.18ms │  1.06x slower │
│ QQuery 35    │  532.88ms │  576.30ms │  1.08x slower │
│ QQuery 36    │  109.92ms │  115.27ms │     no change │
│ QQuery 37    │   57.43ms │   57.22ms │     no change │
│ QQuery 38    │   78.78ms │   80.17ms │     no change │
│ QQuery 39    │  196.90ms │  197.13ms │     no change │
│ QQuery 40    │   26.52ms │   25.77ms │     no change │
│ QQuery 41    │   25.57ms │   25.51ms │     no change │
│ QQuery 42    │   30.03ms │   28.90ms │     no change │
└──────────────┴───────────┴───────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary      ┃            ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (tmp)       │ 28442.44ms │
│ Total Time (tmp)       │ 28319.66ms │
│ Average Time (tmp)     │   661.45ms │
│ Average Time (tmp)     │   658.60ms │
│ Queries Faster         │          8 │
│ Queries Slower         │          4 │
│ Queries with No Change │         31 │
└────────────────────────┴────────────┘
```rust


Using run --release result:


```rust
┏━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query        ┃ main                   ┃ default_enable_utf8view ┃        Change ┃
┡━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0     │                 0.34ms │                  0.33ms │     no change │
│ QQuery 1     │                44.10ms │                 43.49ms │     no change │
│ QQuery 2     │                77.24ms │                 76.79ms │     no change │
│ QQuery 3     │                84.97ms │                 77.06ms │ +1.10x faster │
│ QQuery 4     │               523.60ms │                554.50ms │  1.06x slower │
│ QQuery 5     │               665.56ms │                661.98ms │     no change │
│ QQuery 6     │                38.69ms │                 39.05ms │     no change │
│ QQuery 7     │                47.27ms │                 46.57ms │     no change │
│ QQuery 8     │               702.86ms │                682.91ms │     no change │
│ QQuery 9     │               780.11ms │                770.69ms │     no change │
│ QQuery 10    │               194.09ms │                172.24ms │ +1.13x faster │
│ QQuery 11    │               199.83ms │                191.13ms │     no change │
│ QQuery 12    │               696.13ms │                688.11ms │     no change │
│ QQuery 13    │               890.35ms │               1001.57ms │  1.12x slower │
│ QQuery 14    │               732.92ms │                648.96ms │ +1.13x faster │
│ QQuery 15    │               689.57ms │                633.55ms │ +1.09x faster │
│ QQuery 16    │              1415.16ms │               1468.50ms │     no change │
│ QQuery 17    │              1297.66ms │               1319.06ms │     no change │
│ QQuery 18    │              3272.62ms │               2857.06ms │ +1.15x faster │
│ QQuery 19    │                75.13ms │                 82.66ms │  1.10x slower │
│ QQuery 20    │               743.88ms │                705.83ms │ +1.05x faster │
│ QQuery 21    │               929.50ms │                897.88ms │     no change │
│ QQuery 22    │              2576.76ms │               2506.14ms │     no change │
│ QQuery 23    │              4943.09ms │               4916.55ms │     no change │
│ QQuery 24    │               392.47ms │                384.78ms │     no change │
│ QQuery 25    │               386.58ms │                388.37ms │     no change │
│ QQuery 26    │               423.42ms │                417.96ms │     no change │
│ QQuery 27    │              1050.88ms │                976.19ms │ +1.08x faster │
│ QQuery 28    │              8269.73ms │               8791.73ms │  1.06x slower │
│ QQuery 29    │               439.96ms │                442.74ms │     no change │
│ QQuery 30    │               583.71ms │                541.02ms │ +1.08x faster │
│ QQuery 31    │               632.33ms │                629.25ms │     no change │
│ QQuery 32    │              2580.37ms │               2523.11ms │     no change │
│ QQuery 33    │              2810.58ms │               2848.06ms │     no change │
│ QQuery 34    │              3075.43ms │               3108.88ms │     no change │
│ QQuery 35    │               856.76ms │                891.39ms │     no change │
│ QQuery 36    │               152.80ms │                150.15ms │     no change │
│ QQuery 37    │               117.99ms │                118.82ms │     no change │
│ QQuery 38    │               110.05ms │                112.33ms │     no change │
│ QQuery 39    │               267.64ms │                279.12ms │     no change │
│ QQuery 40    │                41.34ms │                 45.52ms │  1.10x slower │
│ QQuery 41    │                42.04ms │                 41.70ms │     no change │
│ QQuery 42    │                50.40ms │                 45.41ms │ +1.11x faster │
└──────────────┴────────────────────────┴─────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary                      ┃            ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (main)                        │ 43905.93ms │
│ Total Time (default_enable_utf8view)   │ 43779.11ms │
│ Average Time (main)                      │  1021.07ms │
│ Average Time (default_enable_utf8view) │  1018.12ms │
│ Queries Faster                         │          9 │
│ Queries Slower                         │          5 │
│ Queries with No Change                 │         29 │
└────────────────────────────────────────┴────────────┘

alamb · 2025-03-28T18:07:28Z

Small improvement, i think becasue it's parquet format, mostly we already load it as the Utf8View for benchmark:

Yes I would expect no change for the clickbench benchmark as it doesn't use SQL VARCHAR 🤔

alamb added the enhancement New feature or request label Mar 8, 2025

alamb mentioned this issue Mar 8, 2025

feat: Support Utf8View in JSON reader apache/arrow-rs#7244

Closed

alamb mentioned this issue Mar 8, 2025

BUG: schema_force_view_type configuration not working for CREATE EXTERNAL TABLE #14922

Draft

alamb mentioned this issue Mar 8, 2025

schema_force_view_type configuration not working for CREATE EXTERNAL TABLE #14909

Open

zhuqi-lucas mentioned this issue Mar 9, 2025

Config: Add support default sql varchar to view types #15104

Merged

This was referenced Mar 11, 2025

feat: topk functionality for aggregates should support utf8view and largeutf8 #15152

Merged

feat: support ApproxDistinct with utf8view #15200

Merged

alamb mentioned this issue Mar 13, 2025

Release DataFusion 47.0.0 (April 2025) #15072

Open

28 tasks

zhuqi-lucas mentioned this issue Mar 15, 2025

Support logic optimize rule to pass the case that Utf8view datatype combined with Utf8 datatype #15239

Merged

zhuqi-lucas mentioned this issue Mar 16, 2025

Support utf8view datatype for window #15257

Merged

alamb mentioned this issue Mar 17, 2025

March 17, 2025: This week(s) in DataFusion #15269

Open

alamb mentioned this issue Mar 17, 2025

Weekly Plan (Andrew Lamb) March 17, 2025 #15274

Closed

10 tasks

zhuqi-lucas mentioned this issue Mar 17, 2025

feat: Native support utf8view for regex string operators #15275

Merged

zhuqi-lucas mentioned this issue Mar 21, 2025

Perf: Support Utf8View datatype single column comparisons for SortPreservingMergeStream #15348

Merged

alamb mentioned this issue Mar 24, 2025

Weekly Plan (Andrew Lamb) March 24, 2025 #15393

Closed

9 tasks

alamb mentioned this issue Mar 25, 2025

Improve performance sort TPCH q3 with Utf8Vew ( Sort-preserving merging on a single Utf8View ) #15403

Closed

zhuqi-lucas mentioned this issue Apr 1, 2025

Add utf8view benchmark for aggregate topk #15518

Open

alamb mentioned this issue Apr 1, 2025

Weekly Plan (Andrew Lamb) March 31, 2025 #15528

Open

11 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Change mapping of SQL `VARCHAR` from `Utf8` to `Utf8View` #15096

Change mapping of SQL `VARCHAR` from `Utf8` to `Utf8View` #15096

alamb commented Mar 8, 2025 •

edited

Loading

alamb commented Mar 8, 2025

alamb commented Mar 8, 2025

zhuqi-lucas commented Mar 8, 2025

zhuqi-lucas commented Mar 8, 2025

zhuqi-lucas commented Mar 9, 2025 •

edited

Loading

zhuqi-lucas commented Mar 9, 2025 •

edited

Loading

zhuqi-lucas commented Mar 11, 2025

zhuqi-lucas commented Mar 11, 2025 •

edited

Loading

zhuqi-lucas commented Mar 14, 2025 •

edited

Loading

zhuqi-lucas commented Mar 16, 2025 •

edited

Loading

zhuqi-lucas commented Mar 16, 2025 •

edited

Loading

alamb commented Mar 17, 2025

zhuqi-lucas commented Mar 17, 2025

zhuqi-lucas commented Mar 21, 2025 •

edited

Loading

zhuqi-lucas commented Mar 21, 2025

zhuqi-lucas commented Mar 28, 2025 •

edited

Loading

alamb commented Mar 28, 2025

Change mapping of SQL VARCHAR from Utf8 to Utf8View #15096

Change mapping of SQL VARCHAR from Utf8 to Utf8View #15096

Comments

alamb commented Mar 8, 2025 • edited Loading

Is your feature request related to a problem or challenge?

Describe the solution you'd like

Describe alternatives you've considered

Additional context

alamb commented Mar 8, 2025

alamb commented Mar 8, 2025

zhuqi-lucas commented Mar 8, 2025

zhuqi-lucas commented Mar 8, 2025

zhuqi-lucas commented Mar 9, 2025 • edited Loading

zhuqi-lucas commented Mar 9, 2025 • edited Loading

zhuqi-lucas commented Mar 11, 2025

zhuqi-lucas commented Mar 11, 2025 • edited Loading

zhuqi-lucas commented Mar 14, 2025 • edited Loading

zhuqi-lucas commented Mar 16, 2025 • edited Loading

zhuqi-lucas commented Mar 16, 2025 • edited Loading

alamb commented Mar 17, 2025

zhuqi-lucas commented Mar 17, 2025

zhuqi-lucas commented Mar 21, 2025 • edited Loading

zhuqi-lucas commented Mar 21, 2025

zhuqi-lucas commented Mar 28, 2025 • edited Loading

alamb commented Mar 28, 2025

Change mapping of SQL `VARCHAR` from `Utf8` to `Utf8View` #15096

Change mapping of SQL `VARCHAR` from `Utf8` to `Utf8View` #15096

alamb commented Mar 8, 2025 •

edited

Loading

zhuqi-lucas commented Mar 9, 2025 •

edited

Loading

zhuqi-lucas commented Mar 9, 2025 •

edited

Loading

zhuqi-lucas commented Mar 11, 2025 •

edited

Loading

zhuqi-lucas commented Mar 14, 2025 •

edited

Loading

zhuqi-lucas commented Mar 16, 2025 •

edited

Loading

zhuqi-lucas commented Mar 16, 2025 •

edited

Loading

zhuqi-lucas commented Mar 21, 2025 •

edited

Loading

zhuqi-lucas commented Mar 28, 2025 •

edited

Loading