ARROW-11651: [Rust][DataFusion] Implement Postgres String Functions: Length Functions #9509

seddonm1 · 2021-02-16T21:23:24Z

Splitting up #9243

This implements the following functions:

String functions
- bit_Length
- char_length
- character_length
- length
- octet_length

github-actions · 2021-02-16T21:23:46Z

https://issues.apache.org/jira/browse/ARROW-11298

github-actions · 2021-02-16T22:47:55Z

https://issues.apache.org/jira/browse/ARROW-11651

jorgecarleitao

Thanks a lot @seddonm1 . This looks great to me.

I left two comments that imo should be addressed related to reducing the maintenance cost by DRY.

jorgecarleitao · 2021-02-17T02:24:44Z

rust/arrow/src/compute/kernels/bit_length.rs

+fn bit_length_string<OffsetSize>(array: &Array, data_type: DataType) -> ArrayRef
+where
+    OffsetSize: OffsetSizeTrait,
+{
+    // note: offsets are stored as u8, but they can be interpreted as OffsetSize
+    let offsets = &array.data_ref().buffers()[0];
+    // this is a 30% improvement over iterating over u8s and building OffsetSize, which
+    // justifies the usage of `unsafe`.
+    let slice: &[OffsetSize] =
+        &unsafe { offsets.typed_data::<OffsetSize>() }[array.offset()..];
+
+    let bit_size = OffsetSize::from_usize(8).unwrap();
+    let lengths = slice
+        .windows(2)
+        .map(|offset| (offset[1] - offset[0]) * bit_size);
+
+    // JUSTIFICATION
+    //  Benefit
+    //      ~60% speedup
+    //  Soundness
+    //      `values` is an iterator with a known size.
+    let buffer = unsafe { Buffer::from_trusted_len_iter(lengths) };
+


Isn't this a copy of length, with a single difference, that we multiply the lengths by bit_size?

If that is the case, can we create a generic or something that avoids code duplication? Copy-pasting code like this is IMO an unnecessary maintenance burden.

yes i can have a look at doing that.

I have created a macro allowing just the core function to be replaced.

jorgecarleitao · 2021-02-17T02:31:52Z

rust/datafusion/src/physical_plan/functions.rs

+    /// $ARRAY_TYPE is the column type after function applied
+    macro_rules! test_function {
+        ($FUNC:ident, $ARGS:expr, $EXPECTED:expr, $EXPECTED_TYPE:ty, $DATA_TYPE: ident, $ARRAY_TYPE:ident) => {
+            println!("{:?}", BuiltinScalarFunction::$FUNC);


is this println! intended?

it is just there to help debugging but happy to remove

jorgecarleitao · 2021-02-17T02:40:02Z

rust/datafusion/src/physical_plan/string_expressions.rs

+
+/// Returns number of characters in the string.
+/// character_length_i64('josé') = 4
+pub fn character_length_i64(args: &[ArrayRef]) -> Result<ArrayRef> {


I think that this code can be reduced to a single generic. Something like this (untested):

pub fn character_length<T: ArrowPrimitiveType>(args: &[ArrayRef]) -> Result<ArrayRef> where T::Native: StringOffsetSizeTrait { let string_array: &GenericStringArray<T::Native> = args[0] .as_any() .downcast_ref::<GenericStringArray<T::Native>>() .unwrap(); string_array .iter() .map(|x| x.map(|x: &str| T::Native::from_usize(x.graphemes(true).count())).unwrap()) .collect() }

Thanks I will have a go. There are a couple more instances this generic from_usize could help.

Wow. Impressive you can come up with this in your head. I have incorporated this code.

seddonm1 · 2021-02-17T03:28:05Z

Thanks @jorgecarleitao . I will update soon.

seddonm1 · 2021-02-17T08:01:52Z

@jorgecarleitao Thanks for the review. I have addressed all of your points.

The T::Native: StringOffsetSizeTrait will be useful in some of the other string functions so thanks 🙏

jorgecarleitao · 2021-02-17T17:00:22Z

Thanks a lot, @seddonm1 . I went through this PR again and I pull the code locally to understand the reason for the macro.

I now get it that it is not easy to write this. :) I PRed something to your branch with a suggestion to reduce the risk of using that function (which is unsafe due to the typed) on non-string arrays and make it just use generics, that imo are usually easier to read because the constraints are specific.

Added some safety.

codecov-io · 2021-02-17T20:59:52Z

Codecov Report

Merging #9509 (17740e9) into master (e2d6c05) will increase coverage by 0.12%.
The diff coverage is 88.05%.

@@            Coverage Diff             @@
##           master    #9509      +/-   ##
==========================================
+ Coverage   82.15%   82.28%   +0.12%     
==========================================
  Files         243      244       +1     
  Lines       54908    55517     +609     
==========================================
+ Hits        45108    45680     +572     
- Misses       9800     9837      +37

Impacted Files	Coverage Δ
rust/datafusion/src/physical_plan/functions.rs	`72.36% <68.49%> (-3.95%)`	⬇️
rust/datafusion/src/logical_plan/expr.rs	`81.13% <80.00%> (+0.94%)`	⬆️
rust/arrow/src/compute/kernels/length.rs	`100.00% <100.00%> (ø)`
...datafusion/src/physical_plan/string_expressions.rs	`69.62% <100.00%> (+2.49%)`	⬆️
rust/arrow/src/array/equal/utils.rs	`75.49% <0.00%> (-0.99%)`	⬇️
rust/arrow/src/record_batch.rs	`83.70% <0.00%> (-0.63%)`	⬇️
rust/datafusion/src/physical_plan/limit.rs	`57.47% <0.00%> (-0.49%)`	⬇️
rust/datafusion/src/physical_plan/planner.rs	`79.58% <0.00%> (-0.05%)`	⬇️
rust/datafusion/src/logical_plan/plan.rs	`82.45% <0.00%> (-0.04%)`	⬇️
rust/benchmarks/src/bin/tpch.rs	`38.33% <0.00%> (ø)`
... and 12 more

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update e2d6c05...17740e9. Read the comment docs.

seddonm1 · 2021-02-18T20:21:12Z

@jorgecarleitao Sorry I updated this but forgot to update you. Once this is merged I can do more of the big Postgres string functions PRs.

alamb · 2021-02-19T13:49:23Z

The integration test failure appears unrelated https://github.com/apache/arrow/pull/9509/checks?check_run_id=1921983282

I will restart it to try and get a clean run


[INFO] Final Memory: 176M/780M
[INFO] ------------------------------------------------------------------------
Error:  Failed to execute goal org.xolstice.maven.plugins:protobuf-maven-plugin:0.5.0:compile (src) on project flight-core: Execution src of goal org.xolstice.maven.plugins:protobuf-maven-plugin:0.5.0:compile failed: Plugin org.xolstice.maven.plugins:protobuf-maven-plugin:0.5.0 or one of its dependencies could not be resolved: Could not transfer artifact org.apache.maven.plugin-tools:maven-plugin-annotations:jar:3.4 from/to central (https://repo.maven.apache.org/maven2): Connection reset -> [Help 1]
Error:  
Error:  To see the full stack trace of the errors, re-run Maven with the -e switch.
Error:  Re-run Maven using the -X switch to enable full debug logging.
Error:  
Error:  For more information about the errors and possible solutions, please read the following articles:
Error:  [Help 1] http://cwiki.apache.org/confluence/display/MAVEN/PluginResolutionException
Error:  
Error:  After correcting the problems, you can resume the build with the command
Error:    mvn <goals> -rf :flight-core
Destroying 3 processes
Destroying process..
Destroying process..
Destroying process..

alamb

Nice work @seddonm1 -- I think this is ready to go once it is green

alamb · 2021-02-20T13:46:47Z

@seddonm1 we are so close -- now there is a clippy error: https://github.com/apache/arrow/pull/9509/checks?check_run_id=1922005826

ror: this function's return value is unnecessarily wrapped by `Result`
   --> arrow/src/compute/kernels/length.rs:86:1
    |
86  | / fn bit_length_impl<O: StringOffsetSizeTrait, T: ArrowPrimitiveType>(
87  | |     array: &dyn Array,
88  | | ) -> Result<ArrayRef>
89  | | where
...   |
99  | |     }))
100 | | }
    | |_^
    |
    = note: `-D clippy::unnecessary-wraps` implied by `-D warnings`
    = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#unnecessary_wraps
help: remove `Result` from the return type...
    |
88  | ) -> std::sync::Arc<(dyn array::array::Array + 'static)>
    |      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
help: ...and change the returning expressions
    |
97  |     unary_offsets_string::<O, _>(array, T::DATA_TYPE, |x| {
98  |         x * bits_in_bytes
99  |     })
    |

error: aborting due to previous error

error: could not compile `arrow`

To learn more, run the command again with --verbose.
warning: build failed, waiting for other jobs to finish...
error: this function's return value is unnecessarily wrapped by `Result`
   --> arrow/src/compute/kernels/length.rs:86:1
    |
86  | / fn bit_length_impl<O: StringOffsetSizeTrait, T: ArrowPrimitiveType>(
87  | |     array: &dyn Array,
88  | | ) -> Result<ArrayRef>
89  | | where
...   |
99  | |     }))
100 | | }
    | |_^
    |
    = note: `-D clippy::unnecessary-wraps` implied by `-D warnings`
    = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#unnecessary_wraps
help: remove `Result` from the return type...
    |
88  | ) -> std::sync::Arc<(dyn array::array::Array + 'static)>
    |      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
help: ...and change the returning expressions
    |
97  |     unary_offsets_string::<O, _>(array, T::DATA_TYPE, |x| {
98  |         x * bits_in_bytes
99  |     })
    |

seddonm1 · 2021-02-20T22:44:50Z

@alamb done and ready. Thanks to @jorgecarleitao for his elite knowledge of trait implementations.

jorgecarleitao

Thanks a lot, @seddonm1 ! Ready to go from my side.

alamb · 2021-02-21T10:36:20Z

Thanks again @seddonm1 -- this is epic work

length functions

29939a5

github-actions bot added Component: Rust - DataFusion Component: Rust labels Feb 16, 2021

seddonm1 changed the title ~~ARROW-11298: [Rust][DataFusion] Implement Postgres String Functions: Length Functions~~ ARROW-11651: [Rust][DataFusion] Implement Postgres String Functions: Length Functions Feb 16, 2021

jorgecarleitao reviewed Feb 17, 2021

View reviewed changes

address review comments

04f3bba

Added some safety.

23b1329

seddonm1 added 2 commits February 18, 2021 07:17

Merge pull request #1 from jorgecarleitao/safety

328bb24

Added some safety.

clippy

619f2f6

alamb approved these changes Feb 19, 2021

View reviewed changes

clippy

76855f2

fmt

17740e9

jorgecarleitao approved these changes Feb 21, 2021

View reviewed changes

alamb closed this in abf967d Feb 21, 2021

This was referenced Feb 21, 2021

ARROW-11426: [Rust][DataFusion] EXTRACT support #9359

Closed

ARROW-11298: [Rust][DataFusion] Implement Postgres String Functions [Splitting to separate PRs] #9243

Closed

asfimport mentioned this pull request Feb 21, 2021

[Rust][DataFusion] Implement Postgres Length Functions #27514

Closed

asfimport mentioned this pull request Apr 26, 2021

[Rust][DataFusion] Implement Postgres String Functions #27198

Closed

7 tasks

ARROW-11651: [Rust][DataFusion] Implement Postgres String Functions: Length Functions #9509

ARROW-11651: [Rust][DataFusion] Implement Postgres String Functions: Length Functions #9509

Uh oh!

Conversation

seddonm1 commented Feb 16, 2021

Uh oh!

github-actions bot commented Feb 16, 2021

Uh oh!

github-actions bot commented Feb 16, 2021

Uh oh!

jorgecarleitao left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jorgecarleitao Feb 17, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

seddonm1 commented Feb 17, 2021

Uh oh!

seddonm1 commented Feb 17, 2021

Uh oh!

jorgecarleitao commented Feb 17, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

codecov-io commented Feb 17, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

seddonm1 commented Feb 18, 2021

Uh oh!

alamb commented Feb 19, 2021

Uh oh!

alamb left a comment

Choose a reason for hiding this comment

Uh oh!

alamb commented Feb 20, 2021

Uh oh!

seddonm1 commented Feb 20, 2021

Uh oh!

jorgecarleitao left a comment

Choose a reason for hiding this comment

Uh oh!

alamb commented Feb 21, 2021

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

jorgecarleitao left a comment •

edited

Loading

jorgecarleitao Feb 17, 2021 •

edited

Loading

jorgecarleitao commented Feb 17, 2021 •

edited

Loading

codecov-io commented Feb 17, 2021 •

edited

Loading