refactor(ci): bump rust version

[ClSlaid]

Signed-off-by: 蔡略 [email protected]

I hereby agree to the terms of the CLA available at: https://databend.rs/dev/policies/cla/

Summary

bump rust toolchain to rust nightly-2022-12-15

Change log

1. update CI toolchain to nightly-2022-12-15
2. update rust toolchain to nightly-2022-12-15
3. allows clippy::unlined_format_args
4. update stateless-test/10_drivers/10_0001_python_sqlalchemy_driver for sqlalchemy 2.0 compatibilities, this makes this PR a refactor
5. remove unnecessary lifetime annotations
6. fix some higher-ranked lifetime problem introduced in the new toolchain, thanks to @Xuanwo
7. some bonus changes, like changing (b'0'..b'9').contains(k) to k.is_ascii_digit()
8. make cargo-machete ignore match-template in storages-common-index

Closes #issue

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

1 Ignored Deployment

Name	Status	Preview	Updated
databend	⬜️ Ignored (Inspect)		Jan 11, 2023 at 7:26AM (UTC)

[Xuanwo]

Please follow https://databend.rs/doc/faq/routine-maintenance to bump rust toolchain.

[Xuanwo]

This PR is blocked by rust-lang/rust#64552

I tried some workaround but failing with:

error: implementation of `std::marker::Send` is not general enough
   --> src/query/storages/fuse/src/operations/mutation/deletion/deletion_transform.rs:331:53
    |
331 |       async fn async_process(&mut self) -> Result<()> {
    |  _____________________________________________________^
332 | |         match std::mem::replace(&mut self.state, State::None) {
333 | |             State::ReadSegments => {
334 | |                 let ctx = self.ctx.clone();
...   |
368 | |         Ok(())
369 | |     }
    | |_____^ implementation of `std::marker::Send` is not general enough
    |
    = note: `std::marker::Send` would have to be implemented for the type `&'0 [storages_common_table_meta::meta::Location]`, for any lifetime `'0`...
    = note: ...but `std::marker::Send` is actually implemented for the type `&'1 [storages_common_table_meta::meta::Location]`, for some specific lifetime `'1`

Seems we need to refactor SegmentsIO::read_segments to make it pass:

#[tracing::instrument(level = "debug", skip_all)]
pub async fn read_segments(
    &self,
    segment_locations: &[Location],
) -> Result<Vec<Result<Arc<SegmentInfo>>>> {}

[Xuanwo]

Now I'm sure problem happens at this arm:

    async fn async_process(&mut self) -> Result<()> {
        let state = std::mem::replace(&mut self.state, State::None);

        match state {
            // State::ReadSegments => {
            //     let ctx = self.ctx.clone();
            //     let dal = self.dal.clone();
            //     let schema = self.schema.clone();
            //     let segment_locations = self.base_segments.clone();

            //     let fut = Box::pin(async move {
            //         // Read all segments information in parallel.
            //         SegmentsIO::read_segments_work_around(ctx, dal, schema, segment_locations)
            //             .await?
            //             .into_iter()
            //             .collect::<Result<Vec<_>>>()
            //     });

            //     let segments = fut.await?;

            //     self.state = State::GenerateSegments(segments);
            // }
            State::SerializedSegments {
                serialized_data,
                segments,
                summary,
            } => {
                let ctx = self.ctx.clone();
                let dal = self.dal.clone();

                Self::write_segments(ctx, dal, serialized_data).await?;

                self.state = State::Output { segments, summary };
            }
            _ => return Err(ErrorCode::Internal("It's a bug.")),
        }
}

[Xuanwo]

More deeper, this call:

// Self::write_segments(ctx, dal, serialized_data).await?;

[ClSlaid]

Seems we need to refactor SegmentsIO::read_segments to make it pass:

The Location now is just an alias for (String, FormatVersion), maybe changing it to a newtype style could solve it.

pub struct Location(String, FormatVersion)

But it will introduce much more refactor, so if it does solve the problem, I'll make an independent PR for the introduced refactor.

[Xuanwo]

Ok, I got it:

pub async fn try_join_futures<Fut>(
    ctx: Arc<dyn TableContext>,
-    futures: impl IntoIterator<Item = Fut>,
+    futures: Vec<Fut>,
    thread_name: String,
) -> Result<Vec<Fut::Output>>

[Xuanwo]

But it will introduce much more refactor, so if it does solve the problem, I'll make an independent PR for the introduced refactor.

I tried locally, and not fixed.

[Xuanwo]

All warnings in hms_patched.rs can be ignored here:

https://github.com/datafuselabs/databend/blob/231a1e9f7f7ca6e4b1da740be3a8013331b93963/src/query/storages/hive/hive-meta-store/src/lib.rs#L15-L23

13465 |       ()
      |       ^^ help: remove the final `()`
      |
      = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#unused_unit

[Xuanwo]

error: redundant field names in struct initialization
     --> /workspace/target/debug/build/common-hive-meta-store-d2e71787906d9704/out/hms_patched.rs:24315:72
      |
24315 |         let call_args = ThriftHiveMetastoreGetTableNamesByFilterArgs { dbname: dbname, filter: filter, max_tables: max_tables };
      |                                                                        ^^^^^^^^^^^^^^ help: replace it with: `dbname`
      |
      = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#redundant_field_names

clippy is so strict (

[ClSlaid]

clippy is so strict (

Code generator could be improved in the future, let's allow this temporarily.

[Xuanwo]

Code generator could be improved in the future, let's allow this temporarily.

We don't need to improve this part, they are not maintained by human. Disable all lint for generated code LGTM.

[Xuanwo]

All check passed! Thanks for the hard work of @ClSlaid