Codestin Search App

alamb · 2024-03-29T14:39:17Z

Which issue does this PR close?

Related to #9278, follow on to #9292

Rationale for this change

I want DataFusion to compile quickly and be a nice development experience, so the faster the compilation / fewer dependencies the better in general.

I have been wondering why compiling datafusion takes so long and requires compiling crates like gix (which is an implementation of git) #9844 (review) got me inspired to look into this:

...
   Compiling gix-attributes v0.20.1
   Compiling gix-credentials v0.22.0
   Compiling gix-ignore v0.9.1
...

It turns out the dependencies were needed by the depcheck test, added in #9292. There is no need to compile 200 extra crates everytime someone runs cargo test, I think it is enough if this tool runs as part of CI.

What changes are included in this PR?

move the depcheck test into its own binary (not part of the main workspace) and only run that as part of CI

Are these changes tested?

There is a new CI job and I manually tested that it catches issues

Manual Testing Details

Negative Test

I locally introduced a circular dependency:

diff --git a/datafusion/functions/Cargo.toml b/datafusion/functions/Cargo.toml
index 3ae306101..cde0742fc 100644
--- a/datafusion/functions/Cargo.toml
+++ b/datafusion/functions/Cargo.toml
@@ -83,6 +83,7 @@ uuid = { version = "1.7", features = ["v4"], optional = true }

 [dev-dependencies]
 criterion = "0.5"
+datafusion = { workspace = true }
 rand = { workspace = true }
 rstest = { workspace = true }
 tokio = { workspace = true, features = ["macros", "rt", "sync"] }

The test now fails

 andrewlamb@Andrews-MacBook-Pro:~/Software/arrow-datafusion2$ (cd dev/depcheck && cargo run)
    Compiling depcheck v0.0.0 (/Users/andrewlamb/Software/arrow-datafusion2/dev/depcheck)
     Finished dev [unoptimized + debuginfo] target(s) in 0.49s
      Running `target/debug/depcheck`
 Checking for circular dependencies in /Users/andrewlamb/Software/arrow-datafusion2/Cargo.toml
 thread 'main' panicked at src/main.rs:84:9:
 circular dependency detected from datafusion to self via one of {"datafusion-functions", "datafusion-execution", "datafusion-common-runtime", "datafusion-expr", "datafusion-common"}
 note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

When I remove the local dependency, the test passes:

andrewlamb@Andrews-MacBook-Pro:~/Software/arrow-datafusion2$ (cd dev/depcheck && cargo run)
   Compiling depcheck v0.0.0 (/Users/andrewlamb/Software/arrow-datafusion2/dev/depcheck)
    Finished dev [unoptimized + debuginfo] target(s) in 0.50s
     Running `target/debug/depcheck`
Checking for circular dependencies in /Users/andrewlamb/Software/arrow-datafusion2/Cargo.toml
No circular dependencies found

Are there any user-facing changes?

Faster compilation times (200 less crates to compile!)

Previously, running cargo test required compiling 838 crates!

With this change we are down to 647 crates (still a crazy number)

It also seems to make the CI jobs significantly faster:

before this PR: 11m4s
after this PR: 915s

alamb · 2024-03-29T15:14:35Z

@andygrove I wonder if you have any thoughts / reactions to running the circular dependency check as its own CI job?

comphead

Thanks @alamb that looks like addressed to circular dependencies, but how it helps to improve the compile time?

comphead · 2024-03-29T16:15:05Z

windows machine still slow as hell, and it is damned slow there because of compilation latency as well

alamb · 2024-03-30T12:44:52Z

 [dev-dependencies]
 async-trait = { workspace = true }
 bigdecimal = { workspace = true }
-cargo = "0.78.1"


By removing this line, cargo test now requires 200 fewer crates to compile

alamb · 2024-03-30T12:45:47Z

Thanks @alamb that looks like addressed to circular dependencies, but how it helps to improve the compile time?

Sorry -- I should have made it clear, this line is what helps https://github.com/apache/arrow-datafusion/pull/9865/files#r1545339002

Note that the circular dependency check was already present, I just moved it out of the main workspace because the dependency requirements were massive

alamb · 2024-03-30T12:46:11Z

+      - name: Check dependencies
+        run: |
+          cd dev/depcheck
+          cargo run 


Note this check was previously covered by cargo test -p datafusion

alamb · 2024-03-30T12:46:51Z

+/// checking the dependency graph.
+///
+/// See https://github.com/apache/arrow-datafusion/issues/9278 for more details
+fn main() -> CargoResult<()> {


note this check is not new, it just now runs in a separate binary which requires many fewer crates in the main workspace

Co-authored-by: Andrew Lamb <[email protected]>

comphead

lgtm thanks @alamb, macos timing is good but we probably need to investigate why compile time is so slow on windows machines

alamb · 2024-03-30T20:17:21Z

lgtm thanks @alamb, macos timing is good but we probably need to investigate why compile time is so slow on windows machines

I agree the windows test takes too long.

On the other hand I think this PR actually improves the situation:

main takes 44 mins: https://github.com/apache/arrow-datafusion/actions/runs/8492129363/job/23264805971
This branch takes 40 mins: https://github.com/apache/arrow-datafusion/actions/runs/8492313070/job/23265167048?pr=9865

alamb · 2024-03-30T20:17:40Z

Merging this in to improve the Dev experience. Thanks @comphead for the review

…ile) (apache#9865) * Minor: Move depcheck out of main datafusion test * Update dev/depcheck/README.md Co-authored-by: Andrew Lamb <[email protected]> --------- Co-authored-by: comphead <[email protected]>

andygrove · 2024-04-01T19:33:42Z

Thanks for catching this @alamb

github-actions Bot added the core Core DataFusion crate label Mar 29, 2024

alamb added the development-process Related to development process of DataFusion label Mar 29, 2024

Minor: Move depcheck out of main datafusion test

f61a9c2

alamb force-pushed the alamb/move_dep_check branch from 935c53f to f61a9c2 Compare March 29, 2024 14:44

github-actions Bot removed the development-process Related to development process of DataFusion label Mar 29, 2024

alamb changed the title ~~Minor: Move depcheck out of datafusion crate (200 less crates for compilation)~~ Minor: Move depcheck out of datafusion crate (200 less crates to compile) Mar 29, 2024

alamb added the development-process Related to development process of DataFusion label Mar 29, 2024

alamb marked this pull request as ready for review March 29, 2024 15:14

comphead reviewed Mar 29, 2024

View reviewed changes

Comment thread dev/depcheck/README.md Outdated

comphead reviewed Mar 29, 2024

View reviewed changes

alamb commented Mar 30, 2024

View reviewed changes

Update dev/depcheck/README.md

e5e12c2

Co-authored-by: Andrew Lamb <[email protected]>

github-actions Bot removed the development-process Related to development process of DataFusion label Mar 30, 2024

comphead approved these changes Mar 30, 2024

View reviewed changes

alamb merged commit 2159d82 into apache:main Mar 30, 2024

alamb deleted the alamb/move_dep_check branch March 30, 2024 20:17

This was referenced Apr 16, 2024

Minor: only trigger dependency check on changes to Cargo.toml #10099

Merged

Prepare 37.1.0 release (on branch-37) #10128

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Minor: Move depcheck out of datafusion crate (200 less crates to compile)#9865

Minor: Move depcheck out of datafusion crate (200 less crates to compile)#9865
alamb merged 2 commits into
apache:mainfrom
alamb:alamb/move_dep_check

alamb commented Mar 29, 2024 •

edited

Loading

Uh oh!

alamb commented Mar 29, 2024

Uh oh!

Uh oh!

comphead left a comment

Uh oh!

comphead commented Mar 29, 2024

Uh oh!

alamb Mar 30, 2024

Uh oh!

alamb commented Mar 30, 2024

Uh oh!

alamb Mar 30, 2024

Uh oh!

Uh oh!

alamb Mar 30, 2024 •

edited

Loading

Uh oh!

comphead left a comment

Uh oh!

alamb commented Mar 30, 2024

Uh oh!

alamb commented Mar 30, 2024

Uh oh!

andygrove commented Apr 1, 2024

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

alamb commented Mar 29, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Which issue does this PR close?

Rationale for this change

What changes are included in this PR?

Are these changes tested?

Negative Test

Are there any user-facing changes?

Uh oh!

alamb commented Mar 29, 2024

Uh oh!

Uh oh!

comphead left a comment

Choose a reason for hiding this comment

Uh oh!

comphead commented Mar 29, 2024

Uh oh!

alamb Mar 30, 2024

Choose a reason for hiding this comment

Uh oh!

alamb commented Mar 30, 2024

Uh oh!

alamb Mar 30, 2024

Choose a reason for hiding this comment

Uh oh!

Uh oh!

alamb Mar 30, 2024 • 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!

comphead left a comment

Choose a reason for hiding this comment

Uh oh!

alamb commented Mar 30, 2024

Uh oh!

alamb commented Mar 30, 2024

Uh oh!

andygrove commented Apr 1, 2024

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

alamb commented Mar 29, 2024 •

edited

Loading

alamb Mar 30, 2024 •

edited

Loading