fix #6164: Reduce format failure for non default `imports_granularity` #6165

kenoss · 2024-05-19T18:55:39Z

This patch reduces format failure for non default imports_granularity and correct the behavior around some edge cases.

Supports too long line of use ... that contains {}.
Fixes the width calculation of too long lines of use ....

kenoss · 2024-05-19T19:01:13Z

tests/target/issue_3033.rs

-use dom::bindings::codegen::Bindings::BluetoothRemoteGATTServerBinding::BluetoothRemoteGATTServerBinding::
-    BluetoothRemoteGATTServerMethods;
+use dom::bindings::codegen::Bindings::BluetoothRemoteGATTServerBinding::
+    BluetoothRemoteGATTServerBinding::BluetoothRemoteGATTServerMethods;


Exceeds 100 chars. It was preserved because the format failed.

Given that this changes existing behavior these changes need to be version gated

Ah, sorry. I missed the comment. Fixed.

ytmimi

I appreciate you jumping in to work on this, though I'm hoping we can find a way to refactor and simplify the implementation.

I also want to be mindful of your time and say up front that it might be a while before I'm able to get around to a follow up review on this.

ytmimi · 2024-05-21T00:39:14Z

src/imports.rs

+            // Try stacking the next segment, e.g. `use prev::next_segment::...` and
+            // `use prev::{next, segment}`.
+            let can_stack_with_constraint = (|| {
+                // If the segment follows `use ` or `{`, force to consume the segment with overflow.
+                if is_first {
+                    let mut chunk = segment.rewrite(context, shape.infinite_width())?;
+                    let next_shape = if iter.peek().is_some() {
+                        chunk.push_str("::");
+                        shape.offset_left_maybe_overflow(chunk.len())
+                    } else {
+                        shape.clone()
+                    };
+                    Some((chunk, next_shape))
+                } else {
+                    // If the segment follows `use ` or newline, allow overflow by "{".
+                    let s = if prev_is_allow_overflow
+                        && matches!(segment.kind, UseSegmentKind::List(_))
+                    {
+                        shape.add_width(1)
+                    } else {
+                        shape.clone()
+                    };
+                    let mut chunk = segment.rewrite(context, s)?;
+                    let next_shape = match iter.peek().map(|s| &s.kind) {
+                        Some(UseSegmentKind::List(_)) => {
+                            chunk.push_str("::");
+                            let ret = shape.offset_left(chunk.len())?;
+                            // Ensure that there is a room for the next "{".
+                            ret.offset_left(1)?;
+                            ret
+                        }
+                        Some(_) => {
+                            chunk.push_str("::");
+                            shape.offset_left(chunk.len())?
+                        }
+                        None => shape.clone(),
+                    };
+                    Some((chunk, next_shape))
+                }
+            })();
+            match can_stack_with_constraint {
+                Some((chunk, next_shape)) => {
+                    result.push_str(&chunk);
+                    shape = next_shape;
+                    prev_is_allow_overflow = is_first;
+                    is_first = false;
+                }
+                // If the next segment exceeds the given width, continue with newline.
+                None => {
+                    let segment_str = segment.rewrite(context, shape)?;
+                    let mut chunk = format!(
+                        "{}{}",
+                        " ".repeat(shape.indent.block_indent + 4),
+                        segment_str
+                    );
+                    if iter.peek().is_some() {
+                        chunk.push_str("::");
+                    }
+                    result.push_str("\n");
+                    result.push_str(&chunk);
+                    shape = shape_top_level.offset_left_maybe_overflow(segment_str.len());
+                    prev_is_allow_overflow = true;
+                }


I haven't looked at this in too much detail, but my initial thought is that there is a lot of nested logic here that makes this hard to wrap my head around. I think I have a high level understanding of what the code is trying to do, but I'd love to see if you can refactor and simplify the implementation.

Let me think...

Refactored. Looking ahead next + retry.

ytmimi · 2024-05-21T00:39:49Z

src/shape.rs

+    pub(crate) fn add_width(&self, width: usize) -> Shape {
+        Shape {
+            width: self.width + width,
+            ..*self
+        }
+    }


Can you explain a little more why we needed to implement an add_width method on the Shape?

Consider this case:

// With max width 40 // | use long_segment_loooooooooooooooooong::{Foo, Bar}; // should be formated to use long_segment_loooooooooooooooooong::{ Foo, Bar, };

Look at the UseTree::rewrite(), especially

let s = if prev_is_allow_overflow && matches!(segment.kind, UseSegmentKind::List(_)) { shape.add_width(1) } else { shape.clone() };

Let's trace the behavior of the format above.

committed = "use " // 35 = 40 - (4 + 1) shape = {indent: 0, offset: 4, width: 35} // len = 34 segment = "long_segment_loooooooooooooooooong" // 34 + 2 exceeds 35, but the above segment should be stacked to the current line because this is the first segment committed = "use long_segment_loooooooooooooooooong::" // with overflow. shape = {indent: 0, offset: 40, width: 0} prev_is_allow_overflow = true // TIMING1 segment = "{Foo, Bar}" // Here, we must secure the room for "{" since `prev_is_allow_overflow == true`, or `segment.rewrite(context, shape)?` fails. s = {indent: 0, offset: 40, width: 1} committed = "use long_segment_loooooooooooooooooong::{\n Foo,\n Bar,\n}"

Alternative options:

Use segment.rewrite(context, shape.infinite_width())?; at the TIMING1 instead.

We can't do that because it produces

use long_segment_loooooooooooooooooong::{Foo, Bar};

Add a variant of UseSegment::rewirte() that can handle yet another argument should_commit_open_brace_even_if_no_width.

It's not simple compared to adding Shape::add_width()...

So, I added Shape::add_width().

ytmimi · 2024-05-21T00:41:21Z

src/shape.rs

+    pub(crate) fn offset_left_maybe_overflow(&self, width: usize) -> Shape {
+        self.add_offset(width).saturating_sub_width(width)
+    }


Similarly, can you explain what offset_left_maybe_overflow is for?

Similar. There are the cases that we need to allow overflow temporarily. We can proceed Shape with this method without fail.

kenoss · 2024-05-21T14:06:24Z

Thank you for quick look!

I'll rethink and refactor. I'll mention once I've done.

kenoss · 2024-05-23T13:20:02Z

@ytmimi Could you have a look? Thanks!

ytmimi · 2024-05-24T14:52:49Z

@ytmimi Could you have a look? Thanks!

I really appreciate you taking the time to make these changes. As I mentioned in #6165 (review) I likely won't have time to re-review this for a little while, and I'll follow up when I can

kenoss · 2024-05-24T17:34:22Z

np. Take your time.

kenoss · 2024-08-20T17:13:30Z

Rebased and conflict resolved. Updated using RewriteResult/RewriteError.

kenoss · 2024-08-20T17:14:08Z

This is a friendly ping.

Three months passed.

@ytmimi Could you have a look? Let me know if I can help.

ytmimi · 2024-08-20T17:29:29Z

@kenoss I know it's been some time, but I unfortunately haven't had the bandwidth. Right now my main focus is getting rustfmt ready for the upcoming edition release, mentoring the GSoC project, triaging issues, and working on smaller bugfixes. I appreciate your patience on this.

kenoss · 2024-08-20T18:19:10Z

I understand your situation. (But... I wonder why don't you add more reviewers? It looks other PRs also are stuck with review.)

OK. Then, I'll pause to make this PR fresh. Let me know when you can review. Then, I'll rebase it.

``` $ RUSTFMT="./target/debug/rustfmt" cargo run --bin cargo-fmt -- --manifest-path ../../kenoss/sabiniwm/Cargo.toml ``` with rust-lang/rustfmt#6165.

…anularity` This patch reduces format failure for non default `imports_granularity` and correct the behavior around some edge cases. - Supports too long line of `use ...` that contains `{}`. - Fixes the width calculation of too long lines of `use ...`.

… do cargo run --bin rustfmt -- $f ; done

kenoss · 2024-12-17T16:30:00Z

@ytmimi Friendly ping in 2024 year end. I happened to notice that you are reviewing some PRs. Do you mind to take time to review this PR?

I rebased it onto near HEAD. (Commit 8a2c073 is not working on my environment. Rebased to HEAD~.)

I replaced .offset_left() of original PR to .offset_left_opt(). (See commit 1d1fb66.)
It looks OK as this path is not yet properly result-fied. I'm willing to make use results in this path in another PR if you have review bandwidth.

Thanks.

ytmimi · 2024-12-18T21:35:55Z

@kenoss Thanks for taking the time to rebase the PR. At the moment this is a lower priority item for the team and I don't anticipate I'll get around to reviewing the implementation before the end of the year. It's hard to say when exactly I'll be able to revisit this one, but I'll be sure to reach out when I do.

kenoss · 2024-12-21T21:26:54Z

Thank you for taking time!

FYI, it's a blocker for me to upstream it because I need it for CI. I don't want to diverge it.

chansuke · 2024-12-27T14:06:30Z

src/imports.rs

+                let reserved_room_for_brace = match next_segment.map(|s| &s.kind) {
+                    Some(UseSegmentKind::List(_)) => "{".len(),
+                    _ => 0,
+                };


imho: It would be helpful to add a comment explaining why reserved_room_for_brace is calculated?

Added. Thanks!

chansuke

I’ve reviewed your changes, including the test cases, and overall, the changes look good to me.

kenoss · 2024-12-28T14:59:32Z

Thanks! Comment added. All review comments resolved. (I'm not familiar with GitHub and a way of this repository. Which should I or you do Resolve conversation? I'll leave them.)

rustbot added the pr-not-reviewed label May 19, 2024

kenoss commented May 19, 2024

View reviewed changes

ytmimi requested changes May 21, 2024

View reviewed changes

kenoss force-pushed the fix-6164 branch from a68dffe to 19e453d Compare May 23, 2024 13:01

kenoss requested a review from ytmimi May 24, 2024 14:45

ytmimi added pr-follow-up-review-pending and removed pr-not-reviewed labels May 24, 2024

kenoss force-pushed the fix-6164 branch 2 times, most recently from 03c8503 to 435e3c6 Compare June 1, 2024 18:44

kenoss force-pushed the fix-6164 branch from 435e3c6 to f150768 Compare June 4, 2024 17:21

kenoss force-pushed the fix-6164 branch 2 times, most recently from 24de098 to 8b634f2 Compare June 25, 2024 12:37

kenoss force-pushed the fix-6164 branch from 8b634f2 to 6729294 Compare July 10, 2024 14:41

kenoss force-pushed the fix-6164 branch from 6729294 to 667b3dc Compare July 20, 2024 16:03

kenoss force-pushed the fix-6164 branch from 667b3dc to 727179e Compare August 20, 2024 17:11

kenoss added a commit to kenoss/sabiniwm that referenced this pull request Nov 27, 2024

wip: cargo fmt with imporved rustfmt

ec9a98e

``` $ RUSTFMT="./target/debug/rustfmt" cargo run --bin cargo-fmt -- --manifest-path ../../kenoss/sabiniwm/Cargo.toml ``` with rust-lang/rustfmt#6165.

kenoss added 2 commits December 18, 2024 00:02

fix rebase

1d1fb66

kenoss force-pushed the fix-6164 branch from a71be3a to 73fefd9 Compare December 17, 2024 15:37

for f in $(git diff 9f8fcc2 --name-only | pcregrep 'target/.*_2024');…

0eecc48

… do cargo run --bin rustfmt -- $f ; done

kenoss force-pushed the fix-6164 branch from 73fefd9 to 0eecc48 Compare December 17, 2024 15:57

chansuke reviewed Dec 27, 2024

View reviewed changes

chansuke approved these changes Dec 27, 2024

View reviewed changes

comment: reserved_room_for_brace

0c5ee8e

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fix #6164: Reduce format failure for non default `imports_granularity` #6165

fix #6164: Reduce format failure for non default `imports_granularity` #6165

kenoss commented May 19, 2024 •

edited

Loading

kenoss May 19, 2024

ytmimi May 21, 2024

kenoss Aug 20, 2024

ytmimi left a comment

ytmimi May 21, 2024

kenoss May 21, 2024

kenoss May 23, 2024

ytmimi May 21, 2024

kenoss May 21, 2024

ytmimi May 21, 2024

kenoss May 21, 2024 •

edited

Loading

kenoss commented May 21, 2024

kenoss commented May 23, 2024

ytmimi commented May 24, 2024 •

edited

Loading

kenoss commented May 24, 2024

kenoss commented Aug 20, 2024

kenoss commented Aug 20, 2024

ytmimi commented Aug 20, 2024

kenoss commented Aug 20, 2024

kenoss commented Dec 17, 2024

ytmimi commented Dec 18, 2024

kenoss commented Dec 21, 2024

chansuke Dec 27, 2024 •

edited

Loading

kenoss Dec 28, 2024

chansuke left a comment

kenoss commented Dec 28, 2024

fix #6164: Reduce format failure for non default imports_granularity #6165

Are you sure you want to change the base?

fix #6164: Reduce format failure for non default imports_granularity #6165

Conversation

kenoss commented May 19, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ytmimi left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

kenoss May 21, 2024 • edited Loading

Choose a reason for hiding this comment

kenoss commented May 21, 2024

kenoss commented May 23, 2024

ytmimi commented May 24, 2024 • edited Loading

kenoss commented May 24, 2024

kenoss commented Aug 20, 2024

kenoss commented Aug 20, 2024

ytmimi commented Aug 20, 2024

kenoss commented Aug 20, 2024

kenoss commented Dec 17, 2024

ytmimi commented Dec 18, 2024

kenoss commented Dec 21, 2024

chansuke Dec 27, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

chansuke left a comment

Choose a reason for hiding this comment

kenoss commented Dec 28, 2024

fix #6164: Reduce format failure for non default `imports_granularity` #6165

fix #6164: Reduce format failure for non default `imports_granularity` #6165

kenoss commented May 19, 2024 •

edited

Loading

kenoss May 21, 2024 •

edited

Loading

ytmimi commented May 24, 2024 •

edited

Loading

chansuke Dec 27, 2024 •

edited

Loading