Use a flat Python AST in Ruff and ty by MichaReiser · Pull Request #29817 · astral-sh/ruff

MichaReiser · 2026-10-29T17:07:19Z

Summary

Ruff and ty parse Python into 24-byte rows linked by four-byte node IDs. Typed nodes borrow fields from the parsed module, lint fixes build replacement syntax in an arena, and ty stores row IDs without a separate node-numbering pass. Python AST storage and traversal

Review order

The PR starts at the boxed mapping-pattern layout in #29889. The first five commits migrate the AST and its consumers. The remaining 41 isolate subsequent changes. Both Ruff and ty build at the fifth commit. Boxed parent 0f8d2da79c.

Replace boxed Python AST nodes with named arena views. 466cd822cc
Parse and cache Python modules in a flat AST. a19708bfa5
Run Ruff lint rules and build fixes from arena nodes. 16cfcafc3a
Format Python directly from arena nodes. 1863efdc04
Read the flat AST in semantic and IDE queries. 37db863a96

Performance

The boxed and flat revisions both contain #29916, which resolves a call’s qualified name once for the applicable lint rules. Both use source offsets for Ruff’s name-binding map. These measurements predate the comparison optimizations at the base of the stack. The measured flat executable used Git tree 2f1051bec9. Boxed 48b74912c7 · Flat d28e863169.

Ruff lint

The five repository fixtures run with the default rules, all stable rules and all rules including preview. The benchmark parses each module before the timed lint call. Each row sums the same 15 jobs. Expression checking runs inside check_ast, so these scopes overlap.

Scope	Boxed	Flat	Change
Lint instructions before parsed-module destruction	147,763,734	152,327,272	+3.09%
Instructions in `check_ast`	111,651,743	115,944,765	+3.85%
Instructions in expression checks and their callees	54,240,752	59,145,713	+9.04%
Instructions to drop the parsed module	4,299,969	81,527	−98.10%
All timed lint instructions	152,063,703	152,408,799	+0.23%
Simulated L1 instruction-cache misses in `check_ast`	4,797,539	5,469,977	+14.02%
Simulated L1 instruction-cache misses in expression checks	2,848,002	3,269,487	+14.80%
Simulated L1 instruction-cache misses in the full lint call	4,888,425	5,558,886	+13.72%

Both checkers dispatched 26,613 expressions. Among rules with matching native call counts on every fixture, 175 call rules account for 2,120,584 additional instructions and 17 comparison rules for 902,616. The shared callee-name resolution itself ran 2,925 times in each build and cost 22,413 additional instructions. Inlining prevents a separate count for every rule.

For client.send(), S201 checks whether the method name is run, and S610 checks whether it is extra. Each rule receives the decoded call fields. Each still reads the attribute’s identifier. The boxed attribute embeds the identifier. The flat attribute holds a four-byte ID, so the getter loads the ID and resolves the identifier row. S201 · S610

Lint instructions by fixture and rule selection, excluding parsed-module destruction

Workload	Boxed	Flat	Change
`numpy/globals.py` · Default	325,415	343,804	+5.65%
`numpy/globals.py` · All stable	842,195	870,555	+3.37%
`numpy/globals.py` · All with preview	967,922	1,005,134	+3.84%
`numpy/ctypeslib.py` · Default	3,579,185	3,695,833	+3.26%
`numpy/ctypeslib.py` · All stable	7,267,614	7,479,487	+2.92%
`numpy/ctypeslib.py` · All with preview	8,940,307	9,182,057	+2.70%
`unicode/pypinyin.py` · Default	1,051,677	1,108,285	+5.38%
`unicode/pypinyin.py` · All stable	2,975,627	3,079,842	+3.50%
`unicode/pypinyin.py` · All with preview	3,467,273	3,574,212	+3.08%
`pydantic/types.py` · Default	7,508,874	7,917,573	+5.44%
`pydantic/types.py` · All stable	14,112,344	14,686,003	+4.06%
`pydantic/types.py` · All with preview	17,482,955	18,055,002	+3.27%
`large/dataset.py` · Default	14,097,604	14,626,100	+3.75%
`large/dataset.py` · All stable	28,556,834	29,358,466	+2.81%
`large/dataset.py` · All with preview	36,587,908	37,344,919	+2.07%

The first complete flat migration used 161,812,123 instructions before destruction (+9.51% versus boxed). The subsequent commits reduced this to 152,327,272 (−5.86% versus the first flat version). 85f8507b0e

Lint instructions after each of the final five changes

Flat version	Before destruction	Change from preceding row	Simulated L1I misses in the checker
Pass borrowed AST fields to lint rules. `117e774f6f`	153,367,040	—	5,508,683
Reuse decoded call fields in comprehension rules. `b2ba50093f`	153,191,177	−175,863	0
Read method parameters after matching new. `7436ae13a8`	153,180,900	−10,277	0
Keep keyword and positional searches together in call argument lookup. `4e0503771c`	153,030,973	−149,927	0
Reserve room for small expressions in the AST builder. `c72ee7afd8`	152,869,183	−161,790	0
Iterate parameters from the stored list of node IDs. `d28e863169`	152,327,272	−541,911	5,469,977

Compare the two boxed layout PRs and the flat AST on the same 15 lint jobs

Revision	Instructions before destruction	Change from boxed main	Simulated L1I misses in the checker	Instructions including destruction
Boxed main `d71685096a`	148,090,315	—	4,784,764	152,385,916
Boxed with DictItem `fd8861c839`	147,685,674	−0.27%	4,815,095	151,983,841
Boxed with both layout PRs `48b74912c7`	147,763,734	−0.22%	4,797,539	152,063,703
First complete flat AST `85f8507b0e`	161,812,123	+9.27%	5,780,218	161,890,527
Final flat AST `d28e863169`	152,327,272	+2.86%	5,469,977	152,408,799

Complete Ruff and ty commands

These profiling builds run the actual CLI commands with one worker. The before-destruction columns subtract the instructions in the parsed-module destructor from the same process. Both variants produced identical Ruff diagnostics, edits, formatter output and ty diagnostics.

Command	Boxed instructions before destruction	Flat instructions before destruction	Change before destruction	Change including destruction
Ruff lint, CPython default	2,677,589,506	2,769,739,054	+3.44%	+0.40%
Ruff lint, Jinja2 default	169,500,048	177,088,072	+4.48%	+1.53%
Ruff lint, Pygments all stable	2,886,686,226	2,899,463,571	+0.44%	−0.61%
Ruff format, stdlib	10,399,807,146	10,369,784,860	−0.29%	−1.07%
ty, Jinja2	2,228,591,924	2,168,164,728	−2.71%	−2.93%
ty, Pygments	7,969,608,697	7,792,314,876	−2.22%	−2.60%
ty, stdlib	18,878,516,926	18,495,761,323	−2.03%	−2.45%

The lint projects contain 568 CPython files, 25 Jinja2 files and 321 Pygments files. The formatter row uses the median whole-command result from three runs per layout on 568 files. Each ty project uses one matched pair.

Whole-command instructions and parsed-module destruction

Command	Boxed total	Flat total	Boxed destruction	Flat destruction
Ruff lint, CPython default	2,761,197,642	2,772,240,080	83,608,136	2,501,026
Ruff lint, Jinja2 default	174,542,133	177,220,223	5,042,085	132,151
Ruff lint, Pygments all stable	2,917,956,791	2,900,177,927	31,270,565	714,356
Ruff format, stdlib	10,483,958,267	10,372,241,064	84,151,121	2,456,204
ty, Jinja2	2,233,689,913	2,168,337,276	5,097,989	172,548
ty, Pygments	8,000,727,268	7,792,962,969	31,118,571	648,093
ty, stdlib	18,963,120,293	18,497,602,667	84,603,367	1,841,344

Complete Ruff lint commands compared directly with boxed main

Plain boxed main, the boxed parent with both layout changes and the flat head use the same inputs. All three produced byte-identical CLI output.

Project	Boxed main before destruction	Boxed parent before destruction	Flat before destruction	Flat versus main
CPython default	2,679,446,632	2,677,589,506	2,769,739,054	+3.37%
Jinja2 default	169,664,484	169,500,048	177,088,072	+4.38%
Pygments all stable	2,892,102,210	2,886,686,226	2,899,463,571	+0.25%

Hosted CodSpeed

CodSpeed compared flat d28e863169 with its boxed parent 48b74912c7. Its three reported regressions are CPU simulations of linting numpy/globals.py. Across all instruments it reported 167 improvements, 118 unchanged benchmarks and 14 skipped. The report flags different runner hardware for 127 benchmarks with significant changes. Hosted results

The local columns repeat the corresponding repository lint benchmark on the same fixture, including destruction of the parsed module. CodSpeed's CPU simulation models both executed instructions and cache-miss penalties. CPU simulation

NumPy lint rule selection	Hosted CPU efficiency	Boxed instructions	Flat instructions	Boxed L1 instruction-cache misses	Flat L1 instruction-cache misses
Default	−5.60%	333,041	344,343	12,123	13,890
All stable	−5.09%	849,821	871,094	22,767	25,380
All with preview	−6.73%	975,548	1,005,673	24,551	28,066

In the preview benchmark, check_ast accounts for 3,328 of the 3,515 additional L1 instruction-cache misses. L1 data-read misses fall from 2,636 to 2,408.

How instruction-cache capacity affects the same file in the Ruff CLI

These runs collect instructions inside check_path in an isolated, single-file CLI invocation. Each pair produced identical diagnostics. The cache simulations use the same profiling binaries, eight-way instruction caches with 64-byte lines, and fixed data and last-level caches.

Rule selection	Distinct machine instructions executed, boxed	Distinct machine instructions executed, flat	Extra flat L1 instruction misses, 32 KiB	Extra flat L1 instruction misses, 256 KiB	Extra flat L1 instruction misses, 4 MiB
Default	47,438	53,139	1,947	710	516
All stable	105,766	115,249	3,216	1,759	847
All with preview	116,512	126,911	3,630	1,927	882

Increasing the cache from 32 KiB to 4 MiB removed 74–76% of the extra misses. Holding capacity at 32 KiB and making the cache fully associative left gaps of 1,903, 3,892 and 3,408 misses, respectively. The flat CLI executes a larger set of machine instructions, and the smaller cache repeatedly fetches some of that code. These measurements establish the effect of capacity in the simulator. Quantifying how much inlining contributes would require a separate comparison.

Review the 41 commits after the basic migration

Store compact AST IDs in Ruff’s semantic node table. 4671a24d28
Share the list boundary between comparison operators and operands. b28d6670e2
Read the operator of a single comparison directly from its row. 9197d82ec1
Cache recently interned Python identifier spellings while parsing. 5df7b12170
Reserve initial AST buffers from the source length. 0744f8273a
Estimate arena growth from the rows emitted during parsing. d727ad2b49
Reserve AST row zero so child IDs index their arena directly. 7ae6500979
Align Python AST rows to eight bytes. 03d630d026
Read and write AST row fields through a typed payload union. 8206b084bf
[ty] Resolve persisted AST references by parser row and remove the node-index pass. b07cd8977a
Record function body rows for scope-aware return and yield checks. 18af1be6ea
Skip generator lint-rule walks when function rows contain no yield. f7c968677e
Keep ordinary parameter lists inline in AST rows. 24b2e68768
Store parameters in source order in the AST. d3ca11300a
Write parsed parameters into the arena directly from parser scratch storage. 0c6de85d0d
Store call arguments in source order. eb56d0de96
Store class pattern arguments in source order. a604949e30
Borrow remaining child lists during descendant traversal. 69ddb091c7
Borrow adjacent child IDs directly from AST rows. 6308082813
Store conditional expression fields in source order. 6d969ef5ba
[ty] Skip semantic reference walks when no AST keyword label matches. 7c32da470d
Construct child iterators from the known type of a borrowed node. b881497b3d
Use child endpoints and statement traversal in the formatter. 5a1a58efaf
Read expression fields directly when comparing AST nodes. 0d1aecd042
Read version-check literals through their node getters. ebe7c2f704
Write positional arguments immediately before keyword arguments in the arena. 1ac46df1cc
Keep the first assignment target on the stack. ffb5aee129
Keep the first parsed comparison on the stack. 30a2b7ce29
Forward indexed argument lookups to the source-order iterator. 2e9277139d
Bound speculative reservations for AST rows and side tables. 7f761cdc1c
Decode Python literals directly into the AST buffers. c1dc702ec3
Read formatter predicates through typed AST getters. 2dc5c50184
Read lint rule inputs through typed AST getters. 8b85abcec6
[ty] Read inference and IDE inputs through typed AST getters. 86268ac880
Keep borrowed AST nodes in Ruff’s semantic model. 77b925149c
Pass borrowed AST fields to lint rules. 0047cd0beb
Reuse decoded call fields in comprehension rules. 8e9db15765
Read method parameters after matching new. c423b19c27
Keep keyword and positional searches together in call argument lookup. 97b87a9352
Reserve room for small expressions in the AST builder. 1717bf5c83
Iterate parameters from the stored list of node IDs. 1890f9a1bd

Test coverage

Parser fixtures and the existing lint and formatter suites exercise the arena through its consumers. Traversal tests cover source order for valid and recovered syntax. ty’s semantic tests cover resolving stored node IDs, including in quoted annotations.

A boxed-versus-flat check over 1,731 fixture and rule-setting pairs matched all 66,678 diagnostics and 29,731 fixes. A separate check exercised all 70 rules in the shared callee dispatch using direct, aliased, unresolved and shadowed callees and per-file ignores. All 228 diagnostics and 138 fixes matched, including applicability and edit text.

astral-automations-bot · 2026-10-29T17:10:15Z

Typing conformance results

No changes detected ✅

Current numbers

The percentage of diagnostics emitted that were expected errors held steady at 98.60%. The percentage of expected errors that received a diagnostic held steady at 98.86%. The number of fully passing files held steady at 137/146.

Results for 0936621 · Workflow run

astral-automations-bot · 2026-10-29T17:11:36Z

Memory usage report

Summary

Project	Old	New	Diff	Outcome
flake8	38.07MB	33.28MB	-12.58% (4.79MB)	⬇️
trio	91.97MB	84.72MB	-7.88% (7.25MB)	⬇️
sphinx	185.42MB	170.70MB	-7.94% (14.72MB)	⬇️
prefect	455.47MB	426.37MB	-6.39% (29.09MB)	⬇️

Significant changes

Click to expand detailed breakdown

flake8

Name	Old	New	Diff	Outcome
`parsed_module`	14.47MB	9.88MB	-31.73% (4.59MB)	⬇️
`semantic_index`	9.27MB	9.07MB	-2.14% (202.97kB)	⬇️

trio

Name	Old	New	Diff	Outcome
`parsed_module`	22.00MB	15.12MB	-31.25% (6.87MB)	⬇️
`semantic_index`	18.57MB	18.20MB	-2.01% (382.01kB)	⬇️

sphinx

Name	Old	New	Diff	Outcome
`parsed_module`	44.49MB	30.53MB	-31.38% (13.96MB)	⬇️
`semantic_index`	35.06MB	34.30MB	-2.17% (777.89kB)	⬇️

prefect

Name	Old	New	Diff	Outcome
`parsed_module`	87.51MB	60.27MB	-31.12% (27.24MB)	⬇️
`semantic_index`	87.41MB	85.55MB	-2.13% (1.86MB)	⬇️

Results for 0936621 · Workflow run

astral-automations-bot · 2026-10-29T17:13:00Z

`ecosystem-analyzer` results

No diagnostic changes detected ✅

Full report with detailed diff · Timing results

Results for 0936621 · Workflow run

astral-automations-bot · 2026-10-29T17:18:32Z

`ruff-ecosystem` results

Linter (stable)

✅ ecosystem check detected no linter changes.

Linter (preview)

✅ ecosystem check detected no linter changes.

Formatter (stable)

✅ ecosystem check detected no format changes.

Formatter (preview)

✅ ecosystem check detected no format changes.

Linter (stable) · Linter (preview) · Formatter (stable) · Formatter (preview)

Results for 0936621 · Workflow run

astral-automations-bot · 2026-10-29T17:21:22Z

+    let rewritten = owned_pattern_to_expr(pattern.view(storage).to_owned_node());
+    let id = ast::flat::Lower::lower(&rewritten, storage);


[P3] Avoid retaining entire subtrees during pattern recovery

A malformed class pattern whose head is a sequence reaches this conversion at each nesting level. to_owned_node() copies the entire previously recovered subtree, and Lower::lower appends its literal bytes to the same Builder without reclaiming the originals. Nesting around one large string therefore retains memory proportional to literal size × nesting depth, allowing an attacker-controlled file in an untrusted language-server workspace to exhaust host memory. The previous implementation moved owned children. Construct the recovery nodes using existing child handles to avoid retaining these repeated copies.

codspeed · 2026-10-29T17:34:13Z

Merging this PR will regress 5 benchmarks

1 warning

⚠️ 112 benchmarks with a significant change ran on different hardware

Their base and head results were measured on machines with a different CPU, core count or memory, so the changes might not come from the code. Run base and head on the same runner type.

Learn more about Macro Runners

See affected benchmarks

⚡ 176 improved benchmarks
❌ 5 regressed benchmarks
✅ 113 untouched benchmarks
⏩ 14 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	`BASE`	`HEAD`	Efficiency
❌	CPU Simulation	`linter/all-with-preview-rules[numpy/globals.py]`	1.3 ms	1.4 ms	-7.68%
❌	CPU Simulation	`linter/default-rules[numpy/globals.py]`	617.4 µs	659.5 µs	-6.39%
❌	CPU Simulation	`linter/all-rules[numpy/globals.py]`	1.2 ms	1.2 ms	-6.2%
❌	CPU Simulation	`linter/default-rules[pydantic/types.py]`	8.5 ms	8.9 ms	-4.63%
❌	CPU Simulation	`linter/all-with-preview-rules[pydantic/types.py]`	17.6 ms	18.4 ms	-4.53%
⚡	Memory	`parser[numpy/globals.py]`	16.7 KB	11.4 KB	+47.28%
⚡	Memory	`parser[unicode/pypinyin.py]`	44.6 KB	33.8 KB	+31.84%
⚡	Memory	`parser[pydantic/types.py]`	281.7 KB	218.1 KB	+29.12%
⚡	Memory	`parser[large/dataset.py]`	645.4 KB	513.3 KB	+25.73%
⚡	Walltime	`auto_imports[namespace-package/repeated-request]`	2.8 ms	2.2 ms	+23.59%
⚡	Walltime	`auto_imports[regular-package/repeated-request]`	2.9 ms	2.3 ms	+23.37%
⚡	Memory	`parser[numpy/ctypeslib.py]`	134.9 KB	109.8 KB	+22.88%
⚡	Walltime	`auto_imports[regular-package/first-request]`	73.9 ms	61.6 ms	+19.99%
⚡	Walltime	`auto_imports[namespace-package/first-request]`	75.8 ms	63.7 ms	+18.85%
⚡	Memory	`ty_micro[pandas_tdd]`	26.6 MB	23.1 MB	+15.29%
⚡	Walltime	`auto_imports[regular-package/after-module-create]`	4 ms	3.5 ms	+15.14%
⚡	Memory	`ty_micro[very_large_tuple]`	7.4 MB	6.4 MB	+14.49%
⚡	Memory	`ty_micro[async_context_manager_recursive_protocol]`	7.8 MB	6.9 MB	+14.02%
⚡	Memory	`ty_micro[complex_constrained_attributes_3]`	7.6 MB	6.7 MB	+13.85%
⚡	Memory	`ty_micro[split_literal_union_1024]`	4.8 MB	4.2 MB	+13.66%
...	...	...	...	...	...

ℹ️ Only the first 20 benchmarks are displayed. Go to the app to view all benchmarks.

Tip

Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.

_{Comparing micha/flat-python-ast (0936621) with micha/boxed-mapping-pattern-items (3c2225d)}

14 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

dcreager · 2026-10-29T19:38:52Z

YES

…torage

…e arena

Reuse the expression or statement kind across checker dispatch and child traversal. Visitors that specialize a node can forward the kind they already matched. Pattern and type-parameter walkers use the same convention, and the exception-handler walker accepts its concrete node. Keep the capture name across pattern traversal so mapping patterns look up their rest capture once while preserving binding order.

MichaReiser added performance Potential performance improvement parser Related to the parser ty The ty type checker memory An issue regarding poor memory usage, or a PR improving memory usage labels Oct 29, 2026

MichaReiser temporarily deployed to automations October 29, 2026 17:07 — with GitHub Actions Inactive

MichaReiser temporarily deployed to automations October 29, 2026 17:21 — with GitHub Actions Inactive

astral-automations-bot Bot reviewed Oct 29, 2026

View reviewed changes

MichaReiser had a problem deploying to automations October 29, 2026 17:26 — with GitHub Actions Failure

MichaReiser had a problem deploying to automations October 29, 2026 17:47 — with GitHub Actions Failure

AlexWaygood changed the title ~~[ty] Use a flat Python AST in Ruff and ty~~ Use a flat Python AST in Ruff and ty Oct 29, 2026

AlexWaygood removed the ty The ty type checker label Oct 29, 2026

MichaReiser had a problem deploying to automations October 29, 2026 18:04 — with GitHub Actions Failure

MichaReiser had a problem deploying to automations October 29, 2026 18:34 — with GitHub Actions Failure

MichaReiser temporarily deployed to automations October 29, 2026 19:02 — with GitHub Actions Inactive

MichaReiser force-pushed the micha/flat-python-ast branch from bd231d4 to 59e0aa9 Compare October 29, 2026 19:19

MichaReiser temporarily deployed to automations October 29, 2026 19:20 — with GitHub Actions Inactive

MichaReiser temporarily deployed to automations October 29, 2026 19:34 — with GitHub Actions Inactive

MichaReiser temporarily deployed to automations October 29, 2026 20:19 — with GitHub Actions Inactive

MichaReiser temporarily deployed to automations October 29, 2026 20:41 — with GitHub Actions Inactive

MichaReiser force-pushed the micha/flat-python-ast branch from 46a2f59 to 8b89865 Compare October 29, 2026 21:13

MichaReiser had a problem deploying to automations October 29, 2026 21:13 — with GitHub Actions Error

MichaReiser had a problem deploying to automations October 29, 2026 21:31 — with GitHub Actions Error

MichaReiser temporarily deployed to automations October 29, 2026 21:44 — with GitHub Actions Inactive

MichaReiser force-pushed the micha/flat-python-ast branch from 8b89865 to 1d9bbf2 Compare October 29, 2026 23:51

MichaReiser added 30 commits November 4, 2026 17:51

Keep ordinary parameter lists inline in AST rows

66e7d7d

Store parameters in source order in the AST

8c5dd21

Write parsed parameters into the arena directly from parser scratch s…

36ce3cc

…torage

Store call arguments in source order

940934b

Store class pattern arguments in source order

d20b9d1

Borrow remaining child lists during descendant traversal

32674cb

Borrow adjacent child IDs directly from AST rows

44662d7

Store conditional expression fields in source order

73c8d02

[ty] Skip semantic reference walks when no AST keyword label matches

bbb288b

Construct child iterators from the known type of a borrowed node

690ee06

Use child endpoints and statement traversal in the formatter

ea04524

Read expression fields directly when comparing AST nodes

c954f4a

Read version-check literals through their node getters

44e2171

Write positional arguments immediately before keyword arguments in th…

07fc761

…e arena

Keep the first assignment target on the stack

5eb9bf2

Keep the first parsed comparison on the stack

1835c41

Forward indexed argument lookups to the source-order iterator

da7e949

Bound speculative reservations for AST rows and side tables

0f3c522

Decode Python literals directly into the AST buffers

35881e9

Read formatter predicates through typed AST getters

e1a95a9

Read lint rule inputs through typed AST getters

bc08988

[ty] Read inference and IDE inputs through typed AST getters

b4c9701

Keep borrowed AST nodes in Ruff’s semantic model

ed2b737

Pass borrowed AST fields to lint rules

f89adba

Reuse decoded call fields in comprehension rules

820c7f2

Read method parameters after matching __new__

53a2a6b

Keep keyword and positional searches together in call argument lookup

140e51a

Reserve room for small expressions in the AST builder

cea9a0f

Iterate parameters from the stored list of node IDs

0777720

Navigation Menu

Repository navigation

Use a flat Python AST in Ruff and ty - #29817

Use a flat Python AST in Ruff and ty#29817
MichaReiser wants to merge 47 commits into
micha/boxed-mapping-pattern-itemsfrom
micha/flat-python-ast

MichaReiser commented Oct 29, 2026 •

edited

Loading

Uh oh!

astral-automations-bot Bot commented Oct 29, 2026 •

edited

Loading

Uh oh!

astral-automations-bot Bot commented Oct 29, 2026 •

edited

Loading

flake8

trio

sphinx

prefect

Uh oh!

astral-automations-bot Bot commented Oct 29, 2026 •

edited

Loading

Uh oh!

astral-automations-bot Bot commented Oct 29, 2026 •

edited

Loading

Uh oh!

astral-automations-bot Bot Oct 29, 2026

Uh oh!

codspeed Bot commented Oct 29, 2026 •

edited

Loading

Uh oh!

dcreager commented Oct 29, 2026

Uh oh!

This branch was previously deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

		let rewritten = owned_pattern_to_expr(pattern.view(storage).to_owned_node());
		let id = ast::flat::Lower::lower(&rewritten, storage);

Conversation

MichaReiser commented Oct 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Review order

Performance

Ruff lint

Complete Ruff and ty commands

Hosted CodSpeed

Test coverage

Uh oh!

astral-automations-bot Bot commented Oct 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Typing conformance results

No changes detected ✅

Uh oh!

astral-automations-bot Bot commented Oct 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Memory usage report

Summary

Significant changes

flake8

trio

sphinx

prefect

Uh oh!

astral-automations-bot Bot commented Oct 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

ecosystem-analyzer results

Uh oh!

astral-automations-bot Bot commented Oct 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

ruff-ecosystem results

Linter (stable)

Linter (preview)

Formatter (stable)

Formatter (preview)

Uh oh!

astral-automations-bot Bot Oct 29, 2026

Choose a reason for hiding this comment

Uh oh!

codspeed Bot commented Oct 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Merging this PR will regress 5 benchmarks

Performance Changes

Footnotes

Uh oh!

dcreager commented Oct 29, 2026

Uh oh!

This branch was previously deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

MichaReiser commented Oct 29, 2026 •

edited

Loading

astral-automations-bot Bot commented Oct 29, 2026 •

edited

Loading

astral-automations-bot Bot commented Oct 29, 2026 •

edited

Loading

astral-automations-bot Bot commented Oct 29, 2026 •

edited

Loading

`ecosystem-analyzer` results

astral-automations-bot Bot commented Oct 29, 2026 •

edited

Loading

`ruff-ecosystem` results

codspeed Bot commented Oct 29, 2026 •

edited

Loading