gpui-cache-aware

Name: gpui-cache-aware
Author: nowledge-co

496stars37forks

How to install gpui-cache-aware?

nowledge-co/con-terminal/gpui-cache-aware

$npx -y skills add nowledge-co/con-terminal --skill gpui-cache-aware

Installs into the current project.

Summary

Use this skill when reviewing or changing Con UI performance, especially when a view feels janky, expensive, or unexpectedly re-renders during unrelated updates.

Files · 1

View on GitHub

SKILL.md

1	# GPUI Cache-Aware Review
2
3	Use this skill when reviewing or changing Con UI performance, especially when a view feels janky, expensive, or unexpectedly re-renders during unrelated updates.
4
5	## Goal
6
7	Keep Con visually rich without paying repeated GPUI layout/render cost for work that could be cached or isolated.
8
9	## Core Rule
10
11	Cache at the cheapest correct layer first.
12
13	Order of preference:
14
15	1. Cache parsed / normalized data.
16	2. Cache expensive text-run or highlight transforms.
17	3. Isolate expensive subtrees behind stable entity boundaries.
18	4. Use GPUI `AnyView::cached(...)` only when the subtree has a stable size contract.
19
20	Do not jump straight to view caching.
21
22	## Review Checklist
23
24	For any slow UI path, inspect these in order:
25
26	### 1. Parse / transform churn
27
28	- Is the code reparsing markdown / JSON / syntax / layout input on every render?
29	- Is the code rebuilding `SharedString`, `Vec<TextRun>`, highlighted runs, or table cell text on every render?
30	- Can the expensive transform be retained on the model object and invalidated only on real content change?
31
32	Preferred fix:
33
34	- add data-layer caches on parsed document/block/cell/message structs
35	- invalidate only when source text or theme-dependent key changes
36
37	### 2. Render-tree cardinality
38
39	- Is the renderer producing many tiny `div()` children for content that could be one `StyledText`?
40	- Are inline chips, per-token wrappers, nested flex rows, or deep container stacks used for long-form content?
41
42	Preferred fix:
43
44	- collapse long-form prose to text-first rendering
45	- keep decorative element composition for UI chrome and short content only
46
47	### 3. Entity boundaries
48
49	- Is a very large subtree being rebuilt because the parent panel rerendered for unrelated state?
50	- Can the subtree live in its own `Entity<V>` with a narrower invalidation surface?
51
52	Preferred fix:
53
54	- split large stable regions into their own render entities
55	- keep mutation paths explicit so only that entity gets `notify()`
56
57	### 4. GPUI cached view suitability
58
59	Use `AnyView::cached(...)` only if all of these are true:
60
61	- the subtree size is externally constrained or stable
62	- the cached style can describe the layout contract correctly
63	- reusing previous layout/paint is actually valid for the subtree
64
65	Good fits:
66
67	- panes that fill known bounds
68	- fixed-size or externally-sized tool panels
69	- stable canvases / editors / native-host surfaces
70
71	Bad fits:
72
73	- intrinsic-height rich text documents
74	- content whose height depends on wrapping and dynamic width unless that width/height contract is explicitly handled
75
76	If a cached view causes overlap, clipping, or stale layout, the cache boundary is wrong.
77
78	### 5. Lists and scrolling surfaces
79
80	- If there are many repeated items, consider virtualization before micro-optimizing item chrome.
81	- If the item count is small but each item is expensive, focus on item-level caches instead.
82
83	## Con-Specific Guidance
84
85	### Markdown and chat surfaces
86
87	- Prefer parsed markdown caches on the message/document model.
88	- Cache inline text-run generation for paragraphs, headings, and table cells.
89	- Cache syntax-highlight runs for code blocks.
90	- Avoid per-token flex trees for long replies.
91	- For long assistant replies, isolate each markdown block behind a stable `Entity`
92	and reuse that entity across parent renders. Keep per-block UI state such as
93	table scroll handles on that entity, not inside parsed markdown data that is
94	produced off the UI thread.
95	- Do not hide/fold long replies as the primary performance fix. Folding can be a
96	UX affordance, but the expanded state must remain responsive.
97
98	### Terminal-adjacent UI
99
100	- Keep terminal surfaces and heavy side panels isolated.
101	- Avoid reading terminal runtime state during ordinary render unless already cached.
102	- Be cautious with transparency, animation, and resize interactions; measure before adding visual layers.
103	- For hover affordances that only need cursor feedback, prefer a tiny overlay /
104	hitbox and `cursor_pointer()` over repainting terminal text. Keep the overlay
105	state bounded to the hovered link/range.
106
107	### GPUI interaction gotchas
108
109	- `.hover(\|style\| ...)` is a style hook on `Div`; `.on_hover(...)` is an
110	enter/exit listener and requires a `Stateful<Div>`.
111	- If a view needs `.on_hover(...)`, call `.id(stable_id)` first. Use a stable id
112	such as the pane's `FocusHandle`, not a changing row/index.
113	- For platform-selected files (`ghostty_view.rs`, `windows_view.rs`,
114	`linux_view.rs`), a macOS `cargo check -p con` only checks the macOS path.
115	Validate the touched platform path on its target or wait for portable CI
116	before treating the fix as closed.
117
118	## Validation
119
120	After a cache-related change:
121
122	- verify `cargo check -p con`
123	- run targeted tests if the subsystem has them
124	- confirm there is no layout regression
125	- confirm the cache invalidates on real content/theme changes
126	- confirm the cache does not hide stale data
127	- if touching Windows/Linux-only view code from macOS, note that local checks are
128	partial unless the target toolch

People who liked this also viewed

6 picks

Type

find-skillsHelps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skSkillsJul 20262.6M27k
grill-meA relentless interview to sharpen a plan or design.SkillsJul 2026633k182k
vercel-react-best-practicesVercel's official collection of agent skillsSkillsJul 2026573k29k
improve-codebase-architectureScan a codebase for deepening opportunities, present them as a visual HTML report, then grill through whichever one you pick.SkillsJul 2026522k182k
setup-matt-pocock-skillsConfigure this repo for the engineering skills — set up its issue tracker, triage label vocabulary, and domain doc layout. Run once before first use of the other engineering skills.SkillsJul 2026454k182k
triageMove issues and external PRs through a state machine of triage roles — categorise, verify, grill if needed, and write agent-ready briefs.SkillsJul 2026424k182k

gpui-cache-aware

496stars37forks

View on GitHub

How to install gpui-cache-aware?

nowledge-co/con-terminal/gpui-cache-aware

$npx -y skills add nowledge-co/con-terminal --skill gpui-cache-aware

Installs into the current project.

Summary

Use this skill when reviewing or changing Con UI performance, especially when a view feels janky, expensive, or unexpectedly re-renders during unrelated updates.

Files · 1

View on GitHub

SKILL.md

1	# GPUI Cache-Aware Review
2
3	Use this skill when reviewing or changing Con UI performance, especially when a view feels janky, expensive, or unexpectedly re-renders during unrelated updates.
4
5	## Goal
6
7	Keep Con visually rich without paying repeated GPUI layout/render cost for work that could be cached or isolated.
8
9	## Core Rule
10
11	Cache at the cheapest correct layer first.
12
13	Order of preference:
14
15	1. Cache parsed / normalized data.
16	2. Cache expensive text-run or highlight transforms.
17	3. Isolate expensive subtrees behind stable entity boundaries.
18	4. Use GPUI `AnyView::cached(...)` only when the subtree has a stable size contract.
19
20	Do not jump straight to view caching.
21
22	## Review Checklist
23
24	For any slow UI path, inspect these in order:
25
26	### 1. Parse / transform churn
27
28	- Is the code reparsing markdown / JSON / syntax / layout input on every render?
29	- Is the code rebuilding `SharedString`, `Vec<TextRun>`, highlighted runs, or table cell text on every render?
30	- Can the expensive transform be retained on the model object and invalidated only on real content change?
31
32	Preferred fix:
33
34	- add data-layer caches on parsed document/block/cell/message structs
35	- invalidate only when source text or theme-dependent key changes
36
37	### 2. Render-tree cardinality
38
39	- Is the renderer producing many tiny `div()` children for content that could be one `StyledText`?
40	- Are inline chips, per-token wrappers, nested flex rows, or deep container stacks used for long-form content?
41
42	Preferred fix:
43
44	- collapse long-form prose to text-first rendering
45	- keep decorative element composition for UI chrome and short content only
46
47	### 3. Entity boundaries
48
49	- Is a very large subtree being rebuilt because the parent panel rerendered for unrelated state?
50	- Can the subtree live in its own `Entity<V>` with a narrower invalidation surface?
51
52	Preferred fix:
53
54	- split large stable regions into their own render entities
55	- keep mutation paths explicit so only that entity gets `notify()`
56
57	### 4. GPUI cached view suitability
58
59	Use `AnyView::cached(...)` only if all of these are true:
60
61	- the subtree size is externally constrained or stable
62	- the cached style can describe the layout contract correctly
63	- reusing previous layout/paint is actually valid for the subtree
64
65	Good fits:
66
67	- panes that fill known bounds
68	- fixed-size or externally-sized tool panels
69	- stable canvases / editors / native-host surfaces
70
71	Bad fits:
72
73	- intrinsic-height rich text documents
74	- content whose height depends on wrapping and dynamic width unless that width/height contract is explicitly handled
75
76	If a cached view causes overlap, clipping, or stale layout, the cache boundary is wrong.
77
78	### 5. Lists and scrolling surfaces
79
80	- If there are many repeated items, consider virtualization before micro-optimizing item chrome.
81	- If the item count is small but each item is expensive, focus on item-level caches instead.
82
83	## Con-Specific Guidance
84
85	### Markdown and chat surfaces
86
87	- Prefer parsed markdown caches on the message/document model.
88	- Cache inline text-run generation for paragraphs, headings, and table cells.
89	- Cache syntax-highlight runs for code blocks.
90	- Avoid per-token flex trees for long replies.
91	- For long assistant replies, isolate each markdown block behind a stable `Entity`
92	and reuse that entity across parent renders. Keep per-block UI state such as
93	table scroll handles on that entity, not inside parsed markdown data that is
94	produced off the UI thread.
95	- Do not hide/fold long replies as the primary performance fix. Folding can be a
96	UX affordance, but the expanded state must remain responsive.
97
98	### Terminal-adjacent UI
99
100	- Keep terminal surfaces and heavy side panels isolated.
101	- Avoid reading terminal runtime state during ordinary render unless already cached.
102	- Be cautious with transparency, animation, and resize interactions; measure before adding visual layers.
103	- For hover affordances that only need cursor feedback, prefer a tiny overlay /
104	hitbox and `cursor_pointer()` over repainting terminal text. Keep the overlay
105	state bounded to the hovered link/range.
106
107	### GPUI interaction gotchas
108
109	- `.hover(\|style\| ...)` is a style hook on `Div`; `.on_hover(...)` is an
110	enter/exit listener and requires a `Stateful<Div>`.
111	- If a view needs `.on_hover(...)`, call `.id(stable_id)` first. Use a stable id
112	such as the pane's `FocusHandle`, not a changing row/index.
113	- For platform-selected files (`ghostty_view.rs`, `windows_view.rs`,
114	`linux_view.rs`), a macOS `cargo check -p con` only checks the macOS path.
115	Validate the touched platform path on its target or wait for portable CI
116	before treating the fix as closed.
117
118	## Validation
119
120	After a cache-related change:
121
122	- verify `cargo check -p con`
123	- run targeted tests if the subsystem has them
124	- confirm there is no layout regression
125	- confirm the cache invalidates on real content/theme changes
126	- confirm the cache does not hide stale data
127	- if touching Windows/Linux-only view code from macOS, note that local checks are
128	partial unless the target toolch

People who liked this also viewed

6 picks

Type