返回 CodeWhale
FILE_DECOMPOSITION_0_9_0.md
根目录 / docs / rfcs / FILE_DECOMPOSITION_0_9_0.md
1 # RFC: File Decomposition for v0.9.0
2
3 **Status (2026-07-12): still valuable, needs redesign.** The original
4 `config.rs` plan was overtaken by events: `config.rs` shrank to ~7.0k lines via
5 a different module split (`crates/tui/src/config/`), and the provider-churn
6 problem it targeted was mostly solved by the Models.dev catalog + ProviderLake
7 (#4184–#4188) rather than by file moves. Meanwhile the render/CLI monoliths
8 grew — `ui.rs` 9.4k → 13.6k, `main.rs` 8.0k → 12.1k — so any revival of this
9 RFC should re-scope around `ui.rs` and `main.rs`. The migration method
10 (§"Migration strategy": git mv + re-exports, no functional changes) remains
11 the right approach.
12
13 ## Problem
14
15 Six files exceed 5,000 lines. The worst offenders accumulate provider-specific
16 logic, test code, and UI rendering in single translation units. This makes
17 provider additions touch 15+ files and makes code review fragile.
18
19 ### Current state (lines, v0.8.6x era — see status note for v0.8.68 numbers)
20
21 | File | Lines | Contents |
22 |------|-------|----------|
23 | `crates/tui/src/config.rs` | 10,046 | Provider resolution, env handling, model aliases, capability matrix, 2,000+ lines of tests |
24 | `crates/tui/src/tui/ui.rs` | 9,400 | TUI render loop, input handling, command dispatch, /logout clearing |
25 | `crates/tui/src/tui/ui/tests.rs` | 8,360 | Tests for ui.rs |
26 | `crates/tui/src/main.rs` | 7,998 | CLI arg parsing, mode selection, startup |
27 | `crates/tui/src/tui/app.rs` | 7,256 | Application state struct and lifecycle |
28 | `crates/tui/src/runtime_threads.rs` | 5,454 | Async runtime orchestration |
29
30 ## Proposed decomposition
31
32 ### 1. `config.rs` → provider module tree
33
34 Split `crates/tui/src/config.rs` into:
35
36 ```
37 crates/tui/src/config/
38 ├── mod.rs # Re-exports, Config struct, load/save
39 ├── provider.rs # ApiProvider enum, parse/as_str/display_name/all
40 ├── capability.rs # ProviderCapability, provider_capability()
41 ├── model_resolution.rs # wire_model_for_provider, normalize_model_name_for_provider
42 ├── env.rs # EnvGuard, env var precedence, per-provider env handling
43 ├── constants.rs # All DEFAULT_*_MODEL and DEFAULT_*_BASE_URL constants
44 └── tests/ # Test module
45 ├── mod.rs
46 ├── provider.rs
47 ├── capability.rs
48 ├── model_resolution.rs
49 └── env.rs
50 ```
51
52 **Why:** Every new provider currently requires edits to ~20 match arms scattered
53 across one 10K-line file. With constants in their own module and resolution
54 logic isolated, adding a provider becomes: add constants, add enum variant, add
55 one match arm per function. The drift check script can validate each sub-module
56 independently.
57
58 ### 2. `ui.rs` → view modules
59
60 Split `crates/tui/src/tui/ui.rs` into:
61
62 ```
63 crates/tui/src/tui/
64 ├── ui.rs # Core render loop, frame dispatch (keep under 2,000 lines)
65 ├── input.rs # Keyboard/mouse input handling
66 ├── command_dispatch.rs # /command routing, /logout, /config
67 └── status_bar.rs # Status bar rendering
68 ```
69
70 **Why:** The /logout clearing logic, command dispatch, and render loop are
71 independent concerns. `ui.rs` currently has a 6,200-line function body for
72 `execute_command_input` that mixes input parsing, command routing, and state
73 mutation.
74
75 ### 3. `main.rs` → CLI module
76
77 Split `crates/tui/src/main.rs` into:
78
79 ```
80 crates/tui/src/cli/
81 ├── mod.rs # Cli struct, arg parsing
82 ├── args.rs # Argument definitions
83 └── startup.rs # Mode selection, config loading, resume logic
84 ```
85
86 **Why:** `main.rs` at 8K lines suggests the CLI definition has outgrown a
87 single file. Separating arg definitions from startup logic makes the entry
88 point readable.
89
90 ### 4. Provider additions should be data-driven
91
92 The current provider pattern requires touching:
93 - `config.rs`: 20+ match arms
94 - `cli/src/lib.rs`: 4+ match arms
95 - `agent/src/lib.rs`: static registry
96 - `tui/provider_picker.rs`: picker list
97 - `docs/PROVIDERS.md`: registry table
98 - `config.example.toml`: example section
99 - `README.md`: env vars table
100 - `scripts/check-provider-registry.py`: drift check
101
102 A data-driven approach would define each provider as a struct with its
103 constants, env vars, capability metadata, and display name — then derive the
104 match arms from the data. This is a larger refactor but would reduce provider
105 additions to a single file change.
106
107 ## Priority
108
109 1. **config.rs decomposition** — highest impact, most provider churn happens here
110 2. **ui.rs decomposition** — second highest, /logout and command dispatch are independent
111 3. **Data-driven providers** — aspirational for v0.9.0, requires trait design
112
113 ## Migration strategy
114
115 Each decomposition should be a standalone PR that:
116 1. Creates the new module tree
117 2. Moves code with `git mv` (preserves history)
118 3. Adds `pub use` re-exports in the old file location (zero API change)
119 4. Runs the full test suite
120 5. Removes the re-exports in a follow-up PR once consumers are updated
121
122 No functional changes in decomposition PRs. Keep them boring.
123
123 lines MARKDOWN