Skip to content

docs: t27o front-end census of the roadmap repos -- Zig first, then deeper Rust, Python, TypeScript, shell (Closes #10599) #10601

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
gHashTag merged 1 commit into from
Oct 18, 2026
+42 −0
Merged

docs: t27o front-end census of the roadmap repos -- Zig first, then deeper Rust, Python, TypeScript, shell (Closes #10599) #10601

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
42 changes: 42 additions & 0 deletions docs/nona-03-manifest/T27O-COMPETITORS-AND-DOGFOOD.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -55,6 +55,11 @@ First refused construct per item, 17,015 refusals in total, by kind:
The single largest reasons: type `Vec` 1,419, macro `vec!` 1,363, type `String` 680, type `Err` 612,
tuple 550, type `Ok` 473, `self` of a refused type 412, type `Path` 402, for over `&...` 380.

After the five classes of #10526 (PR #10593, measured on its branch against master 0c98c9641 with the
same files and caps): 1,667 fns converted (was 1,573) and 593 tests (was 504); first refusals type `Ok`
474 -> 2, `Err` 609 -> 0, for over `&...` 380 -> 4, `]` 306 -> 59, raw strings 154 -> 0, line
continuations 88 -> 0.

What the numbers say [I]: owned heap values (`Vec`, `vec!`, `String`, `Path`) are about a quarter of
all first refusals, and the cascade multiplies every root refusal. Five classes need no new type at
all (#10526: `Ok`/`Err` in test asserts, `for x in &[..]`, `&[]` arguments, raw strings, line
Expand Down Expand Up @@ -145,3 +150,40 @@ tests execute, but no test pins them one by one.
of refusals.
2. Vector quality: coverage-guided samples and a mutation score in the golden criteria (5 and 6).
3. A Charon/MIR spike for the Rust front end (1).

## 7. Beyond Rust: what the roadmap is written in [M; coarse where marked]

Measured 2026-10-19 at the commits of section 1, tracked files only, excluding `node_modules`, vendored,
`gen/`, `dist/`, `build/`, `target/`. "Distinct" deduplicates function bodies across the five
repositories (trinity and trinity-fpga share most files). A first subset is ours, stated per language
below; "pass" means the structure passes it, not that a converter ran. The measurement scripts were
throw-away (Python's `ast`, the TypeScript 5.9.3 compiler API from a local cache, regex for Zig and
shell); they are not committed, by the Only-t27 rule. #10543's sweep is the committed instrument for Rust.

| Language | Files | Distinct fns | In a first subset | Top first blockers |
|---|---|---|---|---|
| Rust | 1,388 | 16,775 | 1,573 converted (9.4 %), 1,667 after #10593 | types (heap values), methods, macros |
| Zig (coarse) | 2,644 | 36,495 | 3,797 (10.4 %) scalar and slice only; 7,767 (21.3 %) with same-file named types; 13,900 (38.1 %) with `*T` self | error union 41.8 %, pointer 29.9 %, named type 16.8 %, std or allocator 6.9 % |
| TypeScript | 3,654 | 9,746 | 243 (2.5 %); 123 (1.3 %) also fully annotated | async / await 26.4 %, object literal, import, destructured parameter |
| Python | 1,641 | 6,468 | 237 (3.7 %); 58 (0.9 %) also annotated inside the subset | import use, tuple, method outside the table, the cascade, `self` state |
| Shell (coarse) | 812 | - | pure computation is 1.4-2.1 % of scripts | 74-82 % are orchestration (git, gh, curl, docker, cargo) |

Subsets: Rust is `rust_convert.t27`'s. Zig: primitive scalars and slices in signature and body. Python:
no async, yield, lambda, nested def, decorators other than `@staticmethod`, `*args`, `with`, `try`,
`global`; calls only to pure builtins, `math`, a table of str and bytes methods and same-module
functions; no tuple, dict or set, I/O or import use. TypeScript: the same idea on its syntax tree
(no type checker). Zig's 1,947 of 2,644 files carry `test` blocks (19,594 `expectEqual`), which can be
the oracle a Zig front end replays.

Roadmap stages: stage 1 (#4543, the swarm's Python and shell tools) holds 2,395 distinct Python fns, 110
passing (4.6 %); stage 3 (#4545, app.t27.ai backends) 4,027 TypeScript fns, 92 passing; stage 5 (#4547,
the Zig core) 24,062 Zig fns in trinity `src` and 5,828 in trios crates; stage 6 (#4548) is the Rust of
section 1. Stages 2 and 4 live in repositories that were not checked out (unknown); stage 7's 4,899
Verilog files were not measured.

The ranking this gives [I]: (1) a Zig front end (#10597): the largest pool, the closest grammar, its own
tests as the oracle; (2) deeper Rust (`Vec` without a heap #10546, tuples, traits): no new parser;
(3) Python (#8995, #9006): small pool, mostly numeric format models; (4) TypeScript: a quarter of the
fns are async, so progress there means pulling pure decisions out of I/O code, not a subset front end;
(5) shell: orchestration, almost nothing to convert.

Loading