Files
marqdo-cli/README.md
T
cflmy b020904199 test(p4): prove row-store/columnar table equivalence
- Add semantic fixture 72: a .mq.md columnar List/Map table store, converted,
  sorted and filtered; compare headers/cells with the T0 row Table via a
  cell-wise helper (Table == is not structural in this language version).
- Teach the direct-runtime semantic harness to resolve entry std: imports.
- Update ADR-0032, P4 status, docs/changelog/evidence; semantic goldens 72.

Gate: fmt, clippy -D warnings, 430 tests passed.
2026-11-20 23:14:38 +08:00

300 lines
15 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
# Marqdo CLI
Marqdo is a **Markdown-native, independently interpreted programming language**:
standard CommonMark/GFM structures carry program control (headings scope,
`**strong**` executes, `*emph*` returns, ordered lists branch, bullet lists
loop, block quotes call), prose and executable nodes mix in one document, and
all side effects go through a capability-checked host boundary. The official
implementation is a Rust workspace; Python/Node are optional foreign kernels,
never the compiler target.
This repository is the new, independent Marqdo toolchain (M0.5). It is **not**
a compatibility rewrite of `cflmy/marqdo`; the old syntax, package format and
plugin ABI v2 are not claimed to work here.
> **Status: stabilization candidate, not released.** The language, lock v3,
> Analysis v2, Trace v2, Tool v1, Plugin ABI v1 (native + worker) and pinned
> Foreign providers are implemented and covered by tests on Linux. The
> macOS/Windows `E0433` platform-cfg failure and the Linux strict
> probe→launch race are fixed in source at `8af7262`, but the native
> Windows/macOS and GitHub Linux lanes have not been re-run from this
> checkout. **`strict` OS isolation is available on Linux only when the
> runtime Bubblewrap probe passes**; it was verified against the full
> escape/resource matrix in a disposable Ubuntu 26.04 VM at `8af7262`
> (ADR-0021). macOS/Windows fail closed, strict L2 (Worker/Foreign) is not
> implemented, and the broker/worker are not a sandbox. The release decision
> is **DO NOT RELEASE** until the remote gates have same-SHA URLs. See
> [`docs/STATUS.md`](docs/STATUS.md), [`COMPATIBILITY.md`](COMPATIBILITY.md)
> and [`docs/security-gate.md`](docs/security-gate.md). The single current CI
> convergence plan is
> [`docs/NEXT_RELEASE_CI_CONVERGENCE.md`](docs/NEXT_RELEASE_CI_CONVERGENCE.md).
## Build
Requires Rust 1.82+.
```bash
cargo build -p marqdo-cli
# binary: target/debug/marqdo (or $CARGO_TARGET_DIR/debug/marqdo)
```
## Quickstart
```bash
marqdo init demo
cd demo
marqdo check index.mq.md
marqdo run index.mq.md # executes ## main
marqdo inspect index.mq.md --ast
marqdo inspect index.mq.md --signature
marqdo call index.mq.md main
```
Example project from scratch (`index.mq.md`):
````markdown
# 数学
## 阶乘
1. `n <= 1`
*1*
2. `n > 1`
**结果 = n * 阶乘(n=n-1)**
*结果*
````
```bash
marqdo check examples/factorial.mq.md
marqdo call examples/factorial.mq.md 数学.阶乘 --arg n=5 # 120
```
The parser is frozen: `marqdo.md.v1` = CommonMark 0.31.2 + GFM tables,
strikethrough, task lists + `$...$`/`$$...$$` math (comrak `=1.0.0`). A `.mq.md`
file still renders normally in any compatible Markdown reader; rendering never
executes anything.
## What works (implemented and tested)
* **P0 contracts** — frozen Markdown profile, AST → program mapping, strict
expression grammar, evaluation semantics; 160 AST goldens, 72 semantic
goldens, 52 compile-error fixtures and 21 runtime-error fixtures
(`tests/`).
* **P1 interpreter** — native Rust tree-walk interpreter: values, scopes,
functions, static parameter inference, assignments, branches, loops,
returns, recursion, structured errors, fuel/stack/output budgets.
* **Addressable memory (RFC-CM P1)** — opaque run-scoped `Memory` regions with
64 KiB zero-filled pages, grow-only `memory.new/size/grow`, bounded byte
read/write, `m[i] = byte`, run/per-region budgets, host-boundary and wire
rejection (ADR-0022).
* **Minimal structure primitives (RFC-CM CM-A04)** — value-semantics
`list.append/concat`, `map.keys/has/get/set/remove`, Unicode scalar
`text.codepoints/from_codepoint`, and non-negative `table.columns/row/cell`;
pure builtins with no mutation or new syntax (ADR-0023).
* **Recoverable results (RFC-CM CM-A05)** — `{ok:true,value}` /
`{ok:false,code,message}` as an ordinary Map, with a `.mq.md` first-fit
free-list allocator demonstrating reuse, splitting and `alloc.no_space`
recovery (ADR-0024). No `Result` type or exception syntax.
* **Bundled stdlib + print prototype (RFC-CM CM-A07/A10)** — offline
`stdlib/` with explicit `std:io`/`std:format`/`std:text` imports, SHA-256
digest verification, Analysis identity and a frozen-CLI closure test; the
`.mq.md` print chain is the only execution path for `打印`/`print` on an official install (ADR-0031),
covering all 15 value types; `HostOp::Print` and the legacy formatter are
removed (ADR-0027/0028/0029/0030/0031).
* **HTTP/1.1 framing in `.mq.md`** — `stdlib/net/http.mq.md` status/header/
chunked parsing and request construction; a loopback TCP test runs a real
exchange (ADR-0034; no TLS/host HTTP claim).
* **Blocking TCP endpoint (RFC-CM P5 first slice)** — `resource.open("tcp",
"<ip>:<port>", ["read","write"])` with exact IP grants, project mode and
5-second timeouts; DNS/TLS/UDP/server sockets excluded (ADR-0034).
* **In-process ResourceProvider v1 (RFC-CM CM-A09 first slice)** — trusted
providers register extra resource kinds, project-mode + name-grant gated,
using the existing `resource.*`/`Result` ABI; no parser/runtime/`HostOp`
change (ADR-0033).
* **Table stdlib algorithms (RFC-CM CM-A08)** — `stdlib/data/table.mq.md`
select/filter/sort/append/set/join/group over the logical `Table` value;
value-returning, input-immutable, with GFM AST laziness unchanged (ADR-0032).
* **Resource streams (RFC-CM CM-A06)** — opaque host-issued `Resource` handles;
`resource.open/read/write/close` transfer bytes through `Memory`; stdout
write in pure/project, granted project-file read, and buffered project-file
write flushed on `close` in project; strict refuses all resource opens until
verified (ADR-0025/0026).
* **P2 references + CLI** — explicit YAML front-matter imports, `ModuleId`/
`UnitId`/anchors, import cycle detection, multi-module calls,
`run/call/check/inspect/fmt`, reference validation (`check --refs`) and the
static link graph (`inspect --refs`).
* **P3 project** — lock v3 full transitive package-instance graph for path and
pinned Git sources, verified content-addressed Git cache, deterministic
`lock`, strict `sync --locked` with atomic `.marqdo/env` materialization.
`run`/`call`/`check` never fetch or rewrite the lock.
* **P4 security core** — capability-checked host operations, project-relative
FS globs with traversal rejection, budgets, effect trace, `pure`/`project`
modes, and **Linux `strict` mode** using Bubblewrap namespaces, no-new-privs,
capability dropping, rlimits, a wall-clock supervisor and child
self-verification (ADR-0021). It is probe-gated and verified in a disposable
Ubuntu VM; unsupported hosts fail closed.
* **P5 Math/Foreign** — native `Formula` value from `$$...$$`; `math.parse/
eval/substitute` over a documented exact-rational subset; one-shot Python/
Node kernels via `foreign.run/call` with typed wire protocol, timeout and
output limits; pinned provider/environment digests and ADR-0020 environment
binding.
* **P6 plugin ABI** — frozen C ABI v1 header (`include/marqdo_plugin_abi_v1.h`),
host bindings, Rust SDK (`mq_plugin_entry!`), `plugin.yaml` validation,
target-triple + sha256 artifact verification, native and isolated worker
transports, brokered `fs.read`, and explicitly trusted in-process calls
(`--plugin --trust-native-plugin --allow-plugin`). A C ABI is not a sandbox.
* **Public protocols** — Analysis v2 and Trace v2 with JSON Schemas and
independent Python consumers; Tool Protocol v1 with arbitrary command names,
SemVer negotiation and whole-bundle digest verification.
Run everything:
```bash
cargo test --workspace
cargo test -p marqdo-cli # conformance corpus + CLI contract tests
cargo fmt --check
cargo clippy --workspace --all-targets -- -D warnings
```
## Deliberate limits (do not assume)
* **Strict is Linux-only and L1-only.** `--mode strict` enables the
Bubblewrap supervisor only when its runtime probe passes; otherwise it exits
1. It supports pure Marqdo + Math and rejects Foreign, plugins, kernels,
external grants and trace files (exit 2). Worker plugins provide
crash/address-space isolation, not an OS sandbox, and are not allowed in
strict.
* **Native plugins run in-process** and require an explicit unsafe trust flag.
A plugin can crash or read the host; a C ABI is not a sandbox.
* **Package management is path + pinned-Git only.** There is no registry,
floating-branch resolution, conflict solver, `--offline`, `tree`, `update`,
`vendor`, or automatic uv/npm environment creation. Manifest
`capabilities` are requests, never grants.
* **Reference validation is opt-in**: `check --refs` reports broken links;
plain `check` ignores them because links are static knowledge references.
Link-form calls cannot cross modules; cross-module calls use imported alias
expressions such as `geometry.圆面积(r=2)`.
* **`ref("#anchor")` is module-local** and errors with `ref.ambiguous` when a
heading range contains more than one formula/code block instead of picking
the first.
* **`fmt` is minimal** (line endings/trailing newlines), not AST-preserving
reformatting.
* **Math is a limited calculator**, not a CAS/prover: no `simplify`/
`differentiate`/matrices/series; equation solving is affine-only.
* **Foreign kernels are unsandboxed one-shot child processes** (no sessions,
no state bridge); they run with the host user's filesystem/network access.
* Finite floats are IEEE-754 binary64; the typed wire round-trip is exact
(`serde_json/float_roundtrip`), tagged integers never lose digits, and
display formatting is decimal rather than bit-identical.
## Repository layout
```text
crates/marqdo-span shared byte spans/source maps (leaf)
crates/marqdo-value Value model, diagnostics, typed wire encoding
crates/marqdo-expression strict expression/statement parser
crates/marqdo-markdown frozen comrak profile, canonical AST, front matter
crates/marqdo-ir module/unit/function/statement IR, builtin table
crates/marqdo-frontend heading scopes, mapping, inference, static calls
crates/marqdo-math exact-rational Formula subset
crates/marqdo-foreign Python/Node one-shot kernel supervisor
crates/marqdo-sandbox mode parsing + platform isolation detection
crates/marqdo-plugin-abi C ABI v1 mirrors, manifest, artifacts, loader
crates/marqdo-plugin-sdk Rust plugin SDK and entry macro
crates/marqdo-host CapabilityHost, budgets, FS, math/foreign/plugin adapters
crates/marqdo-runtime tree-walk interpreter
crates/marqdo-package marqdo.yaml / marqdo.lock / root discovery
crates/cli CLI composition root (binary `marqdo`)
include/ frozen plugin C ABI header
spec/ normative language/format/ABI specs
docs/ blueprint, ADRs, status, contributor guide
examples/ runnable documents + multi-module + plugin demo
tests/ conformance/negative corpus (see tests/README.md)
```
## Documentation index
* [`AGENTS.md`](AGENTS.md) — operating rules for contributors and coding agents.
* [`docs/DEVELOPMENT_BLUEPRINT.md`](docs/DEVELOPMENT_BLUEPRINT.md) — architecture
baseline (Chinese; the M0.6 plan below supersedes it on command-surface scope,
extension registration, plugin addressing, package identity, repository
decoupling and release gates).
* [`docs/STABLE_RELEASE_IMPLEMENTATION.md`](docs/STABLE_RELEASE_IMPLEMENTATION.md)
— **the single authoritative next-phase plan**: PR order, G-A/G-B/G-C gates,
Windows/macOS fixes, protocol/environment stabilization, strict isolation
requirements and the release checklist.
* [`COMPATIBILITY.md`](COMPATIBILITY.md) — protocol versions, per-platform
feature evidence and explicit unsupported features.
* [`CHANGELOG.md`](CHANGELOG.md) — unreleased stabilization changes and the
`0.1.0` baseline.
* [`docs/security-gate.md`](docs/security-gate.md) — G-C evidence and limits;
[`docs/security-vm.md`](docs/security-vm.md) reproduces the disposable VM and
the self-hosted security runner.
* [`docs/releasing.md`](docs/releasing.md) — artifact, checksum, clean-install
smoke, tagging and rollback procedure.
* [`docs/CORE_STABILIZATION_AND_ECOSYSTEM_GATE.md`](docs/CORE_STABILIZATION_AND_ECOSYSTEM_GATE.md)
— historical stabilization spec (baseline `0adebe4`), superseded on current
priorities by the stable-release plan.
* [`docs/NEXT_PHASE_IMPLEMENTATION.md`](docs/NEXT_PHASE_IMPLEMENTATION.md) —
historical M0.6 plan.
* [`docs/EXTENSIBILITY_AUDIT.md`](docs/EXTENSIBILITY_AUDIT.md) — source-backed
audit issues E01–E16.
* [`docs/contributor-guide.md`](docs/contributor-guide.md) — repo map, workflow,
golden updates, quality gate.
* [`docs/STATUS.md`](docs/STATUS.md) — current work-package status, local
evidence and open blockers.
* [`docs/adr/`](docs/adr/) — accepted decisions (crate graph, inert executables,
`ref` ambiguity, strict refusal, AST normalization, CLI contract, BigInt/wire,
package identity, callable resolution, plugin export routing, capability
broker, kernel providers, generic CLI tools, plugin-handle authority, public Analysis/Trace v1).
* Specs: [`markdown-profile.md`](spec/markdown-profile.md),
[`mapping.md`](spec/mapping.md), [`expression.ebnf`](spec/expression.ebnf),
[`evaluation.md`](spec/evaluation.md), [`references.md`](spec/references.md),
[`value-wire.md`](spec/value-wire.md), [`security.md`](spec/security.md),
[`package-format.md`](spec/package-format.md),
[`plugin-abi.md`](spec/plugin-abi.md), [`math.md`](spec/math.md),
[`foreign.md`](spec/foreign.md), [`tool-v1.md`](spec/tool-v1.md),
[`analysis-v2.md`](spec/analysis-v2.md), [`trace-v2.md`](spec/trace-v2.md);
JSON Schemas in [`spec/schemas/`](spec/schemas/).
* [`tests/README.md`](tests/README.md) — conformance corpus and `UPDATE_GOLDENS`
workflow.
## Examples
```bash
marqdo run examples/hello.mq.md
marqdo call examples/factorial.mq.md 数学.阶乘 --arg n=5
marqdo run examples/multi-module/index.mq.md
marqdo run examples/python-foreign.mq.md --allow-foreign # needs python3 or node
marqdo run examples/packages-multi-version/index.mq.md
marqdo run examples/kernel-mock/demo.mq.md \
--kernel examples/kernel-mock/kernel.json --allow-foreign
marqdo inspect examples/references.mq.md --refs
marqdo inspect examples/hello.mq.md --platform # strict: unavailable
./examples/plugin-demo/build.sh
marqdo plugin inspect examples/plugin-demo/build
marqdo run examples/plugin-demo/plugin-call.mq.md \
--plugin examples/plugin-demo/build --trust-native-plugin --allow-plugin demo.solver
./examples/plugin-multi-export/build.sh
marqdo run examples/plugin-multi-export/plugin-call.mq.md \
--plugin examples/plugin-multi-export/build --trust-native-plugin --allow-plugin demo.solver
marqdo tool install examples/tool-unknown-name/tool.json
marqdo benchmark run .
marqdo tool remove benchmark
```
## References
* [CommonMark 0.31.2](https://spec.commonmark.org/0.31.2/) ·
[GFM](https://github.github.io/gfm/) · [Comrak](https://comrak.ee/)
* [Cargo](https://doc.rust-lang.org/cargo/) · [uv](https://docs.astral.sh/uv/)
* [Wasmtime Component Model](https://component-model.bytecodealliance.org/)
* Legacy Marqdo source (reference only): <https://github.com/cflmy/marqdo>