PLC代码审查

Building from source

npm install

That's the whole setup. The parser ships as WebAssembly: web-tree-sitter boots the tree-sitter runtime and loads the grammar's tree-sitter-iec61131_3_st.wasm (published inside the tree-sitter-iec61131-3-st package). There's no native addon to compile, so no C++ toolchain, no node-gyp, and no platform-specific prerequisites — npm install runs the same on Windows, Linux, and macOS, on any supported Node version.

Quick start

PLCopen Coding Guidelines preset

The repo ships one opinionated preset: presets/plcopen.yml, tuned to PLCopen Coding Guidelines v1.0. Vendor-neutral, standards-body-published — use it as your baseline if your team doesn't already have a style guide:

```yaml

Configuration

Create .plc-st-review.yml at the repo root:

case_sensitive: false          # identifier casing rules; see "Case sensitivity" below

disabled_checks:
  - COMMENT_ONLY

severity_overrides:
  TIMER_VALUE_CHANGED: error   # all timer changes block merge

ignore_paths:
  - "deprecated/**"
  - "third_party/**"

safety_critical_prefixes:
  - SAFETY_
  - INTERLOCK_
  - SIL_

reporting:
  fail_on_severity: error      # exit-nonzero threshold
  comment_style: inline        # inline | summary | both (GitLab/GitHub)

metrics:                       # bands for the metric-regression checks
  thresholds:
    cyclomatic_complexity:
      warn: 15
      error: 25
    nesting_depth:
      warn: 5
      error: 8

parsing:
  max_file_size_bytes: 1000000  # per-file size cap; see "Source-size cap" below

limits:                        # numeric caps for the size-/count-checking PLCopen rules
  max_identifier_length: 32    # PLCopen N6 — drives IDENTIFIER_TOO_LONG (null = off)
  max_globals_used_per_pou: 10 # PLCopen CP18 — drives TOO_MANY_GLOBALS_USED (null = off)
  max_parameters: 8            # PLCopen CP23 — drives TOO_MANY_PARAMETERS (null = off)

identifier_charset: null       # PLCopen N8 — regex every identifier must match; null = off
                               # e.g. '^[A-Za-z_][A-Za-z0-9_]*$' restricts to ASCII + underscore

The metrics block is optional; the values shown are the defaults. COMPLEXITY_INCREASED and NESTING_INCREASED read these bands; LOC_SPIKE fires on any single-PR growth over 50%. See docs/checks-reference.md for the full block (the lines_of_code, comment_ratio, and fan_out keys are accepted now and consumed by the upcoming standalone --metrics mode).

Layer your own preferences on top.

naming_conventions: bool: { prefix: b } # we use b, not PLCopen's x ```

The full rule-by-rule mapping (which PLCopen rules the preset enforces, which would need new checks, which are out of scope) is on the Pages site under Presets → PLCopen Coding Guidelines.

.github/workflows/lint.yml, same idea on GitHub

name: lint on: [push] jobs: lint: runs-on: ubuntu-latest steps: - uses: actions/checkout@v4 - run: npx plc-st-review --lint "src/**/*.st" ```

--lint accepts file paths, directories, or globs (*, **, mixed). It parses each .st file in isolation and runs the 59 single-revision checks: the 21 diff-based ones are auto-disabled because there's no "before" state. Exit code is non-zero when any finding meets reporting.fail_on_severity (default error), so the job fails the pipeline on real bugs.

Diff-based (compare before vs after)