Diagnostic filtering

Source: src/engine/filter-diagnostics.ts

Removes diagnostics that match the user's ignore configuration, after every rule has run. Silencing a known diagnostic or excluding a generated file therefore does not require turning the rule off: no-orm-in-services can run everywhere and still stay quiet inside a legacy module.

Input is the diagnostics, the config, and the project root:

diagnostics: Diagnostic[]          // all diagnostics from rule execution
config: NestjsDoctorConfig         // contains ignore.rules and ignore.files
targetPath: string                 // root directory, used to make paths relative for glob matching

The result is a subset of the input:

Diagnostic[]    // filtered array (subset of input)

The config ignores run first. A second pass then reads suppression comments in the source.

ignore.rules contains rule IDs to suppress:

{
  "ignore": {
    "rules": ["architecture/no-orm-in-services"]
  }
}

The match is exact against the diagnostic's rule field. Every diagnostic from that rule is removed.

ignore.files contains glob patterns for files whose diagnostics should be hidden:

{
  "ignore": {
    "files": ["src/generated/**", "src/legacy/**"]
  }
}

File paths are normalized to forward slashes before matching, so patterns work across platforms.

Source: src/engine/inline-suppressions.ts

After config-based filtering, a second pass (filterSuppressedDiagnostics) removes diagnostics silenced by comments in the source itself. Directives use the ignore verb, or disable as an alias, inside a // or /* */ comment:

const config = eval(raw); // nestjs-doctor-ignore security/no-eval
 
// nestjs-doctor-ignore-next-line security/no-eval
const config = eval(raw);
 
// nestjs-doctor-ignore-file security/no-eval

Each directive covers a different scope:

How a directive is parsed:

• The rule list is space- or comma-separated. Omit it to suppress every rule for that scope.
• Only tokens containing / count as rule ids, since every id is category/name. A -- reason trailer is therefore ignored wherever it sits.
• Only real comments are honored. String, template and regex contents are blanked via the AST before parsing, so a directive inside a string cannot suppress anything.

What it reads, and what it can reach:

• TypeScript comes from the in-memory AST project, with no extra disk reads. Prisma .prisma files are not in that project, so they are read on demand.
• Each file is parsed at most once per pass.
• Line-scoped directives key off the diagnostic's line, so they only apply to code diagnostics. Schema diagnostics have no line and can only be suppressed with nestjs-doctor-ignore-file, including one placed anywhere in schema.prisma.

If no ignore config is set, which is the default, the filter returns the original array immediately without iterating. The inline-suppression pass skips any file whose source contains no nestjs-doctor-ignore/-disable token.

• config.rules: disables a rule entirely. It never runs.
• config.ignore.rules: lets the rule run, then hides its output.

The distinction matters for project-scoped rules that cross-reference data. For example, no-missing-injectable checks providers against the module graph.

Disabling it with config.rules means the rule never runs. Ignoring it with config.ignore.rules means the rule runs and its diagnostics are hidden.

• If diagnostics are disappearing unexpectedly, check ignore.rules and ignore.files in your config.
• File patterns use glob syntax via picomatch. Double-star ** matches any depth.