Struct CandidateScanner 

pub struct CandidateScanner { /* private fields */ }

What: The loaded scanner. The field is private: the only way to obtain a value is load. Why: Holding the loaded rules in one value is what makes “load once, scan many” a property of the type: scanning borrows it and cannot reload.

In TS you’d write (pseudocode):

class CandidateScanner { readonly #scanner: Scanner }

Implementations

impl CandidateScanner

What: impl CandidateScanner { ... } attaches the adapter’s operations, like class methods. Why: Loading is the only fallible, expensive step; scanning afterwards is cheap and repeatable.

In TS you’d write (pseudocode):

class CandidateScanner { static load(rules, builtinRules) {} scan(candidate, bytes) {} }

pub fn load( rules: &RulesSource, builtin_rules: bool, ) -> Result<CandidateScanner, ScannerError>

What: Load the runtime rules file and, when asked, the built-in baseline. &RulesSource borrows the selection; Result<T, E> is “a value or a failure”. Why: The scanner’s loader validates rules, reuses its compiled-rule cache and catches its own panics. Any failure leaves no scanner at all, so nothing can be scanned with half a rule set.

In TS you’d write (pseudocode):

static load(rules: RulesSource, builtinRules: boolean): CandidateScanner

pub fn cache_warnings(&self) -> &[CacheWarning]

What: The cache diagnostics of this load, as the scanner’s own fixed-token values. &[CacheWarning] borrows the list read-only. Why: A stale or missing rule cache is recovered by recompiling, which is slow; the caller reports that once per invocation.

In TS you’d write (pseudocode):

get cacheWarnings(): readonly CacheWarning[]

pub fn scan( &self, candidate: &Candidate, bytes: &[u8], ) -> Result<CandidateScan, ScannerError>

What: Scan one candidate’s pathname and exact bytes. &[u8] borrows the bytes; the result is the scanner’s own CandidateScan. Why: The scan’s identity is the candidate’s position in its version, so a finding is attributed without reading the pathname, which may be masked. Fail-closed findings (an engine error, a line break in a pathname) are returned like any other finding and are never filtered here.

In TS you’d write (pseudocode):

scan(candidate: Candidate, bytes: Uint8Array): CandidateScan