fn_registry/functions/infra/audit_uses_functions.go

package infra

import (
	"bufio"
	"database/sql"
	"encoding/json"
	"fmt"
	"os"
	"path/filepath"
	"regexp"
	"strings"
	"unicode"

	_ "github.com/mattn/go-sqlite3"
)

// UsesFunctionsAudit holds the drift report for a single app.
type UsesFunctionsAudit struct {
	AppID   string   // registry id, e.g. "kanban_go_tools"
	Lang    string   // "go" or "py"
	DirPath string   // dir_path as stored in registry.db
	Missing []string // function IDs found in imports but absent from app.md uses_functions
	Unused  []string // function IDs declared in app.md but not detected in code
}

// auditFnMeta holds registry metadata for a single function.
type auditFnMeta struct {
	id     string
	name   string
	domain string
	lang   string
}

// AuditUsesFunctions checks every Go and Python app registered in registry.db
// and compares the uses_functions declared in the app.md manifest against the
// functions actually imported by the app's source code.
//
// For Go apps it greps for "fn-registry/functions/<domain>" import paths, then
// searches the source for the exported symbol derived from each function name
// (snake_case → PascalCase) to achieve per-function granularity within a package.
//
// For Python apps it scans for "from <pkg> import X" patterns where <pkg> matches
// a known registry domain, then resolves X to a function ID by matching the name
// field in registry.db.
//
// Returns an error only if registry.db cannot be opened. Apps where dir_path
// does not exist on disk are reported with Missing/Unused = nil (cannot inspect).
func AuditUsesFunctions(registryRoot string) ([]UsesFunctionsAudit, error) {
	dbPath := filepath.Join(registryRoot, "registry.db")
	dsn := fmt.Sprintf("file:%s?mode=ro&_foreign_keys=on", dbPath)
	db, err := sql.Open("sqlite3", dsn)
	if err != nil {
		return nil, fmt.Errorf("audit_uses_functions: open db: %w", err)
	}
	defer db.Close()
	if err := db.Ping(); err != nil {
		return nil, fmt.Errorf("audit_uses_functions: ping db: %w", err)
	}

	// Load all Go/Python functions from registry: id → name, domain, lang.
	rows, err := db.Query(`SELECT id, name, domain, lang FROM functions WHERE lang IN ('go','py')`)
	if err != nil {
		return nil, fmt.Errorf("audit_uses_functions: query functions: %w", err)
	}
	allFunctions := make(map[string]auditFnMeta) // id → meta
	for rows.Next() {
		var m auditFnMeta
		if err := rows.Scan(&m.id, &m.name, &m.domain, &m.lang); err != nil {
			continue
		}
		allFunctions[m.id] = m
	}
	rows.Close()

	// Load apps with lang go or py.
	type appRow struct {
		id           string
		lang         string
		dirPath      string
		usesFunctions []string
	}
	rows2, err := db.Query(`SELECT id, lang, dir_path, uses_functions FROM apps WHERE lang IN ('go','py')`)
	if err != nil {
		return nil, fmt.Errorf("audit_uses_functions: query apps: %w", err)
	}
	var apps []appRow
	for rows2.Next() {
		var a appRow
		var ufJSON string
		if err := rows2.Scan(&a.id, &a.lang, &a.dirPath, &ufJSON); err != nil {
			continue
		}
		_ = json.Unmarshal([]byte(ufJSON), &a.usesFunctions)
		apps = append(apps, a)
	}
	rows2.Close()

	var results []UsesFunctionsAudit
	for _, app := range apps {
		absDir := app.dirPath
		if !filepath.IsAbs(absDir) {
			absDir = filepath.Join(registryRoot, app.dirPath)
		}
		audit := UsesFunctionsAudit{
			AppID:   app.id,
			Lang:    app.lang,
			DirPath: app.dirPath,
		}

		if _, err := os.Stat(absDir); os.IsNotExist(err) {
			// Cannot inspect — skip diff, leave Missing/Unused nil.
			results = append(results, audit)
			continue
		}

		var importedIDs []string
		switch app.lang {
		case "go":
			importedIDs = auditGoApp(absDir, allFunctions)
		case "py":
			importedIDs = auditPyApp(absDir, allFunctions)
		}

		declaredSet := make(map[string]bool)
		for _, id := range app.usesFunctions {
			declaredSet[id] = true
		}
		importedSet := make(map[string]bool)
		for _, id := range importedIDs {
			importedSet[id] = true
		}

		for id := range importedSet {
			if !declaredSet[id] {
				audit.Missing = append(audit.Missing, id)
			}
		}
		for id := range declaredSet {
			if !importedSet[id] {
				audit.Unused = append(audit.Unused, id)
			}
		}

		results = append(results, audit)
	}
	return results, nil
}

// auditGoApp returns function IDs detected in the Go source files of appDir.
// Strategy:
//  1. Find all "fn-registry/functions/<domain>" import paths.
//  2. For each domain, collect registry functions in that domain.
//  3. Grep source files for the exported symbol (PascalCase of name).
//     If any source file contains the token, the function is considered used.
func auditGoApp(appDir string, all map[string]auditFnMeta) []string {
	// Step 1: collect imported domains.
	importedDomains := collectGoImportedDomains(appDir)
	if len(importedDomains) == 0 {
		return nil
	}

	// Step 2: for each function in those domains, grep for its exported name.
	var used []string
	// Read all .go source once into a single blob for fast search.
	blob := readGoSourceBlob(appDir)
	if blob == "" {
		return nil
	}

	for _, m := range all {
		if m.lang != "go" {
			continue
		}
		if !importedDomains[m.domain] {
			continue
		}
		exported := snakeToPascal(m.name)
		// Use word-boundary-like check: look for the token as a standalone identifier.
		// We check the domain qualifier pattern e.g. "infra.SQLiteOpen" or bare "SQLiteOpen(".
		if containsToken(blob, exported) {
			used = append(used, m.id)
		}
	}
	return used
}

// collectGoImportedDomains returns the set of registry domains imported by .go files.
var goImportRe = regexp.MustCompile(`"fn-registry/functions/([a-z]+)"`)

func collectGoImportedDomains(appDir string) map[string]bool {
	domains := make(map[string]bool)
	_ = filepath.Walk(appDir, func(path string, info os.FileInfo, err error) error {
		if err != nil || info.IsDir() || !strings.HasSuffix(path, ".go") {
			return nil
		}
		f, err := os.Open(path)
		if err != nil {
			return nil
		}
		defer f.Close()
		sc := bufio.NewScanner(f)
		for sc.Scan() {
			line := sc.Text()
			if m := goImportRe.FindStringSubmatch(line); m != nil {
				domains[m[1]] = true
			}
		}
		return nil
	})
	return domains
}

// readGoSourceBlob concatenates all .go file contents in appDir.
func readGoSourceBlob(appDir string) string {
	var sb strings.Builder
	_ = filepath.Walk(appDir, func(path string, info os.FileInfo, err error) error {
		if err != nil || info.IsDir() || !strings.HasSuffix(path, ".go") {
			return nil
		}
		data, err := os.ReadFile(path)
		if err != nil {
			return nil
		}
		sb.Write(data)
		sb.WriteByte('\n')
		return nil
	})
	return sb.String()
}

// containsToken reports whether the exported symbol appears as an identifier
// in src (preceded and followed by non-letter/non-digit/non-underscore runes,
// or at string boundaries). This avoids matching substrings inside longer names.
func containsToken(src, token string) bool {
	idx := 0
	for {
		pos := strings.Index(src[idx:], token)
		if pos < 0 {
			return false
		}
		abs := idx + pos
		before := abs == 0 || !isIdentRune(rune(src[abs-1]))
		after := abs+len(token) >= len(src) || !isIdentRune(rune(src[abs+len(token)]))
		if before && after {
			return true
		}
		idx = abs + 1
	}
}

func isIdentRune(r rune) bool {
	return r == '_' || unicode.IsLetter(r) || unicode.IsDigit(r)
}

// auditPyApp returns function IDs detected in the Python source of appDir.
// Looks for: "from <pkg> import X, Y" patterns and resolves X, Y to function IDs.
var pyFromImportRe = regexp.MustCompile(`from\s+(\w+)\s+import\s+(.+)`)

func auditPyApp(appDir string, all map[string]auditFnMeta) []string {
	// Build name→id map for py functions.
	nameToID := make(map[string]string) // "metabase_auth" → "metabase_auth_py_infra"
	for _, m := range all {
		if m.lang == "py" {
			nameToID[m.name] = m.id
		}
	}

	usedSet := make(map[string]bool)

	_ = filepath.Walk(appDir, func(path string, info os.FileInfo, err error) error {
		if err != nil || info.IsDir() || !strings.HasSuffix(path, ".py") {
			return nil
		}
		f, err := os.Open(path)
		if err != nil {
			return nil
		}
		defer f.Close()
		sc := bufio.NewScanner(f)
		for sc.Scan() {
			line := strings.TrimSpace(sc.Text())
			if m := pyFromImportRe.FindStringSubmatch(line); m != nil {
				// m[2] = "X, Y, Z" or "X"
				names := strings.Split(m[2], ",")
				for _, nm := range names {
					nm = strings.TrimSpace(nm)
					nm = strings.Fields(nm)[0] // strip "as alias"
					if id, ok := nameToID[nm]; ok {
						usedSet[id] = true
					}
				}
			}
		}
		return nil
	})

	var used []string
	for id := range usedSet {
		used = append(used, id)
	}
	return used
}

// snakeToPascal converts snake_case to PascalCase (Go exported name).
// E.g. "sqlite_open" → "SQLiteOpen", "http_json_response" → "HTTPJSONResponse".
// Common abbreviations are uppercased in full.
var commonAbbrevs = map[string]string{
	"http":   "HTTP",
	"https":  "HTTPS",
	"sql":    "SQL",
	"sqlite": "SQLite",
	"url":    "URL",
	"api":    "API",
	"id":     "ID",
	"db":     "DB",
	"tls":    "TLS",
	"tcp":    "TCP",
	"udp":    "UDP",
	"ip":     "IP",
	"json":   "JSON",
	"yaml":   "YAML",
	"xml":    "XML",
	"html":   "HTML",
	"css":    "CSS",
	"csv":    "CSV",
	"ssh":    "SSH",
	"jwt":    "JWT",
	"oauth":  "OAuth",
	"oauth2": "OAuth2",
	"spa":    "SPA",
	"cors":   "CORS",
	"rbac":   "RBAC",
	"crud":   "CRUD",
	"cli":    "CLI",
	"cpu":    "CPU",
	"gpu":    "GPU",
	"os":     "OS",
	"s3":     "S3",
	"gcs":    "GCS",
	"bq":     "BQ",
	"ttl":    "TTL",
	"rgb":    "RGB",
	"rgba":   "RGBA",
	"sse":    "SSE",
	"ws":     "WS",
	"smtp":   "SMTP",
	"imap":   "IMAP",
	"pop3":   "POP3",
	"dns":    "DNS",
	"vpn":    "VPN",
	"cmd":    "Cmd",
	"ctx":    "Ctx",
	"cfg":    "Cfg",
	"env":    "Env",
	"io":     "IO",
	"ok":     "OK",
	"ui":     "UI",
}

func snakeToPascal(s string) string {
	parts := strings.Split(s, "_")
	var sb strings.Builder
	for _, p := range parts {
		if p == "" {
			continue
		}
		if abbr, ok := commonAbbrevs[strings.ToLower(p)]; ok {
			sb.WriteString(abbr)
		} else {
			sb.WriteString(strings.ToUpper(p[:1]) + p[1:])
		}
	}
	return sb.String()
}