feat(assets): add native Go asset/file loader for migrate-apply

Implements a new `relspec assets` command (list/execute subcommands) that
loads local binary and text asset files into PostgreSQL by binding file bytes
as native pgx query parameters — never as SQL text literals — so binary data
stays byte-exact with no escaping overhead.

Key design points:
- YAML manifest (assets.yaml) colocated with files describes each entry:
  file path, SQL call with :bytes/:filename/:param named placeholders, and
  optional static params map.
- Placeholder substitution converts :name to positional $N params; PostgreSQL
  ::cast syntax is protected before substitution to avoid false matches.
- Directory scan follows the existing {priority}_{sequence}_{name} naming
  convention, enabling asset-loading steps to be correctly interleaved with
  relspec scripts execute in a migrate-apply pipeline.
- Symlink components and path traversal (../) are silently skipped to prevent
  directory escape attacks.
- 14 unit tests cover manifest loading, directory scanning, ordering, symlink
  skipping, path traversal rejection, placeholder substitution edge cases
  (repeated, cast protection, binary byte-exact, unknown).

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

pkg/assetloader/executor.go

@@ -0,0 +1,107 @@
+package assetloader
+
+import (
+	"context"
+	"fmt"
+	"os"
+	"path/filepath"
+	"regexp"
+	"strings"
+
+	"github.com/jackc/pgx/v5"
+)
+
+// namedPlaceholder matches :identifier patterns (but not ::cast syntax).
+var namedPlaceholder = regexp.MustCompile(`:([a-zA-Z_][a-zA-Z0-9_]*)`)
+
+// pgCastMarker temporarily replaces :: to protect PostgreSQL cast syntax.
+const pgCastMarker = "\x00PGCAST\x00"
+
+// BuildQuery converts a SQL call that uses :name named placeholders into a
+// pgx-compatible positional-parameter query ($1, $2, …) and returns the
+// corresponding argument slice.
+//
+// Built-in placeholders:
+//   - :bytes    → fileBytes ([]byte)
+//   - :filename → filename (string, base name only)
+//   - :any_key  → staticParams["any_key"] (string)
+//
+// A placeholder that appears more than once maps to the same $N. An unknown
+// placeholder (not built-in and not in staticParams) returns an error.
+// PostgreSQL cast syntax (::type) is left untouched.
+func BuildQuery(call string, fileBytes []byte, filename string, staticParams map[string]string) (string, []any, error) {
+	// Protect :: casts before running the placeholder regex.
+	protected := strings.ReplaceAll(call, "::", pgCastMarker)
+
+	paramIndex := map[string]int{} // name → 1-based position
+	var args []any
+	var firstErr error
+
+	result := namedPlaceholder.ReplaceAllStringFunc(protected, func(match string) string {
+		if firstErr != nil {
+			return match
+		}
+		name := match[1:] // strip leading ':'
+
+		// Return existing positional param for repeated placeholders.
+		if idx, seen := paramIndex[name]; seen {
+			return fmt.Sprintf("$%d", idx)
+		}
+
+		// Resolve the placeholder value.
+		var val any
+		switch name {
+		case "bytes":
+			val = fileBytes
+		case "filename":
+			val = filename
+		default:
+			if staticParams != nil {
+				if v, ok := staticParams[name]; ok {
+					val = v
+				}
+			}
+			if val == nil {
+				firstErr = fmt.Errorf("unknown placeholder %q in SQL call (not a built-in and not listed in params)", match)
+				return match
+			}
+		}
+
+		idx := len(args) + 1
+		paramIndex[name] = idx
+		args = append(args, val)
+		return fmt.Sprintf("$%d", idx)
+	})
+
+	if firstErr != nil {
+		return "", nil, firstErr
+	}
+
+	// Restore :: casts.
+	result = strings.ReplaceAll(result, pgCastMarker, "::")
+
+	return result, args, nil
+}
+
+// ExecuteItem reads the asset file referenced by item.Entry.File (which is the
+// absolute path set by ScanDir) and executes the configured SQL call via conn.
+// The file's raw bytes are bound as a []byte parameter — no encoding or escaping.
+func ExecuteItem(ctx context.Context, conn *pgx.Conn, item Item) error {
+	data, err := os.ReadFile(item.Entry.File)
+	if err != nil {
+		return fmt.Errorf("reading asset file %s: %w", item.Entry.File, err)
+	}
+
+	filename := filepath.Base(item.Entry.File)
+
+	sql, args, err := BuildQuery(item.Entry.Call, data, filename, item.Entry.Params)
+	if err != nil {
+		return fmt.Errorf("building query for %s: %w", filename, err)
+	}
+
+	if _, err := conn.Exec(ctx, sql, args...); err != nil {
+		return fmt.Errorf("executing asset %s: %w", filename, err)
+	}
+
+	return nil
+}

pkg/assetloader/loader.go

@@ -0,0 +1,200 @@
+// Package assetloader implements a native Go asset/file loader that binds
+// local binary and text files as pgx query parameters during database seeding.
+// Files are bound as actual []byte query parameters — never converted to SQL
+// text literals — so binary data stays byte-exact and no escaping is needed.
+//
+// Manifests are small YAML files (assets.yaml) that describe, per file, the
+// SQL call to invoke and the named placeholders for :bytes, :filename, and any
+// static column values. Manifests live inside directories that follow the same
+// {priority}_{sequence}_{name} naming convention used by the sqldir reader,
+// so asset-loading steps can be interleaved with SQL scripts in a migrate-apply
+// run.
+package assetloader
+
+import (
+	"fmt"
+	"os"
+	"path/filepath"
+	"regexp"
+	"sort"
+	"strconv"
+	"strings"
+
+	"gopkg.in/yaml.v3"
+)
+
+// ManifestEntry describes a single file to load from an assets.yaml manifest.
+type ManifestEntry struct {
+	// File is the path to the asset file, relative to the manifest directory.
+	File string `yaml:"file"`
+	// Call is the SQL statement to execute. Use :bytes for file content,
+	// :filename for the base name, and :param_name for static params.
+	Call string `yaml:"call"`
+	// Params holds optional static named parameters referenced in Call.
+	Params map[string]string `yaml:"params,omitempty"`
+}
+
+// Item combines a manifest entry with its ordering metadata and the resolved
+// directory where the manifest and asset file reside.
+type Item struct {
+	// Priority and Sequence come from the parent directory's naming pattern.
+	Priority int
+	Sequence uint
+	// DirName is the last path component of the manifest's directory.
+	DirName string
+	// Dir is the absolute path to the directory containing assets.yaml and files.
+	Dir string
+	// Entry is the parsed manifest entry.
+	Entry ManifestEntry
+}
+
+// dirPattern matches {priority}_{sequence}_{name} or {priority}-{sequence}-{name}
+// directory names, e.g. "1_010_seed_templates" or "2-001-branding".
+var dirPattern = regexp.MustCompile(`^(\d+)[_-](\d+)[_-](.+)$`)
+
+// LoadManifest reads and parses the assets.yaml file in dir, returning
+// the ordered list of manifest entries. Returns an error if assets.yaml is
+// absent or contains invalid YAML.
+func LoadManifest(dir string) ([]ManifestEntry, error) {
+	manifestPath := filepath.Join(dir, "assets.yaml")
+	data, err := os.ReadFile(manifestPath)
+	if err != nil {
+		return nil, fmt.Errorf("reading %s: %w", manifestPath, err)
+	}
+
+	var entries []ManifestEntry
+	if err := yaml.Unmarshal(data, &entries); err != nil {
+		return nil, fmt.Errorf("parsing %s: %w", manifestPath, err)
+	}
+
+	return entries, nil
+}
+
+// ScanDir recursively walks baseDir, finds all assets.yaml manifests, resolves
+// each file entry (skipping symlinks and path traversal), and returns the
+// resulting Items sorted by (Priority, Sequence, DirName).
+//
+// Each manifest must reside in a directory whose name follows the
+// {priority}_{sequence}_{name} pattern. Manifests in directories that do not
+// follow this convention are assigned Priority=0, Sequence=0 and sorted last.
+func ScanDir(baseDir string) ([]Item, error) {
+	absBase, err := filepath.Abs(baseDir)
+	if err != nil {
+		return nil, fmt.Errorf("resolving base dir: %w", err)
+	}
+
+	var items []Item
+
+	err = filepath.WalkDir(absBase, func(path string, d os.DirEntry, walkErr error) error {
+		if walkErr != nil {
+			return walkErr
+		}
+		if d.IsDir() {
+			return nil
+		}
+		if d.Name() != "assets.yaml" {
+			return nil
+		}
+
+		manifestDir := filepath.Dir(path)
+		priority, sequence, dirName := parseDirName(filepath.Base(manifestDir))
+
+		entries, err := LoadManifest(manifestDir)
+		if err != nil {
+			return fmt.Errorf("loading manifest in %s: %w", manifestDir, err)
+		}
+
+		for _, entry := range entries {
+			if entry.File == "" || entry.Call == "" {
+				continue
+			}
+
+			// Resolve and validate the asset file path.
+			resolved, skip, err := resolveAssetPath(absBase, manifestDir, entry.File)
+			if err != nil {
+				return err
+			}
+			if skip {
+				continue
+			}
+
+			items = append(items, Item{
+				Priority: priority,
+				Sequence: sequence,
+				DirName:  dirName,
+				Dir:      manifestDir,
+				Entry: ManifestEntry{
+					File:   resolved, // absolute path, safe to read
+					Call:   entry.Call,
+					Params: entry.Params,
+				},
+			})
+		}
+
+		return nil
+	})
+	if err != nil {
+		return nil, err
+	}
+
+	sort.SliceStable(items, func(i, j int) bool {
+		if items[i].Priority != items[j].Priority {
+			return items[i].Priority < items[j].Priority
+		}
+		if items[i].Sequence != items[j].Sequence {
+			return items[i].Sequence < items[j].Sequence
+		}
+		return items[i].DirName < items[j].DirName
+	})
+
+	return items, nil
+}
+
+// parseDirName extracts (priority, sequence, name) from a directory name that
+// follows the {priority}[_-]{sequence}[_-]{name} convention. Returns (0, 0, dir)
+// when the name does not match.
+func parseDirName(dir string) (priority int, sequence uint, name string) {
+	m := dirPattern.FindStringSubmatch(dir)
+	if m == nil {
+		return 0, 0, dir
+	}
+	p, _ := strconv.Atoi(m[1])
+	s, _ := strconv.ParseUint(m[2], 10, 64)
+	return p, uint(s), m[3]
+}
+
+// resolveAssetPath resolves a manifest-relative file path and checks that:
+//   - it does not escape the base directory (path traversal prevention)
+//   - none of its path components are symlinks
+//
+// Returns the absolute path, a skip flag (true when the entry should be silently
+// dropped), and any hard error.
+func resolveAssetPath(absBase, manifestDir, file string) (absPath string, skip bool, err error) {
+	// Clean and join before any symlink resolution so we can detect traversal.
+	joined := filepath.Join(manifestDir, filepath.Clean(file))
+
+	// Ensure the cleaned path is still inside absBase.
+	rel, err := filepath.Rel(absBase, joined)
+	if err != nil || strings.HasPrefix(rel, "..") {
+		// Path escapes the base directory; skip silently.
+		return "", true, nil
+	}
+
+	// Walk each component to detect symlinks.
+	parts := strings.Split(rel, string(filepath.Separator))
+	current := absBase
+	for _, part := range parts {
+		current = filepath.Join(current, part)
+		info, statErr := os.Lstat(current)
+		if statErr != nil {
+			// File doesn't exist; skip.
+			return "", true, nil
+		}
+		if info.Mode()&os.ModeSymlink != 0 {
+			// Symlink in path; skip silently.
+			return "", true, nil
+		}
+	}
+
+	return joined, false, nil
+}

pkg/assetloader/loader_test.go

@@ -0,0 +1,341 @@
+package assetloader_test
+
+import (
+	"os"
+	"path/filepath"
+	"testing"
+
+	"git.warky.dev/wdevs/relspecgo/pkg/assetloader"
+)
+
+func TestLoadManifest_ValidList(t *testing.T) {
+	dir := t.TempDir()
+	writeFile(t, dir, "assets.yaml", `
+- file: hello.txt
+  call: INSERT INTO files (name, data) VALUES (:filename, :bytes)
+- file: logo.png
+  call: UPDATE branding SET logo = :bytes WHERE id = 1
+`)
+	writeFile(t, dir, "hello.txt", "hello world")
+	writeFile(t, dir, "logo.png", "\x89PNG\r\n\x1a\n")
+
+	m, err := assetloader.LoadManifest(dir)
+	if err != nil {
+		t.Fatalf("LoadManifest failed: %v", err)
+	}
+	if len(m) != 2 {
+		t.Fatalf("expected 2 entries, got %d", len(m))
+	}
+	if m[0].File != "hello.txt" {
+		t.Errorf("entry 0 file: got %q, want %q", m[0].File, "hello.txt")
+	}
+	if m[1].File != "logo.png" {
+		t.Errorf("entry 1 file: got %q, want %q", m[1].File, "logo.png")
+	}
+}
+
+func TestLoadManifest_WithStaticParams(t *testing.T) {
+	dir := t.TempDir()
+	writeFile(t, dir, "assets.yaml", `
+- file: template.md
+  call: INSERT INTO templates (owner_id, name, data) VALUES (:owner_id, :filename, :bytes)
+  params:
+    owner_id: "42"
+`)
+	writeFile(t, dir, "template.md", "# Template")
+
+	m, err := assetloader.LoadManifest(dir)
+	if err != nil {
+		t.Fatalf("LoadManifest failed: %v", err)
+	}
+	if len(m) != 1 {
+		t.Fatalf("expected 1 entry, got %d", len(m))
+	}
+	if m[0].Params["owner_id"] != "42" {
+		t.Errorf("static param owner_id: got %q, want %q", m[0].Params["owner_id"], "42")
+	}
+}
+
+func TestLoadManifest_MissingFile(t *testing.T) {
+	dir := t.TempDir()
+	// No assets.yaml present
+	_, err := assetloader.LoadManifest(dir)
+	if err == nil {
+		t.Fatal("expected error for missing assets.yaml, got nil")
+	}
+}
+
+func TestLoadManifest_InvalidYAML(t *testing.T) {
+	dir := t.TempDir()
+	writeFile(t, dir, "assets.yaml", `{not: [valid yaml`)
+	_, err := assetloader.LoadManifest(dir)
+	if err == nil {
+		t.Fatal("expected error for invalid YAML, got nil")
+	}
+}
+
+func TestScanDir_FindsManifests(t *testing.T) {
+	root := t.TempDir()
+
+	// Directory named with priority-sequence pattern
+	dir1 := filepath.Join(root, "1_010_seed_templates")
+	if err := os.MkdirAll(dir1, 0o755); err != nil {
+		t.Fatal(err)
+	}
+	writeFile(t, dir1, "assets.yaml", `
+- file: a.txt
+  call: INSERT INTO t (data) VALUES (:bytes)
+`)
+	writeFile(t, dir1, "a.txt", "aaa")
+
+	dir2 := filepath.Join(root, "2_001_branding")
+	if err := os.MkdirAll(dir2, 0o755); err != nil {
+		t.Fatal(err)
+	}
+	writeFile(t, dir2, "assets.yaml", `
+- file: logo.png
+  call: UPDATE branding SET logo = :bytes
+`)
+	writeFile(t, dir2, "logo.png", "PNG")
+
+	items, err := assetloader.ScanDir(root)
+	if err != nil {
+		t.Fatalf("ScanDir failed: %v", err)
+	}
+	if len(items) != 2 {
+		t.Fatalf("expected 2 items, got %d", len(items))
+	}
+	// Should be ordered by priority then sequence
+	if items[0].Priority != 1 || items[0].Sequence != 10 {
+		t.Errorf("item[0]: got priority=%d seq=%d, want 1,10", items[0].Priority, items[0].Sequence)
+	}
+	if items[1].Priority != 2 || items[1].Sequence != 1 {
+		t.Errorf("item[1]: got priority=%d seq=%d, want 2,1", items[1].Priority, items[1].Sequence)
+	}
+}
+
+func TestScanDir_OrdersByPriorityThenSequence(t *testing.T) {
+	root := t.TempDir()
+
+	for _, d := range []string{"2_002_b", "1_001_a", "2_001_c", "1_002_d"} {
+		dir := filepath.Join(root, d)
+		if err := os.MkdirAll(dir, 0o755); err != nil {
+			t.Fatal(err)
+		}
+		writeFile(t, dir, "assets.yaml", `
+- file: x.txt
+  call: SELECT :bytes
+`)
+		writeFile(t, dir, "x.txt", "x")
+	}
+
+	items, err := assetloader.ScanDir(root)
+	if err != nil {
+		t.Fatalf("ScanDir failed: %v", err)
+	}
+	if len(items) != 4 {
+		t.Fatalf("expected 4 items, got %d", len(items))
+	}
+
+	type ps struct{ p int; s uint }
+	want := []ps{{1, 1}, {1, 2}, {2, 1}, {2, 2}}
+	for i, w := range want {
+		got := ps{items[i].Priority, items[i].Sequence}
+		if got != w {
+			t.Errorf("items[%d]: got {%d,%d}, want {%d,%d}", i, got.p, got.s, w.p, w.s)
+		}
+	}
+}
+
+func TestScanDir_SkipsSymlinks(t *testing.T) {
+	root := t.TempDir()
+
+	dir1 := filepath.Join(root, "1_001_real")
+	if err := os.MkdirAll(dir1, 0o755); err != nil {
+		t.Fatal(err)
+	}
+	writeFile(t, dir1, "assets.yaml", `
+- file: a.txt
+  call: SELECT :bytes
+`)
+	writeFile(t, dir1, "a.txt", "real")
+
+	// Symlink to an asset file - should be skipped during file read
+	realFile := filepath.Join(root, "real.txt")
+	writeFile(t, root, "real.txt", "symlink target")
+	symlink := filepath.Join(dir1, "link.txt")
+	if err := os.Symlink(realFile, symlink); err != nil {
+		t.Skip("symlinks not supported:", err)
+	}
+
+	// Add a manifest entry that references the symlink
+	writeFile(t, dir1, "assets.yaml", `
+- file: a.txt
+  call: SELECT :bytes
+- file: link.txt
+  call: SELECT :bytes
+`)
+
+	items, err := assetloader.ScanDir(root)
+	if err != nil {
+		t.Fatalf("ScanDir failed: %v", err)
+	}
+	// The symlink entry should be skipped; only a.txt should remain
+	if len(items) != 1 {
+		t.Fatalf("expected 1 item after symlink skip, got %d", len(items))
+	}
+	if filepath.Base(items[0].Entry.File) != "a.txt" {
+		t.Errorf("expected non-symlink entry, got %q", items[0].Entry.File)
+	}
+}
+
+func TestScanDir_RejectsPathTraversal(t *testing.T) {
+	root := t.TempDir()
+
+	dir1 := filepath.Join(root, "1_001_evil")
+	if err := os.MkdirAll(dir1, 0o755); err != nil {
+		t.Fatal(err)
+	}
+	writeFile(t, dir1, "assets.yaml", `
+- file: ../../etc/passwd
+  call: SELECT :bytes
+`)
+
+	items, err := assetloader.ScanDir(root)
+	if err != nil {
+		t.Fatalf("ScanDir failed: %v", err)
+	}
+	// Path traversal entry should be skipped
+	if len(items) != 0 {
+		t.Fatalf("expected 0 items after path traversal rejection, got %d", len(items))
+	}
+}
+
+func TestBuildQuery_BasicPlaceholders(t *testing.T) {
+	sql, args, err := assetloader.BuildQuery(
+		"INSERT INTO t (name, data) VALUES (:filename, :bytes)",
+		[]byte("hello"),
+		"hello.txt",
+		nil,
+	)
+	if err != nil {
+		t.Fatalf("BuildQuery failed: %v", err)
+	}
+	if sql != "INSERT INTO t (name, data) VALUES ($1, $2)" {
+		t.Errorf("unexpected SQL: %s", sql)
+	}
+	if len(args) != 2 {
+		t.Fatalf("expected 2 args, got %d", len(args))
+	}
+	if string(args[0].(string)) != "hello.txt" {
+		t.Errorf("args[0]: got %q, want %q", args[0], "hello.txt")
+	}
+	if string(args[1].([]byte)) != "hello" {
+		t.Errorf("args[1]: got %v, want %v", args[1], []byte("hello"))
+	}
+}
+
+func TestBuildQuery_StaticParams(t *testing.T) {
+	sql, args, err := assetloader.BuildQuery(
+		"INSERT INTO t (owner, name, data) VALUES (:owner_id, :filename, :bytes)",
+		[]byte("data"),
+		"file.bin",
+		map[string]string{"owner_id": "99"},
+	)
+	if err != nil {
+		t.Fatalf("BuildQuery failed: %v", err)
+	}
+	if sql != "INSERT INTO t (owner, name, data) VALUES ($1, $2, $3)" {
+		t.Errorf("unexpected SQL: %s", sql)
+	}
+	if len(args) != 3 {
+		t.Fatalf("expected 3 args, got %d: %v", len(args), args)
+	}
+	if args[0].(string) != "99" {
+		t.Errorf("args[0]: got %q, want %q", args[0], "99")
+	}
+}
+
+func TestBuildQuery_RepeatedPlaceholder(t *testing.T) {
+	sql, args, err := assetloader.BuildQuery(
+		"SELECT length(:bytes), encode(:bytes, 'base64')",
+		[]byte("abc"),
+		"f.bin",
+		nil,
+	)
+	if err != nil {
+		t.Fatalf("BuildQuery failed: %v", err)
+	}
+	// :bytes appears twice but maps to same $1
+	if sql != "SELECT length($1), encode($1, 'base64')" {
+		t.Errorf("unexpected SQL: %s", sql)
+	}
+	if len(args) != 1 {
+		t.Fatalf("expected 1 arg, got %d", len(args))
+	}
+}
+
+func TestBuildQuery_PostgresCastNotMatched(t *testing.T) {
+	// ::text should NOT be treated as a placeholder
+	sql, args, err := assetloader.BuildQuery(
+		"SELECT :bytes::text, :filename",
+		[]byte("data"),
+		"f.txt",
+		nil,
+	)
+	if err != nil {
+		t.Fatalf("BuildQuery failed: %v", err)
+	}
+	if sql != "SELECT $1::text, $2" {
+		t.Errorf("unexpected SQL: %s", sql)
+	}
+	if len(args) != 2 {
+		t.Fatalf("expected 2 args, got %d", len(args))
+	}
+}
+
+func TestBuildQuery_UnknownPlaceholder(t *testing.T) {
+	_, _, err := assetloader.BuildQuery(
+		"SELECT :unknown_param",
+		[]byte("data"),
+		"f.txt",
+		nil,
+	)
+	if err == nil {
+		t.Fatal("expected error for unknown placeholder, got nil")
+	}
+}
+
+func TestBuildQuery_BinaryFileByteExact(t *testing.T) {
+	// Binary data with null bytes, high bytes - must pass through unchanged
+	binary := []byte{0x00, 0xFF, 0x80, 0x01, 0xFE}
+	_, args, err := assetloader.BuildQuery(
+		"INSERT INTO blobs (data) VALUES (:bytes)",
+		binary,
+		"blob.bin",
+		nil,
+	)
+	if err != nil {
+		t.Fatalf("BuildQuery failed: %v", err)
+	}
+	if len(args) != 1 {
+		t.Fatalf("expected 1 arg")
+	}
+	got := args[0].([]byte)
+	if len(got) != len(binary) {
+		t.Fatalf("byte count: got %d, want %d", len(got), len(binary))
+	}
+	for i, b := range binary {
+		if got[i] != b {
+			t.Errorf("byte[%d]: got 0x%02x, want 0x%02x", i, got[i], b)
+		}
+	}
+}
+
+// writeFile is a test helper that writes content to a file.
+func writeFile(t *testing.T, dir, name, content string) {
+	t.Helper()
+	if err := os.WriteFile(filepath.Join(dir, name), []byte(content), 0o644); err != nil {
+		t.Fatalf("writeFile %s: %v", name, err)
+	}
+}