Move all regexp MustCompiles outside (#510)

Lots of CPU time spent compiling reused regexps in CPU profiles

atproto/syntax/cid.go

@@ -13,14 +13,16 @@ import (
 // Always use [ParseCID] instead of wrapping strings directly, especially when working with network input.
 type CID string
 
+var cidRegex = regexp.MustCompile(`^[a-zA-Z0-9+=]{8,256}$`)
+
 func ParseCID(raw string) (CID, error) {
 	if len(raw) > 256 {
 		return "", fmt.Errorf("CID is too long (256 chars max)")
 	}
 	if len(raw) < 8 {
 		return "", fmt.Errorf("CID is too short (8 chars min)")
 	}
-	var cidRegex = regexp.MustCompile(`^[a-zA-Z0-9+=]{8,256}$`)
+
 	if !cidRegex.MatchString(raw) {
 		return "", fmt.Errorf("CID syntax didn't validate via regex")
 	}

atproto/syntax/datetime.go

@@ -21,11 +21,13 @@ const (
 // Syntax is specified at: https://atproto.com/specs/lexicon#datetime
 type Datetime string
 
+var datetimeRegex = regexp.MustCompile(`^[0-9]{4}-[01][0-9]-[0-3][0-9]T[0-2][0-9]:[0-6][0-9]:[0-6][0-9](.[0-9]{1,20})?(Z|([+-][0-2][0-9]:[0-5][0-9]))$`)
+
 func ParseDatetime(raw string) (Datetime, error) {
 	if len(raw) > 64 {
 		return "", fmt.Errorf("Datetime too long (max 64 chars)")
 	}
-	var datetimeRegex = regexp.MustCompile(`^[0-9]{4}-[01][0-9]-[0-3][0-9]T[0-2][0-9]:[0-6][0-9]:[0-6][0-9](.[0-9]{1,20})?(Z|([+-][0-2][0-9]:[0-5][0-9]))$`)
+
 	if !datetimeRegex.MatchString(raw) {
 		return "", fmt.Errorf("Datetime syntax didn't validate via regex")
 	}
@@ -53,6 +55,8 @@ func ParseDatetimeTime(raw string) (time.Time, error) {
 // Similar to ParseDatetime, but more flexible about some parsing.
 //
 // Note that this may mutate the internal string, so a round-trip will fail. This is intended for working with legacy/broken records, not to be used in an ongoing way.
+var hasTimezoneRegex = regexp.MustCompile(`^.*(([+-]\d\d:?\d\d)|[a-zA-Z])$`)
+
 func ParseDatetimeLenient(raw string) (Datetime, error) {
 	// fast path: it is a valid overall datetime
 	valid, err := ParseDatetime(raw)
@@ -71,7 +75,6 @@ func ParseDatetimeLenient(raw string) (Datetime, error) {
 	}
 
 	// try adding timezone if it is missing
-	var hasTimezoneRegex = regexp.MustCompile(`^.*(([+-]\d\d:?\d\d)|[a-zA-Z])$`)
 	if !hasTimezoneRegex.MatchString(raw) {
 		withTZ, err := ParseDatetime(raw + "Z")
 		if nil == err {

atproto/syntax/did.go

@@ -13,11 +13,12 @@ import (
 // Syntax specification: https://atproto.com/specs/did
 type DID string
 
+var didRegex = regexp.MustCompile(`^did:[a-z]+:[a-zA-Z0-9._:%-]*[a-zA-Z0-9._-]$`)
+
 func ParseDID(raw string) (DID, error) {
 	if len(raw) > 2*1024 {
 		return "", fmt.Errorf("DID is too long (2048 chars max)")
 	}
-	var didRegex = regexp.MustCompile(`^did:[a-z]+:[a-zA-Z0-9._:%-]*[a-zA-Z0-9._-]$`)
 	if !didRegex.MatchString(raw) {
 		return "", fmt.Errorf("DID syntax didn't validate via regex")
 	}

atproto/syntax/language.go

@@ -12,11 +12,12 @@ import (
 // The syntax is BCP-47. This is a partial/naive parsing implementation, designed for fast validation and exact-string passthrough with no normaliztion. For actually working with BCP-47 language specifiers in atproto code bases, we recommend the golang.org/x/text/language package.
 type Language string
 
+var langRegex = regexp.MustCompile(`^(i|[a-z]{2,3})(-[a-zA-Z0-9]+)*$`)
+
 func ParseLanguage(raw string) (Language, error) {
 	if len(raw) > 128 {
 		return "", fmt.Errorf("Language is too long (128 chars max)")
 	}
-	var langRegex = regexp.MustCompile(`^(i|[a-z]{2,3})(-[a-zA-Z0-9]+)*$`)
 	if !langRegex.MatchString(raw) {
 		return "", fmt.Errorf("Language syntax didn't validate via regex")
 	}

atproto/syntax/tid.go

@@ -24,11 +24,12 @@ func Base32Sort() *base32.Encoding {
 // Syntax specification: https://atproto.com/specs/record-key
 type TID string
 
+var tidRegex = regexp.MustCompile(`^[234567abcdefghij][234567abcdefghijklmnopqrstuvwxyz]{12}$`)
+
 func ParseTID(raw string) (TID, error) {
 	if len(raw) != 13 {
 		return "", fmt.Errorf("TID is wrong length (expected 13 chars)")
 	}
-	var tidRegex = regexp.MustCompile(`^[234567abcdefghij][234567abcdefghijklmnopqrstuvwxyz]{12}$`)
 	if !tidRegex.MatchString(raw) {
 		return "", fmt.Errorf("TID syntax didn't validate via regex")
 	}

search/indexing.go

@@ -48,6 +48,8 @@ func (s *Server) deletePost(ctx context.Context, ident *identity.Identity, rkey
 	return nil
 }
 
+var tidRegex = regexp.MustCompile(`^[234567abcdefghijklmnopqrstuvwxyz]{13}$`)
+
 func (s *Server) indexPost(ctx context.Context, ident *identity.Identity, rec *appbsky.FeedPost, path string, rcid cid.Cid) error {
 	ctx, span := tracer.Start(ctx, "indexPost")
 	defer span.End()
@@ -56,7 +58,6 @@ func (s *Server) indexPost(ctx context.Context, ident *identity.Identity, rec *a
 	log := s.logger.With("repo", ident.DID, "path", path, "op", "indexPost")
 	parts := strings.SplitN(path, "/", 3)
 	// TODO: replace with an atproto/syntax package type for TID
-	var tidRegex = regexp.MustCompile(`^[234567abcdefghijklmnopqrstuvwxyz]{13}$`)
 	if len(parts) != 2 || !tidRegex.MatchString(parts[1]) {
 		log.Warn("skipping index post record with weird path/TID", "did", ident.DID, "path", path)
 		return nil