Revert "refactor: Improve robots.txt parsing and caching"

This reverts commit 6a96fb26cc.

gemini/network.go

@@ -6,6 +6,7 @@ import (
 	"gemini-grc/config"
 	"io"
 	"net"
+	go_url "net/url"
 	"regexp"
 	"slices"
 	"strconv"
@@ -14,6 +15,14 @@ import (
 	"github.com/guregu/null/v5"
 )
 
+type GeminiPageData struct {
+	ResponseCode int
+	MimeType     string
+	Lang         string
+	GemText      string
+	Data         []byte
+}
+
 // Resolve the URL hostname and
 // check if we already have an open
 // connection to this host.
@@ -31,51 +40,53 @@ func getHostIPAddresses(hostname string) ([]string, error) {
 	return addrs, nil
 }
 
-// Connect to given URL, using the Gemini protocol.
-// Return a Snapshot with the data or the error.
-// Any errors are stored within the snapshot.
-func Visit(s *Snapshot) {
+func ConnectAndGetData(url string) ([]byte, error) {
+	parsedUrl, err := go_url.Parse(url)
+	if err != nil {
+		return nil, fmt.Errorf("Could not parse URL, error %w", err)
+	}
+	hostname := parsedUrl.Hostname()
+	port := parsedUrl.Port()
+	if port == "" {
+		port = "1965"
+	}
+	host := fmt.Sprintf("%s:%s", hostname, port)
 	// Establish the underlying TCP connection.
-	host := fmt.Sprintf("%s:%d", s.Host, s.URL.Port)
 	dialer := &net.Dialer{
-		Timeout:   time.Duration(config.CONFIG.ResponseTimeout) * time.Second, // Set the overall connection timeout
-		KeepAlive: 30 * time.Second,
+		Timeout:   time.Duration(config.CONFIG.ResponseTimeout) * time.Second,
+		KeepAlive: 10 * time.Second,
 	}
 	conn, err := dialer.Dial("tcp", host)
 	if err != nil {
-		s.Error = null.StringFrom(fmt.Sprintf("TCP connection failed: %v", err))
-		return
+		return nil, fmt.Errorf("TCP connection failed: %w", err)
 	}
 	// Make sure we always close the connection.
 	defer func() {
 		err := conn.Close()
 		if err != nil {
-			s.Error = null.StringFrom(fmt.Sprintf("Error closing connection: %s", err))
+			// Do nothing! Connection will timeout eventually if still open somehow.
 		}
 	}()
 
 	// Set read and write timeouts on the TCP connection.
 	err = conn.SetReadDeadline(time.Now().Add(time.Duration(config.CONFIG.ResponseTimeout) * time.Second))
 	if err != nil {
-		s.Error = null.StringFrom(fmt.Sprintf("Error setting connection deadline: %s", err))
-		return
+		return nil, fmt.Errorf("Error setting connection deadline: %w", err)
 	}
 	err = conn.SetWriteDeadline(time.Now().Add(time.Duration(config.CONFIG.ResponseTimeout) * time.Second))
 	if err != nil {
-		s.Error = null.StringFrom(fmt.Sprintf("Error setting connection deadline: %s", err))
-		return
+		return nil, fmt.Errorf("Error setting connection deadline: %w", err)
 	}
 
 	// Perform the TLS handshake
 	tlsConfig := &tls.Config{
 		InsecureSkipVerify: true,              // Accept all TLS certs, even if insecure.
-		ServerName:         s.URL.Hostname, // SNI
+		ServerName:         parsedUrl.Hostname(), // SNI should not include port
 		// MinVersion:         tls.VersionTLS12, // Use a minimum TLS version. Warning breaks a lot of sites.
 	}
 	tlsConn := tls.Client(conn, tlsConfig)
 	if err := tlsConn.Handshake(); err != nil {
-		s.Error = null.StringFrom(fmt.Sprintf("TLS handshake error: %v", err))
-		return
+		return nil, fmt.Errorf("TLS handshake error: %w", err)
 	}
 
 	// We read `buf`-sized chunks and add data to `data`.
@@ -83,10 +94,9 @@ func Visit(s *Snapshot) {
 	var data []byte
 
 	// Send Gemini request to trigger server response.
-	_, err = tlsConn.Write([]byte(fmt.Sprintf("%s\r\n", s.URL.String())))
+	_, err = tlsConn.Write([]byte(fmt.Sprintf("%s\r\n", url)))
 	if err != nil {
-		s.Error = null.StringFrom(fmt.Sprintf("Error sending network request: %s", err))
-		return
+		return nil, fmt.Errorf("Error sending network request: %w", err)
 	}
 	// Read response bytes in len(buf) byte chunks
 	for {
@@ -96,21 +106,40 @@ func Visit(s *Snapshot) {
 		}
 		if len(data) > config.CONFIG.MaxResponseSize {
 			data = []byte{}
-			s.Error = null.StringFrom(fmt.Sprintf("Response size exceeded maximum of %d bytes", config.CONFIG.MaxResponseSize))
+			return nil, fmt.Errorf("Response size exceeded maximum of %d bytes", config.CONFIG.MaxResponseSize)
 		}
 		if err != nil {
 			if err == io.EOF {
 				break
 			} else {
-				s.Error = null.StringFrom(fmt.Sprintf("Network error: %s", err))
-				return
+				return nil, fmt.Errorf("Network error: %s", err)
 			}
 		}
 	}
-	// Great, response data received.
-	err = processResponse(s, data)
+	return data, nil
+}
+
+// Connect to given URL, using the Gemini protocol.
+// Mutate given Snapshot with the data or the error.
+func Visit(s *Snapshot) {
+	data, err := ConnectAndGetData(s.URL.String())
 	if err != nil {
 		s.Error = null.StringFrom(err.Error())
+		return
+	}
+	pageData, err := processData(data)
+	if err != nil {
+		s.Error = null.StringFrom(err.Error())
+		return
+	}
+	s.ResponseCode = null.IntFrom(int64(pageData.ResponseCode))
+	s.MimeType = null.StringFrom(pageData.MimeType)
+	s.Lang = null.StringFrom(pageData.Lang)
+	if pageData.GemText != "" {
+		s.GemText = null.StringFrom(string(pageData.GemText))
+	}
+	if pageData.Data != nil {
+		s.Data = null.ValueFrom(pageData.Data)
 	}
 	return
 }
@@ -118,31 +147,33 @@ func Visit(s *Snapshot) {
 // Update given snapshot with the
 // Gemini header data: response code,
 // mime type and lang (optional)
-func processResponse(snapshot *Snapshot, data []byte) error {
+func processData(data []byte) (*GeminiPageData, error) {
 	headers, body, err := getHeadersAndData(data)
 	if err != nil {
-		return err
+		return nil, err
 	}
 	code, mimeType, lang := getMimeTypeAndLang(headers)
 	geminiError := checkGeminiStatusCode(code)
 	if geminiError != nil {
-		return geminiError
+		return nil, geminiError
+	}
+	pageData := GeminiPageData{
+		ResponseCode: code,
+		MimeType:     mimeType,
+		Lang:         lang,
 	}
-	snapshot.ResponseCode = null.IntFrom(int64(code))
-	snapshot.MimeType = null.StringFrom(mimeType)
-	snapshot.Lang = null.StringFrom(lang)
 	// If we've got a Gemini document, populate
 	// `GemText` field, otherwise raw data goes to `Data`.
 	if mimeType == "text/gemini" {
 		validBody, err := EnsureValidUTF8(body)
 		if err != nil {
-			return fmt.Errorf("UTF-8 error: %w", err)
+			return nil, fmt.Errorf("UTF-8 error: %w", err)
 		}
-		snapshot.GemText = null.StringFrom(string(validBody))
+		pageData.GemText = validBody
 	} else {
-		snapshot.Data = null.ValueFrom(body)
+		pageData.Data = body
 	}
-	return nil
+	return &pageData, nil
 }
 
 // Checks for a Gemini header, which is

gemini/persistence.go

@@ -57,16 +57,16 @@ func SaveSnapshotToDB(tx *sqlx.Tx, s *Snapshot) error {
 	return nil
 }
 
-func SaveLinkToDB(tx *sqlx.Tx, s *Snapshot) error {
+func SaveLinksToDB(tx *sqlx.Tx, snapshots []*Snapshot) error {
 	query := `
         INSERT INTO snapshots (uid, url, host, timestamp, mimetype, data, gemtext, links, lang, response_code, error)
         VALUES (:uid, :url, :host, :timestamp, :mimetype, :data, :gemtext, :links, :lang, :response_code, :error)
         ON CONFLICT (uid) DO NOTHING
     `
-	_, err := tx.NamedExec(query, s)
+	_, err := tx.NamedExec(query, snapshots)
 	if err != nil {
-		logging.LogError("[%s] [%s] Error upserting snapshot: %w", s.URL, s.MimeType.String, err)
-		return fmt.Errorf("DB error: %w", err) // Return the error instead of panicking
+		logging.LogError("Error batch inserting snapshots: %w", err)
+		return fmt.Errorf("DB error: %w", err)
 	}
 	return nil
 }

gemini/robots.go

@@ -0,0 +1,31 @@
+package gemini
+
+import (
+	"fmt"
+	"strings"
+)
+
+// Takes robots.txt content and a host, and
+// returns a list of full URLs that shouldn't
+// be visited.
+// TODO Also take into account the user agent?
+// Check gemini://geminiprotocol.net/docs/companion/robots.gmi
+func ParseRobotsTxt(content string, host string) []string {
+	var disallowedPaths []string
+	for _, line := range strings.Split(content, "\n") {
+		line = strings.TrimSpace(line)
+		line = strings.ToLower(line)
+		if strings.HasPrefix(line, "disallow:") {
+			parts := strings.SplitN(line, ":", 2)
+			if len(parts) == 2 {
+				path := strings.TrimSpace(parts[1])
+				if path != "" {
+					// Construct full Gemini URL
+					disallowedPaths = append(disallowedPaths,
+						fmt.Sprintf("gemini://%s%s", host, path))
+				}
+			}
+		}
+	}
+	return disallowedPaths
+}

gemini/robots_test.go

@@ -0,0 +1,25 @@
+package gemini
+
+import (
+	"testing"
+	"reflect"
+)
+
+func TestParseRobotsTxt(t *testing.T) {
+	input := `User-agent: *
+Disallow: /cgi-bin/wp.cgi/view
+Disallow: /cgi-bin/wp.cgi/media
+User-agent: googlebot
+Disallow: /admin/`
+
+	expected := []string{
+		"gemini://example.com/cgi-bin/wp.cgi/view",
+		"gemini://example.com/cgi-bin/wp.cgi/media",
+	}
+
+	result := ParseRobotsTxt(input, "example.com")
+
+	if !reflect.DeepEqual(result, expected) {
+		t.Errorf("ParseRobotsTxt() = %v, want %v", result, expected)
+	}
+}

gemini/worker.go

@@ -84,19 +84,31 @@ func workOnSnapshot(id int, tx *sqlx.Tx, s *Snapshot) (err error) {
 		return fmt.Errorf("[%d] DB Error: %w", id, err)
 	}
 
-	// Store links
+	// Store links in batch
 	if s.Links != nil {
+		var batchSnapshots []*Snapshot
+		timestamp := null.TimeFrom(time.Now())
+		
 		for _, link := range *s.Links {
-			newSnapshot := Snapshot{UID: uid.UID(), URL: link, Host: link.Hostname, Timestamp: null.TimeFrom(time.Now())}
 			if shouldPersistURL(tx, link) {
-				logging.LogDebug("[%d] Saving link %s", id, link)
-				err = SaveLinkToDB(tx, &newSnapshot)
+				newSnapshot := &Snapshot{
+					UID: uid.UID(),
+					URL: link,
+					Host: link.Hostname,
+					Timestamp: timestamp,
+				}
+				batchSnapshots = append(batchSnapshots, newSnapshot)
+			}
+		}
+		
+		if len(batchSnapshots) > 0 {
+			logging.LogDebug("[%d] Batch saving %d links", id, len(batchSnapshots))
+			err = SaveLinksToDB(tx, batchSnapshots)
 			if err != nil {
 				return fmt.Errorf("[%d] DB Error: %w", id, err)
 			}
 		}
 	}
-	}
 	return nil
 }