feat: offer a non-Mews web page as a simplified view or in the browser

app/lib/application/actors/fetch_actor.dart

@@ -8,6 +8,7 @@ import 'package:http/http.dart' as http;
 import '../../mews/atom.dart';
 import '../../mews/document.dart';
 import '../../mews/gemtext.dart';
+import '../../mews/html.dart';
 import '../../mews/parser.dart';
 import '../../protocols/gemini.dart' as gemini;
 import '../../protocols/spartan.dart' as spartan;
@@ -153,12 +154,18 @@ class FetchHttpStatusError extends FetchResult {
 }
 
 class FetchParseError extends FetchResult {
-  const FetchParseError(this.message, this.source);
+  const FetchParseError(this.message, {this.simplified, this.declaresMews = false});
 
   final String message;
 
-  /// The markup that failed to parse, for the "show page text" option.
-  final String source;
+  /// The page as the app can show it, for the simplified view, or null when
+  /// nothing legible came out of the markup.
+  final MewsPage? simplified;
+
+  /// Whether the markup declared the Mews conformance marker. A page that
+  /// claims to be Mews and still will not parse has a real markup error; one
+  /// that never claimed it is just an ordinary web page.
+  final bool declaresMews;
 }
 
 /// A Gemini or Spartan status carrying a message for the reader: a temporary
@@ -304,7 +311,8 @@ class FetchHandler with Handler<FetchCommand, FetchResult> {
 
     final newEtag = response.headers['etag'];
     final newLastModified = response.headers['last-modified'];
-    return _parse(response.body, url, etag: newEtag, lastModified: newLastModified);
+    return _parse(response.body, url,
+        bytes: response.bodyBytes, etag: newEtag, lastModified: newLastModified);
   }
 
   /// The page could not be revalidated; use the cached copy when there is one.
@@ -315,14 +323,32 @@ class FetchHandler with Handler<FetchCommand, FetchResult> {
         etag: etag, lastModified: lastModified, notModified: true);
   }
 
+  /// [bytes] are the undecoded body when there is one. The HTML conversion
+  /// reads the document's own charset from them, which an already-decoded
+  /// string has lost.
   FetchResult _parse(String source, String url,
-      {String? etag, String? lastModified, bool notModified = false}) {
+      {List<int>? bytes, String? etag, String? lastModified, bool notModified = false}) {
     try {
       final page = parseMewsPage(source: source, url: url);
       return FetchPageOk(
           page: page, source: source, etag: etag, lastModified: lastModified, notModified: notModified);
     } on MewsParseException catch (e) {
-      return FetchParseError(e.message, source);
+      return _parseFailure(e, source, bytes, url);
+    }
+  }
+
+  /// Markup the strict parse refused is converted for the simplified view, so
+  /// the reader is offered the structure Mews supports rather than nothing.
+  /// Converting here keeps the work off the main isolate.
+  FetchResult _parseFailure(
+      MewsParseException error, String source, List<int>? bytes, String url) {
+    try {
+      final converted = convertHtmlPage(url: url, source: source, bytes: bytes);
+      return FetchParseError(error.message,
+          simplified: converted.page, declaresMews: converted.declaresMews);
+    } on Exception {
+      // Nothing legible in it; the error page stands on its own.
+      return FetchParseError(error.message);
     }
   }
 
@@ -497,7 +523,7 @@ class FetchHandler with Handler<FetchCommand, FetchResult> {
         entries: parseAtomEntries(source: response.body, feedUrl: url),
       );
     } on MewsParseException catch (e) {
-      return FetchParseError(e.message, response.body);
+      return FetchParseError(e.message);
     }
   }
 
@@ -532,7 +558,7 @@ class FetchHandler with Handler<FetchCommand, FetchResult> {
       return const FetchNetworkError();
     }
     if (response.statusCode != 200) return FetchHttpStatusError(response.statusCode);
-    return _parse(response.body, url);
+    return _parse(response.body, url, bytes: response.bodyBytes);
   }
 
   @override

app/lib/application/controllers/browser_controller.dart

@@ -23,11 +23,15 @@ class LoadFailed extends BrowseError {
   const LoadFailed();
 }
 
+/// A page the strict Mews parse refused. [simplified] is the same markup read
+/// as far as the Mews subset reaches, for the reader who asks to see it, and is
+/// null when nothing legible came out. [declaresMews] separates a page with a
+/// real markup error from an ordinary web page.
 class BrokenPage extends BrowseError {
-  const BrokenPage(this.source);
+  const BrokenPage({this.simplified, this.declaresMews = false});
 
-  /// The markup that failed to parse; "show page text" recovers its content.
-  final String source;
+  final MewsPage? simplified;
+  final bool declaresMews;
 }
 
 /// A protocol status with the site's own message for the reader: a temporary
@@ -435,9 +439,9 @@ class BrowserController {
           ));
           await _db.recordVisit(url, page.title);
         }
-      case FetchParseError(:final source):
+      case FetchParseError(:final simplified, :final declaresMews):
         currentPage.value = null;
-        error.value = BrokenPage(source);
+        error.value = BrokenPage(simplified: simplified, declaresMews: declaresMews);
       case FetchProtocolError(:final message):
         currentPage.value = null;
         error.value = ProtocolFailure(message);