improve reverse bang matching

This commit is contained in:
Fabian Freund
2026-06-02 12:02:52 +02:00
parent 630b1dfbe0
commit 6108b24f21
4 changed files with 89 additions and 8 deletions
@@ -26,6 +26,7 @@ import 'package:weblibre/features/bangs/data/models/bang.dart';
import 'package:weblibre/features/bangs/data/models/bang_data.dart';
import 'package:weblibre/features/bangs/data/models/bang_group.dart';
import 'package:weblibre/features/bangs/data/models/bang_key.dart';
import 'package:weblibre/utils/uri_parser.dart';
@DriftAccessor()
class BangDao extends DatabaseAccessor<BangDatabase> with $BangDaoMixin {
@@ -129,14 +130,19 @@ class BangDao extends DatabaseAccessor<BangDatabase> with $BangDaoMixin {
}
Selectable<BangData> getBangDataByTemplateHost(String host) {
final httpsPrefix = 'https://$host/%';
final httpPrefix = 'http://$host/%';
// Search engines redirect results between generic subdomains (e.g.
// bing.com → www.bing.com, www.youtube.com → m.youtube.com), so match the
// template against every equivalent host spelling, not just the live one.
final clauses = hostVariants(host)
.map(
(variant) =>
db.bangDataView.urlTemplate.like('https://$variant/%') |
db.bangDataView.urlTemplate.like('http://$variant/%'),
)
.toList();
return select(db.bangDataView)
..where(
(t) =>
t.urlTemplate.like(httpsPrefix) | t.urlTemplate.like(httpPrefix),
);
..where((_) => clauses.reduce((a, b) => a | b));
}
Selectable<BangData> queryBangs(String searchString) {
@@ -92,7 +92,7 @@ class BangUrlPattern {
/// Extracts the user query from [input] if it matches this pattern. Returns
/// null otherwise.
String? match(Uri input) {
if (input.host.toLowerCase() != host) return null;
if (uri_parser.normalizeHost(input.host) != host) return null;
final mainInput = _componentsFromUri(input);
if (mainInput == null) return null;
@@ -178,7 +178,9 @@ class BangUrlPattern {
if (!parsed.hasScheme || parsed.host.isEmpty) return null;
if (parsed.host.contains(_sentinel)) return null;
final host = parsed.host.toLowerCase();
// Normalize so a template host and a redirected results host that differ
// only by a generic subdomain (www./m.) still match.
final host = uri_parser.normalizeHost(parsed.host);
final fragmentSlot = parsed.fragment.contains(_sentinel);
final templateMain = _componentsFromUri(parsed);
+31
View File
@@ -28,3 +28,34 @@ Uri? tryParseUrl(String? input, {bool eagerParsing = false}) {
allowSchemelessHosts: eagerParsing,
);
}
/// Generic subdomain labels that don't identify a distinct site. Search engines
/// routinely redirect between these (e.g. `bing.com` → `www.bing.com`,
/// `www.youtube.com` → `m.youtube.com`), so they are treated as equivalent when
/// matching a live URL against a bang template host.
const _genericHostPrefixes = {'www', 'm', 'mobile'};
/// Strips a single leading generic subdomain label ([_genericHostPrefixes])
/// from [host] so hosts differing only by such a prefix compare equal.
///
/// `www.bing.com` and `bing.com` → `bing.com`; `m.youtube.com` → `youtube.com`.
/// Non-generic subdomains (e.g. `cn.bing.com`) are left untouched.
String normalizeHost(String host) {
final lower = host.toLowerCase();
final dot = lower.indexOf('.');
if (dot <= 0) return lower;
final label = lower.substring(0, dot);
if (_genericHostPrefixes.contains(label)) {
return lower.substring(dot + 1);
}
return lower;
}
/// All host spellings that should be considered the same site as [host] for the
/// purposes of bang template lookup: the normalized base plus each generic
/// subdomain variant.
Set<String> hostVariants(String host) {
final base = normalizeHost(host);
return {base, for (final prefix in _genericHostPrefixes) '$prefix.$base'};
}