feat: chatbot trip search finds trips by trip name, hotel name and code

src/Chatbot/TripCatalog.php

@@ -5,7 +5,9 @@ declare(strict_types=1);
 namespace App\Chatbot;
 
 use App\BusProNet\Utility\DateCodeUtility;
+use App\BusProNet\XmlLoader\HotelLoader;
 use App\Chatbot\Search\ContentSearchQuery;
+use App\Chatbot\Search\NameQuery;
 use App\Entity\ContentDocument;
 use App\Enum\HotelType;
 use App\Enum\RoomType;
@@ -24,9 +26,11 @@ use Symfony\Component\Clock\ClockInterface;
  * The join is on the product code; several TYPO3 products may share one, and each is its
  * own trip. Hotels are joined on any of the TYPO3 hotel's `codes`, because BusPro uses the
  * matchcode aliases; a code carried by more than one hotel resolves to the one whose own
- * `code` it is.
+ * `code` it is. BusPro's own hotel names (hotel.xml) are internal wording, so they are
+ * only searched, and shown just where TYPO3 has no hotel for a code.
  *
- * search() mirrors the website's travel search, get() details one trip. A room is bookable
+ * search() mirrors the website's travel search, get() details one trip. Free text is read
+ * as a code or a name before it is looked up in the page texts, see resolveQuery(). A room is bookable
  * unless its status is `Buchungsstop`; places and prices are derived per date x hotel by
  * TravelDateFacts. Room counts never leave this class: the website does not show them, so
  * the tools say only whether a date is freely bookable or on request.
@@ -47,6 +51,7 @@ class TripCatalog
     public function __construct(
         private readonly ProductExportBuilder $exportBuilder,
         private readonly ContentDocumentRepository $repository,
+        private readonly HotelLoader $hotelLoader,
         private readonly ClockInterface $clock,
         private readonly LoggerInterface $logger,
     ) {
@@ -63,6 +68,7 @@ class TripCatalog
     {
         $catalog = $this->load();
         $today = $this->today();
+        $named = null === $criteria->query ? null : $this->resolveQuery($catalog, $criteria->query);
         $results = [];
 
         foreach ($this->candidateProducts($catalog, $criteria) as $product) {
@@ -75,6 +81,15 @@ class TripCatalog
 
                 $hotel = $catalog['hotelsByCode'][$this->normalize($row['hotel']['code'] ?? null)][0] ?? null;
                 $hotelData = $hotel?->getData() ?? [];
+
+                if (null !== $named
+                    && !isset($named['products'][$product->getUid()])
+                    && !isset($named['hotels'][$hotel?->getUid() ?? 0])
+                    && !isset($named['busProHotels'][$row['hotel']['busProId'] ?? 0])
+                    && (null === $named['dateCode'] || $named['dateCode'] !== $this->normalize($row['date']['code'] ?? null))) {
+                    continue;
+                }
+
                 $facts = TravelDateFacts::fromExportRow($row);
 
                 if (!$this->matches($row, $facts, HotelType::tryFrom((int) ($hotelData['type'] ?? 0)), $criteria)) {
@@ -87,7 +102,8 @@ class TripCatalog
                     'product' => $header['name'],
                     'code' => $header['code'],
                     'dateCode' => $row['date']['code'],
-                    'hotel' => $hotel?->getTitle() ?? $row['hotel']['code'],
+                    'hotel' => $this->hotelName($catalog, $hotel, $row),
+                    'hotelCode' => $row['hotel']['code'],
                     'accommodation' => HotelType::tryFrom((int) ($hotelData['type'] ?? 0))?->label(),
                     'destination' => $header['destination'],
                     'concept' => $header['concept'],
@@ -228,6 +244,7 @@ class TripCatalog
      *     productsByCode: array<string, list<ContentDocument>>,
      *     hotelsByCode: array<string, list<ContentDocument>>,
      *     rowsByCode: array<string, list<array<string, mixed>>>,
+     *     busProHotels: array<int, array{code: string, name: ?string}>,
      * }
      */
     private function load(): array
@@ -264,11 +281,21 @@ class TripCatalog
         }
 
         $rowsByCode = [];
+        $busProHotels = [];
+        $masterData = $this->hotelLoader->loadAll();
+
         foreach ($export->rows as $row) {
             $code = $this->normalize($row['product']['code'] ?? null);
             if ('' !== $code) {
                 $rowsByCode[$code][] = $row;
             }
+
+            $hotelId = (int) ($row['hotel']['busProId'] ?? 0);
+            if (0 !== $hotelId && !isset($busProHotels[$hotelId])) {
+                // BusPro breaks names over lines ("Vacanceole\nRésidence L'Edelweiss").
+                $name = trim((string) preg_replace('/\s+/u', ' ', isset($masterData[$hotelId]) ? (string) $masterData[$hotelId]->name : ''));
+                $busProHotels[$hotelId] = ['code' => $this->normalize($row['hotel']['code'] ?? null), 'name' => '' === $name ? null : $name];
+            }
         }
 
         return [
@@ -277,12 +304,14 @@ class TripCatalog
             'productsByCode' => $productsByCode,
             'hotelsByCode' => $hotelsByCode,
             'rowsByCode' => $rowsByCode,
+            'busProHotels' => $busProHotels,
         ];
     }
 
     /**
      * Published products narrowed by everything the content index can resolve: site, day
-     * trip, concept, destination and free text.
+     * trip, concept and destination. Free text is applied per travel date, since it may
+     * name one hotel of a trip.
      *
      * @param array<string, mixed> $catalog
      *
@@ -297,7 +326,6 @@ class TripCatalog
         $countryUids = null === $criteria->country ? null : $this->matchingUids($records, ['country'], $criteria->country, matchCode: true);
         $regionUids = null === $criteria->region ? null : $this->matchingUids($records, ['region'], $criteria->region);
         $cityUids = null === $criteria->region ? null : $this->matchingUids($records, ['city'], $criteria->region);
-        $queryUids = null === $criteria->query ? null : $this->productUidsMatchingText($records, $criteria->query);
 
         $products = [];
 
@@ -315,8 +343,7 @@ class TripCatalog
                 || (null !== $criteria->site && $record->getSite() !== $criteria->site)
                 || (null !== $criteria->daytrip && (bool) ($data['daytrip'] ?? false) !== $criteria->daytrip)
                 || (null !== $conceptUids && !in_array($data['conceptUid'] ?? null, $conceptUids, true))
-                || (null !== $countryUids && !in_array($data['countryUid'] ?? null, $countryUids, true))
-                || (null !== $queryUids && !in_array($record->getUid(), $queryUids, true))) {
+                || (null !== $countryUids && !in_array($data['countryUid'] ?? null, $countryUids, true))) {
                 continue;
             }
 
@@ -366,27 +393,155 @@ class TripCatalog
     }
 
     /**
-     * Products reached by a full-text match on themselves or on a hotel, destination or
-     * concept they belong to.
+     * What a free-text query asks for: whole trips, single hotels of trips, or one travel
+     * date. Null when it restricts nothing.
      *
-     * @param array<string, ContentDocument> $records
+     * A code or a name is taken at its word, so naming a trip or hotel returns that one and
+     * not every page that mentions it. In order, stopping at the first that finds anything:
+     * a product, hotel or date code; names holding every word of the query; names the query
+     * spells out among other words; and only then the page texts.
      *
-     * @return list<int>
+     * @param array<string, mixed> $catalog
+     *
+     * Hotels are known twice: by their TYPO3 record and by BusPro's own name and code, which
+     * also covers a hotel TYPO3 has no record for
+     *
+     * @return array{products: array<int, true>, hotels: array<int, true>, busProHotels: array<int, true>, dateCode: ?string}|null
      */
-    private function productUidsMatchingText(array $records, string $text): array
+    private function resolveQuery(array $catalog, string $text): ?array
     {
-        $booleanQuery = ContentSearchQuery::booleanQuery($text);
+        /** @var array<string, ContentDocument> $records */
+        $records = $catalog['records'];
 
-        if (null === $booleanQuery) {
-            return array_values(array_map(static fn (ContentDocument $record): int => $record->getUid(), array_filter($records, static fn (ContentDocument $record): bool => 'product' === $record->getType())));
+        $matched = $this->recordsWithCode($catalog, $text) ?? $this->recordsNamed($catalog, $text);
+
+        if (null !== $matched) {
+            return $this->tripsOf($records, $matched);
         }
 
+        $dateCode = (new DateCodeUtility())->sanitize(trim($text));
+
+        foreach ($catalog['export']->rows as $row) {
+            if ($this->normalize($row['date']['code'] ?? null) === $dateCode) {
+                return ['products' => [], 'hotels' => [], 'busProHotels' => [], 'dateCode' => $dateCode];
+            }
+        }
+
+        $terms = ContentSearchQuery::terms($text);
+
+        if ([] === $terms) {
+            return null;
+        }
+
+        // Every term first: "Sauna Whirlpool" means both. A question rarely uses the page's
+        // own words for everything, so any term is the fallback.
+        foreach (count($terms) > 1 ? [true, false] : [false] as $allTerms) {
+            $matched = [];
+            foreach ($this->repository->matchRecords((string) ContentSearchQuery::booleanQuery($text, $allTerms), self::RECORD_TYPES, self::LANGUAGE) as $hit) {
+                $matched[$hit['type']][$hit['uid']] = true;
+            }
+
+            if ([] !== $matched) {
+                break;
+            }
+        }
+
+        return $this->tripsOf($records, $matched);
+    }
+
+    /**
+     * The products and the hotel carrying the text as their code, hotel aliases included.
+     * Matches are keyed by record type, BusPro's hotels as `busProHotel` by their id.
+     *
+     * @param array<string, mixed> $catalog
+     *
+     * @return array<string, array<int, true>>|null
+     */
+    private function recordsWithCode(array $catalog, string $text): ?array
+    {
+        $code = (new DateCodeUtility())->sanitize(trim($text));
         $matched = [];
-        foreach ($this->repository->matchRecords($booleanQuery, self::RECORD_TYPES, self::LANGUAGE) as $hit) {
-            $matched[$hit['type']][$hit['uid']] = true;
+
+        foreach ($catalog['productsByCode'][$code] ?? [] as $product) {
+            $matched['product'][$product->getUid()] = true;
         }
 
-        $uids = [];
+        if (isset($catalog['hotelsByCode'][$code][0])) {
+            $matched['hotel'][$catalog['hotelsByCode'][$code][0]->getUid()] = true;
+        }
+
+        foreach ($catalog['busProHotels'] as $hotelId => $busProHotel) {
+            if ($busProHotel['code'] === $code) {
+                $matched['busProHotel'][$hotelId] = true;
+            }
+        }
+
+        return [] === $matched ? null : $matched;
+    }
+
+    /**
+     * The records the text names. A name holding every word of the query wins: "Lederer"
+     * is "Sportclub Lederer". Otherwise the query may spell a name out among other words,
+     * and there the most specific kind wins, so that "Sportclub Lederer Saalbach" is the
+     * hotel and not everything in Saalbach.
+     *
+     * @param array<string, mixed> $catalog
+     *
+     * @return array<string, array<int, true>>|null
+     */
+    private function recordsNamed(array $catalog, string $text): ?array
+    {
+        $query = NameQuery::tokens($text);
+        $names = [];
+
+        foreach ($catalog['records'] as $record) {
+            $names[] = [$record->getType(), $record->getUid(), $record->getTitle()];
+        }
+
+        foreach ($catalog['busProHotels'] as $hotelId => $busProHotel) {
+            $names[] = ['busProHotel', $hotelId, $busProHotel['name'] ?? ''];
+        }
+
+        $within = [];
+        $contained = [];
+
+        foreach ($names as [$type, $id, $title]) {
+            $name = NameQuery::tokens($title);
+
+            if (NameQuery::within($query, $name)) {
+                $within[$type][$id] = true;
+            } elseif (NameQuery::contains($query, $name)) {
+                $contained[$type][$id] = true;
+            }
+        }
+
+        if ([] !== $within) {
+            return $within;
+        }
+
+        foreach ([['product', 'hotel', 'busProHotel'], ['city'], ['region'], ['country', 'concept']] as $types) {
+            $matched = array_intersect_key($contained, array_flip($types));
+
+            if ([] !== $matched) {
+                return $matched;
+            }
+        }
+
+        return null;
+    }
+
+    /**
+     * Matched records as trips: a product, destination or concept stands for the whole
+     * trips, a hotel only for itself, since a trip may have many.
+     *
+     * @param array<string, ContentDocument>  $records
+     * @param array<string, array<int, true>> $matched by type and uid
+     *
+     * @return array{products: array<int, true>, hotels: array<int, true>, busProHotels: array<int, true>, dateCode: null}
+     */
+    private function tripsOf(array $records, array $matched): array
+    {
+        $products = [];
 
         foreach ($records as $record) {
             if ('product' !== $record->getType()) {
@@ -394,33 +549,18 @@ class TripCatalog
             }
 
             $data = $record->getData() ?? [];
-            $hotelUids = is_array($data['hotelUids'] ?? null) ? $data['hotelUids'] : [];
             $regionUids = array_merge([$data['regionUid'] ?? null], is_array($data['additionalRegionUids'] ?? null) ? $data['additionalRegionUids'] : []);
 
-            $hit = isset($matched['product'][$record->getUid()])
-                || [] !== array_intersect($hotelUids, array_keys($matched['hotel'] ?? []))
+            if (isset($matched['product'][$record->getUid()])
                 || [] !== array_intersect($regionUids, array_keys($matched['region'] ?? []))
                 || isset($matched['country'][$data['countryUid'] ?? -1])
                 || isset($matched['city'][$data['cityUid'] ?? -1])
-                || isset($matched['concept'][$data['conceptUid'] ?? -1]);
-
-            // Hotels also name their products, for products whose hotelUids miss them.
-            if (!$hit) {
-                foreach (array_keys($matched['hotel'] ?? []) as $hotelUid) {
-                    $hotelProducts = ($records['hotel:'.$hotelUid] ?? null)?->getData()['productUids'] ?? [];
-                    if (is_array($hotelProducts) && in_array($record->getUid(), $hotelProducts, true)) {
-                        $hit = true;
-                        break;
-                    }
-                }
-            }
-
-            if ($hit) {
-                $uids[] = $record->getUid();
+                || isset($matched['concept'][$data['conceptUid'] ?? -1])) {
+                $products[$record->getUid()] = true;
             }
         }
 
-        return $uids;
+        return ['products' => $products, 'hotels' => $matched['hotel'] ?? [], 'busProHotels' => $matched['busProHotel'] ?? [], 'dateCode' => null];
     }
 
     /**
@@ -509,8 +649,9 @@ class TripCatalog
             }
         }
 
-        // A trip without upcoming dates still describes its hotels.
-        if ([] === $hotels) {
+        // A trip without upcoming dates still describes its hotels. One whose dates are at
+        // a hotel the website does not know has none to describe.
+        if ([] === $hotelCodes) {
             foreach ((array) ($product->getData()['hotelUids'] ?? []) as $uid) {
                 $hotel = is_int($uid) ? ($catalog['records']['hotel:'.$uid] ?? null) : null;
                 if (null !== $hotel) {
@@ -568,7 +709,7 @@ class TripCatalog
             ];
 
             $dates[$dateId]['hotels'][] = [
-                'hotel' => $hotel?->getTitle() ?? $row['hotel']['code'],
+                'hotel' => $this->hotelName($catalog, $hotel, $row),
                 'hotelCode' => $row['hotel']['code'],
                 'rooms' => $rooms,
                 'bookable' => [] !== $rooms,
@@ -644,6 +785,19 @@ class TripCatalog
         return array_values(array_filter(array_unique(array_map($this->normalize(...), $codes))));
     }
 
+    /**
+     * The website's name for the hotel of a travel date, else BusPro's, else its code.
+     *
+     * @param array<string, mixed> $catalog
+     * @param array<string, mixed> $row
+     */
+    private function hotelName(array $catalog, ?ContentDocument $hotel, array $row): string
+    {
+        return $hotel?->getTitle()
+            ?? $catalog['busProHotels'][$row['hotel']['busProId'] ?? 0]['name']
+            ?? (string) $row['hotel']['code'];
+    }
+
     private function dataAsOf(Export $export): string
     {
         return ($export->sourceGeneratedAt ?? $export->generatedAt)->format('c');