Improve approved museum URL normalizing

lib/Museum.php

@@ -22,10 +22,13 @@ class Museum extends PropertiesBase{
 			throw new Exceptions\InvalidUrlException($url);
 		}
 
-		if(stripos($parsedUrl['host'], 'rijksmuseum.nl') !== false){
+		// We can't match on TLD because extracting the TLD for double-barrel TLDs, like .gov.uk, requires a whitelist.
+
+		if(preg_match('/\brijksmuseum\.nl$/ius', $parsedUrl['host'])){
 			// Rijksmuseum has several URL variants we want to resolve
 			// https://www.rijksmuseum.nl/en/search/objects?q=hals&p=6&ps=12&st=Objects&ii=2#/SK-A-1246,59
 			// https://www.rijksmuseum.nl/nl/collectie/SK-A-1246
+			// https://www.rijksmuseum.nl/en/rijksstudio/artists/jean-baptiste-vanmour/objects#/SK-A-1998,8
 
 			$exampleUrl = 'https://www.rijksmuseum.nl/en/collection/SK-A-1246';
 
@@ -33,8 +36,23 @@ class Museum extends PropertiesBase{
 				throw new Exceptions\InvalidMuseumUrlException($url, $exampleUrl);
 			}
 
-			if(preg_match('|^/en/search/objects$|ius', $parsedUrl['path']) && isset($parsedUrl['fragment'])){
+			if(	(
-				$outputUrl = 'https://' . $parsedUrl['host'] . '/en/collection/' . preg_replace('/,+$/ius', '', $parsedUrl['fragment']);
+					preg_match('|^/en/search/objects$|ius', $parsedUrl['path'])
+					||
+					preg_match('|^/en/rijksstudio/[^/]+/[^/]+/objects$|ius', $parsedUrl['path'])
+				)
+				&&
+				isset($parsedUrl['fragment'])
+			){
+				$id = preg_replace('/,.+$/ius', '', $parsedUrl['fragment']);
+				$id = preg_replace('|^/|ius', '', $id);
+				$outputUrl = 'https://' . $parsedUrl['host'] . '/en/collection/' . $id;
+			}
+
+			if(preg_match('|^en/rijksstudio/[^/]+/[^/]+/objects$|ius', $parsedUrl['path']) && isset($parsedUrl['fragment'])){
+				$id = preg_replace('/,.+$/ius', '', $parsedUrl['fragment']);
+				$id = preg_replace('|^/|ius', '', $id);
+				$outputUrl = 'https://' . $parsedUrl['host'] . '/en/collection/' . $id;
 			}
 
 			if(preg_match('|^/nl/collectie/[^/]+$|ius', $parsedUrl['path'])){
@@ -51,24 +69,23 @@ class Museum extends PropertiesBase{
 
 			return $outputUrl;
 		}
-
+		elseif(preg_match('/\bmetmuseum\.org$/ius', $parsedUrl['host'])){
-		if(stripos($parsedUrl['host'], 'metmuseum.org') !== false){
 			$exampleUrl = 'https://www.metmuseum.org/art/collection/search/13180';
 
-			if($parsedUrl['host'] != 'www.metmuseum.org'){
+			$host = preg_replace('|^metmuseum.org|ius', 'www.metmuseum.org', $parsedUrl['host']);
+			if($host != 'www.metmuseum.org'){
 				throw new Exceptions\InvalidMuseumUrlException($url, $exampleUrl);
 			}
 
-			if(!preg_match('|^/art/collection/search/\d+$|ius', $parsedUrl['path'])){
+			if(!preg_match('|^/art/collection/search/[\d]+$|ius', $parsedUrl['path'])){
 				throw new Exceptions\InvalidMuseumUrlException($url, $exampleUrl);
 			}
 
-			$outputUrl = 'https://' . $parsedUrl['host'] . $parsedUrl['path'];
+			$outputUrl = 'https://' . $host . $parsedUrl['path'];
 
 			return $outputUrl;
 		}
-
+		elseif(preg_match('/\bnationalmuseum\.se$/ius', $parsedUrl['host'])){
-		if(stripos($parsedUrl['host'], 'nationalmuseum.se') !== false){
 			$exampleUrl = 'https://collection.nationalmuseum.se/eMP/eMuseumPlus?objectId=18217';
 
 			if($parsedUrl['host'] != 'collection.nationalmuseum.se'){
@@ -93,8 +110,7 @@ class Museum extends PropertiesBase{
 
 			return $outputUrl;
 		}
-
+		elseif(preg_match('/\bartsmia\.org$/ius', $parsedUrl['host'])){
-		if(stripos($parsedUrl['host'], 'artsmia.org') !== false){
 			$exampleUrl = 'https://collections.artsmia.org/art/3729/castle-and-watermill-by-a-river-jacob-van-ruisdael';
 
 			if($parsedUrl['host'] != 'collections.artsmia.org'){
@@ -109,8 +125,7 @@ class Museum extends PropertiesBase{
 
 			return $outputUrl;
 		}
-
+		elseif(preg_match('/\bthewalters\.org$/ius', $parsedUrl['host'])){
-		if(stripos($parsedUrl['host'], 'thewalters.org') !== false){
 			$exampleUrl = 'https://art.thewalters.org/detail/4695/boston-street-scene-boston-common/';
 
 			if($parsedUrl['host'] != 'art.thewalters.org'){
@@ -125,8 +140,7 @@ class Museum extends PropertiesBase{
 
 			return $outputUrl;
 		}
-
+		elseif(preg_match('/\bartic\.edu$/ius', $parsedUrl['host'])){
-		if(stripos($parsedUrl['host'], 'artic.edu') !== false){
 			$exampleUrl = 'https://www.artic.edu/artworks/133864/the-defense-of-paris';
 
 			if($parsedUrl['host'] != 'www.artic.edu'){
@@ -141,8 +155,7 @@ class Museum extends PropertiesBase{
 
 			return $outputUrl;
 		}
-
+		elseif(preg_match('/\bclevelandart\.org$/ius', $parsedUrl['host'])){
-		if(stripos($parsedUrl['host'], 'clevelandart.org') !== false){
 			$exampleUrl = 'https://www.clevelandart.org/art/1969.54';
 
 			if($parsedUrl['host'] != 'www.clevelandart.org'){
@@ -157,13 +170,13 @@ class Museum extends PropertiesBase{
 
 			return $outputUrl;
 		}
-
+		elseif(preg_match('/\bparis\.fr$/ius', $parsedUrl['host'])){
-		if(stripos($parsedUrl['host'], 'paris.fr') !== false){
 			// Variant: https://www.parismuseescollections.paris.fr/en/musee-carnavalet/oeuvres/portrait-del-singer-simon-chenard-1758-1832-wearing-the-clothes-of-a-sans
+			// May also be missing www.
 
 			$exampleUrl = 'https://www.parismuseescollections.paris.fr/en/node/226154';
 
-			if($parsedUrl['host'] != 'www.parismuseescollections.paris.fr'){
+			if(!preg_match('|^(www\.)?parismuseescollections\.paris\.fr$|ius', $parsedUrl['host'])){
 				throw new Exceptions\InvalidMuseumUrlException($url, $exampleUrl);
 			}
 
@@ -171,34 +184,27 @@ class Museum extends PropertiesBase{
 				throw new Exceptions\InvalidMuseumUrlException($url, $exampleUrl);
 			}
 
-			$outputUrl = 'https://' . $parsedUrl['host'] . $parsedUrl['path'];
+			$outputUrl = 'https://' . preg_replace('|^(www\.)?parismuseescollections\.paris\.fr$|ius', 'www.parismuseescollections.paris.fr', $parsedUrl['host']) . $parsedUrl['path'];
 
 			return $outputUrl;
 		}
-
+		elseif(preg_match('/\bwww\.si\.edu$/ius', $parsedUrl['host'])){
-		if(stripos($parsedUrl['host'], 'www.si.edu') !== false){
 			// URLs can look like <https://www.si.edu/object/cave-scene:saam_XX98H> in which the text before : is for SEO and can be cut
 			$exampleUrl = 'https://www.si.edu/object/saam_1983.95.90';
 
-			if(!preg_match('|/object/[^/]+?:[^/:]+$|ius', $parsedUrl['path'])){
+			if(!preg_match('|^/object/[^/]+?(:[^/:]+)?$|ius', $parsedUrl['path'])){
 				throw new Exceptions\InvalidMuseumUrlException($url, $exampleUrl);
 			}
 
-			$path = preg_replace('^|/object/[^/]+:([^/:]+)$|ius', '/object/\1', $parsedUrl['path']);
+			$path = preg_replace('|^/object/[^/]+:([^/:]+)$|ius', '/object/\1', $parsedUrl['path']);
 
 			$outputUrl = 'https://' . $parsedUrl['host'] . $path;
 
 			return $outputUrl;
 		}
-
+		elseif(preg_match('/\bamericanart\.si\.edu$/ius', $parsedUrl['host'])){
-		if(stripos($parsedUrl['host'], 'americanart.si.edu') !== false){
 			$exampleUrl = 'https://americanart.si.edu/artwork/study-apotheosis-washington-rotunda-united-states-capitol-building-84517';
 
-			$path = $parsedUrl['path'];
-			if(!preg_match('|/object/[^/]+?:[^/:]+$|ius', $path)){
-				throw new Exceptions\InvalidMuseumUrlException($url, $exampleUrl);
-			}
-
 			if(!preg_match('|^/artwork/[^/]+$|ius', $parsedUrl['path'])){
 				throw new Exceptions\InvalidMuseumUrlException($url, $exampleUrl);
 			}
@@ -207,8 +213,7 @@ class Museum extends PropertiesBase{
 
 			return $outputUrl;
 		}
-
+		elseif(preg_match('/\bcollections\.si\.edu$/ius', $parsedUrl['host'])){
-		if(stripos($parsedUrl['host'], 'collections.si.edu') !== false){
 			// These URLs can actually be normalized to a www.si.edu URL by pulling out the object ID
 			$exampleUrl = 'https://collections.si.edu/search/detail/edanmdm:saam_1981.146.1';
 
@@ -223,8 +228,7 @@ class Museum extends PropertiesBase{
 
 			return $outputUrl;
 		}
-
+		elseif(preg_match('/\bnpg\.si\.edu$/ius', $parsedUrl['host'])){
-		if(stripos($parsedUrl['host'], 'npg.si.edu') !== false){
 			// These URLs can actually be normalized to a www.si.edu URL by pulling out the object ID
 			$exampleUrl = 'https://npg.si.edu/object/npg_NPG.2008.5';
 
@@ -237,8 +241,7 @@ class Museum extends PropertiesBase{
 
 			return $outputUrl;
 		}
-
+		elseif(preg_match('/\bbirminghammuseums\.org\.uk$/ius', $parsedUrl['host'])){
-		if(stripos($parsedUrl['host'], 'birminghammuseums.org.uk') !== false){
 			$exampleUrl = 'https://dams.birminghammuseums.org.uk/asset-bank/action/viewAsset?id=6726';
 
 			if($parsedUrl['host'] != 'dams.birminghammuseums.org.uk'){
@@ -259,56 +262,60 @@ class Museum extends PropertiesBase{
 
 			return $outputUrl;
 		}
-
+		elseif(preg_match('/\bmnk\.pl$/ius', $parsedUrl['host'])){
-		if(stripos($parsedUrl['host'], 'mnk.pl') !== false){
+			$exampleUrl = 'https://zbiory.mnk.pl/en/catalog/333584';
-			$exampleUrl = 'https://zbiory.mnk.pl/en/search-result/catalog/333584';
 
 			if($parsedUrl['host'] != 'zbiory.mnk.pl'){
 				throw new Exceptions\InvalidMuseumUrlException($url, $exampleUrl);
 			}
 
-			if(!preg_match('|^/en/search-result/catalog/[^/]+$|ius', $parsedUrl['path'])){
+			// Somtimes the path may have '/search-result/advance' or '/search-result' in it, cut that here
+			$path = preg_replace('~^/en(/search-result/advance|/search-result)?~ius', '/en', $parsedUrl['path']);
+
+			if(!preg_match('|^/en/catalog/[^/]+$|ius', $path)){
 				throw new Exceptions\InvalidMuseumUrlException($url, $exampleUrl);
 			}
 
-			$outputUrl = 'https://' . $parsedUrl['host'] . $parsedUrl['path'];
+			$outputUrl = 'https://' . $parsedUrl['host'] . $path;
 
 			return $outputUrl;
 		}
-
+		elseif(preg_match('/\bsmk\.dk$/ius', $parsedUrl['host'])){
-		if(stripos($parsedUrl['host'], 'smk.dk') !== false){
 			$exampleUrl = 'https://open.smk.dk/artwork/image/KMS1884';
 
 			if($parsedUrl['host'] != 'open.smk.dk'){
 				throw new Exceptions\InvalidMuseumUrlException($url, $exampleUrl);
 			}
 
-			if(!preg_match('|^/artwork/image/[^/]+$|ius', $parsedUrl['path'])){
+			// Somtimes the path may have 'en' in it, cut that here
+			$path = preg_replace('|^/en/|ius', '/', $parsedUrl['path']);
+
+			if(!preg_match('|^/artwork/image/[^/]+$|ius', $path)){
 				throw new Exceptions\InvalidMuseumUrlException($url, $exampleUrl);
 			}
 
-			$outputUrl = 'https://' . $parsedUrl['host'] . $parsedUrl['path'];
+			$outputUrl = 'https://' . $parsedUrl['host'] . $path;
 
 			return $outputUrl;
 		}
-
+		elseif(preg_match('/\bkansallisgalleria\.fi$/ius', $parsedUrl['host'])){
-		if(stripos($parsedUrl['host'], 'kansallisgalleria.fi') !== false){
 			$exampleUrl = 'https://www.kansallisgalleria.fi/en/object/429609';
 
 			if($parsedUrl['host'] != 'www.kansallisgalleria.fi'){
 				throw new Exceptions\InvalidMuseumUrlException($url, $exampleUrl);
 			}
 
-			if(!preg_match('|^/en/object/[^/]+$|ius', $parsedUrl['path'])){
+			$path = preg_replace('|^/fi/|ius', '/en/', $parsedUrl['path']);
+
+			if(!preg_match('|^/en/object/[^/]+$|ius', $path)){
 				throw new Exceptions\InvalidMuseumUrlException($url, $exampleUrl);
 			}
 
-			$outputUrl = 'https://' . $parsedUrl['host'] . $parsedUrl['path'];
+			$outputUrl = 'https://' . $parsedUrl['host'] . $path;
 
 			return $outputUrl;
 		}
-
+		elseif(preg_match('/\bnga\.gov$/ius', $parsedUrl['host'])){
-		if(stripos($parsedUrl['host'], 'nga.gov') !== false){
 			$exampleUrl = 'https://www.nga.gov/collection/art-object-page.46522.html';
 
 			if($parsedUrl['host'] != 'www.nga.gov'){
@@ -323,8 +330,7 @@ class Museum extends PropertiesBase{
 
 			return $outputUrl;
 		}
-
+		elseif(preg_match('/\bnivaagaard\.dk$/ius', $parsedUrl['host'])){
-		if(stripos($parsedUrl['host'], 'nivaagaard.dk') !== false){
 			$exampleUrl = 'https://www.nivaagaard.dk/en/vare/lundstroem-vilhelm/';
 
 			if($parsedUrl['host'] != 'www.nivaagaard.dk'){
@@ -339,8 +345,7 @@ class Museum extends PropertiesBase{
 
 			return $outputUrl;
 		}
-
+		elseif(preg_match('/\brisdmuseum\.org$/ius', $parsedUrl['host'])){
-		if(stripos($parsedUrl['host'], 'risdmuseum.org') !== false){
 			$exampleUrl = 'https://risdmuseum.org/art-design/collection/portrait-christiana-carteaux-bannister-2016381';
 
 			if($parsedUrl['host'] != 'risdmuseum.org'){
@@ -355,8 +360,7 @@ class Museum extends PropertiesBase{
 
 			return $outputUrl;
 		}
-
+		elseif(preg_match('/\baberdeencity\.gov\.uk$/ius', $parsedUrl['host'])){
-		if(stripos($parsedUrl['host'], 'aberdeencity.gov.uk') !== false){
 			// All we need is the int object ID, the last slug is SEO
 			$exampleUrl = 'https://emuseum.aberdeencity.gov.uk/objects/3215/james-cromar-watt-lld';
 
@@ -374,8 +378,7 @@ class Museum extends PropertiesBase{
 
 			return $outputUrl;
 		}
-
+		elseif(preg_match('/\bbrightonmuseums\.org\.uk$/ius', $parsedUrl['host'])){
-		if(stripos($parsedUrl['host'], 'brightonmuseums.org.uk') !== false){
 			// All we need is the int object ID, the last slug is SEO
 			$exampleUrl = 'https://collections.brightonmuseums.org.uk/records/63caa90083d50a00184b8e90';
 
@@ -391,8 +394,7 @@ class Museum extends PropertiesBase{
 
 			return $outputUrl;
 		}
-
+		elseif(preg_match('/\bgrpmcollections\.org$/ius', $parsedUrl['host'])){
-		if(stripos($parsedUrl['host'], 'grpmcollections.org') !== false){
 			// All we need is the int object ID, the last slug is SEO
 			$exampleUrl = 'https://www.grpmcollections.org/Detail/objects/130684';
 
@@ -408,8 +410,7 @@ class Museum extends PropertiesBase{
 
 			return $outputUrl;
 		}
-
+		elseif(preg_match('/\bthorvaldsensmuseum\.dk$/ius', $parsedUrl['host'])){
-		if(stripos($parsedUrl['host'], 'thorvaldsensmuseum.dk') !== false){
 			// All we need is the int object ID, the last slug is SEO
 			$exampleUrl = 'https://kataloget.thorvaldsensmuseum.dk/en/B122';
 
@@ -425,8 +426,7 @@ class Museum extends PropertiesBase{
 
 			return $outputUrl;
 		}
-
+		elseif(preg_match('/\bmuseabrugge\.be$/ius', $parsedUrl['host'])){
-		if(stripos($parsedUrl['host'], 'museabrugge.be') !== false){
 			// All we need is the int object ID, the last slug is SEO
 			$exampleUrl = 'https://collectie.museabrugge.be/en/collection/work/id/2013_GRO0013_I';
 
@@ -442,8 +442,7 @@ class Museum extends PropertiesBase{
 
 			return $outputUrl;
 		}
-
+		elseif(preg_match('/\bbritishart\.yale\.edu$/ius', $parsedUrl['host'])){
-		if(stripos($parsedUrl['host'], 'britishart.yale.edu') !== false){
 			// All we need is the int object ID, the last slug is SEO
 			$exampleUrl = 'https://collections.britishart.yale.edu/catalog/tms:1010';
 
@@ -459,8 +458,7 @@ class Museum extends PropertiesBase{
 
 			return $outputUrl;
 		}
-
+		elseif(preg_match('/\bkunsthalle-karlsruhe\.de$/ius', $parsedUrl['host'])){
-		if(stripos($parsedUrl['host'], 'kunsthalle-karlsruhe.de') !== false){
 			// All we need is the int object ID, the last slug is SEO
 			$exampleUrl = 'https://www.kunsthalle-karlsruhe.de/kunstwerke/Ferdinand-Keller/K%C3%BCstenlandschaft-bei-Rio-de-Janeiro/C066F030484D7D09148891B0E70524B8/';
 
@@ -476,8 +474,7 @@ class Museum extends PropertiesBase{
 
 			return $outputUrl;
 		}
-
+		elseif(preg_match('/\bgetty\.edu$/ius', $parsedUrl['host'])){
-		if(stripos($parsedUrl['host'], 'getty.edu') !== false){
 			// All we need is the int object ID, the last slug is SEO
 			$exampleUrl = 'https://www.getty.edu/art/collection/object/103RG0';
 
@@ -493,8 +490,7 @@ class Museum extends PropertiesBase{
 
 			return $outputUrl;
 		}
-
+		elseif(preg_match('/\bartgallery\.yale\.edu$/ius', $parsedUrl['host'])){
-		if(stripos($parsedUrl['host'], 'artgallery.yale.edu') !== false){
 			// All we need is the int object ID, the last slug is SEO
 			$exampleUrl = 'https://artgallery.yale.edu/collections/objects/44306';
 
@@ -510,6 +506,52 @@ class Museum extends PropertiesBase{
 
 			return $outputUrl;
 		}
+		elseif(preg_match('/\bwebumenia\.sk$/ius', $parsedUrl['host'])){
+			// All we need is the int object ID, the last slug is SEO
+			$exampleUrl = 'https://www.webumenia.sk/en/dielo/SVK:SNG.O_85';
+
+			if($parsedUrl['host'] != 'www.webumenia.sk'){
+				throw new Exceptions\InvalidMuseumUrlException($url, $exampleUrl);
+			}
+
+			if(!preg_match('|^/en/dielo/[^/]+?$|ius', $parsedUrl['path'])){
+				throw new Exceptions\InvalidMuseumUrlException($url, $exampleUrl);
+			}
+
+			$outputUrl = 'https://' . $parsedUrl['host'] . $parsedUrl['path'];
+
+			return $outputUrl;
+		}
+		elseif(preg_match('/\bdigitaltmuseum\.no$/ius', $parsedUrl['host'])){
+			$exampleUrl = 'https://digitaltmuseum.no/021048495118/fra-saxegardsgaten-maleri';
+
+			if($parsedUrl['host'] != 'digitaltmuseum.no'){
+				throw new Exceptions\InvalidMuseumUrlException($url, $exampleUrl);
+			}
+
+			if(!preg_match('|^/[^/]+?/[^/]+?$|ius', $parsedUrl['path'])){
+				throw new Exceptions\InvalidMuseumUrlException($url, $exampleUrl);
+			}
+
+			$outputUrl = 'https://' . $parsedUrl['host'] . $parsedUrl['path'];
+
+			return $outputUrl;
+		}
+		elseif(preg_match('/\bart\.pl$/ius', $parsedUrl['host'])){
+			$exampleUrl = 'https://cyfrowe.mnw.art.pl/en/catalog/445066';
+
+			if($parsedUrl['host'] != 'cyfrowe.mnw.art.pl'){
+				throw new Exceptions\InvalidMuseumUrlException($url, $exampleUrl);
+			}
+
+			if(!preg_match('|^/en/catalog/[^/]+?$|ius', $parsedUrl['path'])){
+				throw new Exceptions\InvalidMuseumUrlException($url, $exampleUrl);
+			}
+
+			$outputUrl = 'https://' . $parsedUrl['host'] . $parsedUrl['path'];
+
+			return $outputUrl;
+		}
 
 		return $outputUrl;
 	}