Merge pull request #3553 from wallabag/url-3529

Swap entry url with origin url if graby provides an updated one

2 files changed, 244 insertions, 1 deletions

diff --git a/src/Wallabag/CoreBundle/Helper/ContentProxy.php b/src/Wallabag/CoreBundle/Helper/ContentProxy.php
index d4ea608f..d38811a2 100644
--- a/src/Wallabag/CoreBundle/Helper/ContentProxy.php
+++ b/src/Wallabag/CoreBundle/Helper/ContentProxy.php
@@ -66,6 +66,13 @@ class ContentProxy
         // so we'll be able to refetch it in the future
         $content['url'] = !empty($content['url']) ? $content['url'] : $url;
 
+        // In one case (at least in tests), url is empty here
+        // so we set it using $url provided in the updateEntry call.
+        // Not sure what are the other possible cases where this property is empty
+        if (empty($entry->getUrl()) && !empty($url)) {
+            $entry->setUrl($url);
+        }
+
         $this->stockEntry($entry, $content);
     }
 
@@ -239,7 +246,7 @@ class ContentProxy
      */
     private function stockEntry(Entry $entry, array $content)
     {
-        $entry->setUrl($content['url']);
+        $this->updateOriginUrl($entry, $content['url']);
 
         $this->setEntryDomainName($entry);
 
@@ -306,6 +313,115 @@ class ContentProxy
     }
 
     /**
+     * Update the origin_url field when a redirection occurs
+     * This field is set if it is empty and new url does not match ignore list.
+     *
+     * @param Entry  $entry
+     * @param string $url
+     */
+    private function updateOriginUrl(Entry $entry, $url)
+    {
+        if (empty($url) || $entry->getUrl() === $url) {
+            return false;
+        }
+
+        $parsed_entry_url = parse_url($entry->getUrl());
+        $parsed_content_url = parse_url($url);
+
+        /**
+         * The following part computes the list of part changes between two
+         * parse_url arrays.
+         *
+         * As array_diff_assoc only computes changes to go from the left array
+         * to the right one, we make two differents arrays to have both
+         * directions. We merge these two arrays and sort keys before passing
+         * the result to the switch.
+         *
+         * The resulting array gives us all changing parts between the two
+         * urls: scheme, host, path, query and/or fragment.
+         */
+        $diff_ec = array_diff_assoc($parsed_entry_url, $parsed_content_url);
+        $diff_ce = array_diff_assoc($parsed_content_url, $parsed_entry_url);
+
+        $diff = array_merge($diff_ec, $diff_ce);
+        $diff_keys = array_keys($diff);
+        sort($diff_keys);
+
+        if ($this->ignoreUrl($entry->getUrl())) {
+            $entry->setUrl($url);
+
+            return false;
+        }
+
+        /**
+         * This switch case lets us apply different behaviors according to
+         * changing parts of urls.
+         *
+         * As $diff_keys is an array, we provide arrays as cases. ['path'] means
+         * 'only the path is different between the two urls' whereas
+         * ['fragment', 'query'] means 'only fragment and query string parts are
+         * different between the two urls'.
+         *
+         * Note that values in $diff_keys are sorted.
+         */
+        switch ($diff_keys) {
+            case ['path']:
+                if (($parsed_entry_url['path'] . '/' === $parsed_content_url['path']) // diff is trailing slash, we only replace the url of the entry
+                    || ($url === urldecode($entry->getUrl()))) { // we update entry url if new url is a decoded version of it, see EntryRepository#findByUrlAndUserId
+                    $entry->setUrl($url);
+                }
+                break;
+            case ['scheme']:
+                $entry->setUrl($url);
+                break;
+            case ['fragment']:
+                // noop
+                break;
+            default:
+                if (empty($entry->getOriginUrl())) {
+                    $entry->setOriginUrl($entry->getUrl());
+                }
+                $entry->setUrl($url);
+                break;
+        }
+    }
+
+    /**
+     * Check entry url against an ignore list to replace with content url.
+     *
+     * XXX: move the ignore list in the database to let users handle it
+     *
+     * @param string $url url to test
+     *
+     * @return bool true if url matches ignore list otherwise false
+     */
+    private function ignoreUrl($url)
+    {
+        $ignored_hosts = ['feedproxy.google.com', 'feeds.reuters.com'];
+        $ignored_patterns = ['https?://www\.lemonde\.fr/tiny.*'];
+
+        $parsed_url = parse_url($url);
+
+        $filtered = array_filter($ignored_hosts, function ($var) use ($parsed_url) {
+            return $var === $parsed_url['host'];
+        });
+
+        if ([] !== $filtered) {
+            return true;
+        }
+
+        $filtered = array_filter($ignored_patterns, function ($var) use ($url) {
+            return preg_match("`$var`i", $url);
+        });
+
+        if ([] !== $filtered) {
+            return true;
+        }
+
+        return false;
+    }
+
+    /**
      * Validate that the given content has at least a title, an html and a url.
      *
      * @param array $content
diff --git a/tests/Wallabag/CoreBundle/Helper/ContentProxyTest.php b/tests/Wallabag/CoreBundle/Helper/ContentProxyTest.php
index 3f3c60d0..3dd9273c 100644
--- a/tests/Wallabag/CoreBundle/Helper/ContentProxyTest.php
+++ b/tests/Wallabag/CoreBundle/Helper/ContentProxyTest.php
@@ -740,6 +740,133 @@ class ContentProxyTest extends TestCase
     }
 
     /**
+     * Data provider for testWithChangedUrl.
+     *
+     * Arrays contain the following values:
+     * $entry_url
+     * $origin_url
+     * $content_url
+     * $expected_entry_url
+     * $expected_origin_url
+     * $expected_domain
+     */
+    public function dataForChangedUrl()
+    {
+        return [
+            'normal' => [
+                'http://0.0.0.0',
+                null,
+                'http://1.1.1.1',
+                'http://1.1.1.1',
+                'http://0.0.0.0',
+                '1.1.1.1',
+            ],
+            'origin already set' => [
+                'http://0.0.0.0',
+                'http://hello',
+                'http://1.1.1.1',
+                'http://1.1.1.1',
+                'http://hello',
+                '1.1.1.1',
+            ],
+            'trailing slash' => [
+                'https://example.com/hello-world',
+                null,
+                'https://example.com/hello-world/',
+                'https://example.com/hello-world/',
+                null,
+                'example.com',
+            ],
+            'query string in fetched content' => [
+                'https://example.org/hello',
+                null,
+                'https://example.org/hello?world=1',
+                'https://example.org/hello?world=1',
+                'https://example.org/hello',
+                'example.org',
+            ],
+            'fragment in fetched content' => [
+                'https://example.org/hello',
+                null,
+                'https://example.org/hello#world',
+                'https://example.org/hello',
+                null,
+                'example.org',
+            ],
+            'fragment and query string in fetched content' => [
+                'https://example.org/hello',
+                null,
+                'https://example.org/hello?foo#world',
+                'https://example.org/hello?foo#world',
+                'https://example.org/hello',
+                'example.org',
+            ],
+            'different path and query string in fetch content' => [
+                'https://example.org/hello',
+                null,
+                'https://example.org/world?foo',
+                'https://example.org/world?foo',
+                'https://example.org/hello',
+                'example.org',
+            ],
+            'feedproxy ignore list test' => [
+                'http://feedproxy.google.com/~r/Wallabag/~3/helloworld',
+                null,
+                'https://example.org/hello-wallabag',
+                'https://example.org/hello-wallabag',
+                null,
+                'example.org',
+            ],
+            'feedproxy ignore list test with origin url already set' => [
+                'http://feedproxy.google.com/~r/Wallabag/~3/helloworld',
+                'https://example.org/this-is-source',
+                'https://example.org/hello-wallabag',
+                'https://example.org/hello-wallabag',
+                'https://example.org/this-is-source',
+                'example.org',
+            ],
+            'lemonde ignore pattern test' => [
+                'http://www.lemonde.fr/tiny/url',
+                null,
+                'http://example.com/hello-world',
+                'http://example.com/hello-world',
+                null,
+                'example.com',
+            ],
+        ];
+    }
+
+    /**
+     * @dataProvider dataForChangedUrl
+     */
+    public function testWithChangedUrl($entry_url, $origin_url, $content_url, $expected_entry_url, $expected_origin_url, $expected_domain)
+    {
+        $tagger = $this->getTaggerMock();
+        $tagger->expects($this->once())
+            ->method('tag');
+
+        $proxy = new ContentProxy((new Graby()), $tagger, $this->getValidator(), $this->getLogger(), $this->fetchingErrorMessage, true);
+        $entry = new Entry(new User());
+        $entry->setOriginUrl($origin_url);
+        $proxy->updateEntry(
+            $entry,
+            $entry_url,
+            [
+                'html' => false,
+                'title' => '',
+                'url' => $content_url,
+                'content_type' => '',
+                'language' => '',
+            ],
+            true
+        );
+
+        $this->assertSame($expected_entry_url, $entry->getUrl());
+        $this->assertSame($expected_domain, $entry->getDomainName());
+        $this->assertSame($expected_origin_url, $entry->getOriginUrl());
+    }
+
+    /**
      * https://stackoverflow.com/a/18506801.
      *
      * @param $string