Tests for ad hoc content sniffing

4 years ago · 251249b11e
2 changed files with 20 additions and 6 deletions
--- a/lib/Parser/Parser.php
+++ b/lib/Parser/Parser.php
@ -35,21 +35,21 @@ abstract class Parser {
            return "application/xml";
        } elseif (preg_match('/^\s*</s', $data)) {
            // distinguish between XML feeds and HTML; first skip any comments before the root element
-            $offset = preg_match('/^\s*(?:<!--(?:[^\-]|-(?!->)*-->\s*)*/s', $data, $match) ? strlen($match[0]) : 0;
+            $offset = preg_match('/^\s*(<!(?=--)([^-]|-(?!->))*-->\s*)*/s', $data, $match) ? strlen($match[0]) : 0;
            $prefix = substr($data, $offset, 100);
            if (preg_match('/^<(?:!DOCTYPE\s+html|html|body|head|table|div|title|p|link|meta)[\s>]/si', $prefix)) {
                return "text/html";
            } elseif (preg_match('/^<rss[\s>\/]/', $prefix)) {
                return "application/rss+xml";
-            } elseif (preg_match('/^<(?:[A-Za-z0-9\-\._]+:)?(feed|RDF)\s/', $prefix)) {
+            } elseif (preg_match('/^<(?:[A-Za-z0-9\-\._]+:)?(feed|RDF)\s/', $prefix, $match)) {
                if ($match[1] === "feed") {
                    return "application/atom+xml";
                } else {
-                    return "applicatiojn/rdf+xml";
+                    return "application/rdf+xml";
                }
            } else {
                // FIIXME: Is there a better fallback that could used here?
-                "application/xml";
+                return "application/xml";
            }
        } else {
            return "application/octet-stream";
--- a/tests/cases/Parser/ParserTest.php
+++ b/tests/cases/Parser/ParserTest.php
@ -46,8 +46,22 @@ class ParserTest extends \PHPUnit\Framework\TestCase {

    public function provideDetectableContent(): iterable {
        return [
-            ['{""}',      "application/json"],
-            [" \n  {\"v", "application/json"],
+            ['{""}',                             "application/json"],
+            [" \n  {\"v",                        "application/json"],
+            ["<?xml",                            "application/xml"],
+            [" \n <?xml",                        "application/xml"],
+            ["<!DOCTYPE html>",                  "text/html"],
+            ["<!DOCTYPE html ",                  "text/html"],
+            [" \n <!DOCTYPE html>",              "text/html"],
+            [" \n <!DOCTYPE html\n",             "text/html"],
+            [" <!-- --> <!-- oops -->\n <rss>",  "application/rss+xml"],
+            [" <!--> <!-- oops -->\n <rss>",     "application/rss+xml"],
+            ["<feed ",                           "application/atom+xml"],
+            ["<atom:feed ",                      "application/atom+xml"],
+            ["<RDF ",                            "application/rdf+xml"],
+            ["<rdf:RDF ",                        "application/rdf+xml"],
+            ["<opml>",                           "application/xml"],
+            ["plain text",                       "application/octet-stream"],
        ];
    }
 }