fix: send searched words only as file names

A search that found nothing sent any short text, which a name could be.
Its words now go out only as a file name or a hash, every other search
is counted as "other" without them.

tests/test_site_privacy.py

@@ -44,9 +44,9 @@ class TheCounterAsksNothingOfTheDevice(unittest.TestCase):
         self.assertIn("from.origin", self.code)
         self.assertNotIn("from.pathname", self.code)
 
-    def test_a_search_with_personal_details_is_not_sent(self) -> None:
-        self.assertIn("!query.includes('@')", self.code)
-        self.assertIn("\\d{7,}", self.code)
+    def test_a_search_is_sent_only_as_a_file_name_or_a_hash(self) -> None:
+        self.assertIn("isHash(query) || isFileName(query) || isFileStem(query) ? query : 'other'", self.code)
+        self.assertIn("countEvent(`search-none/${searchWords(query)}`)", self.code)
         self.assertIn("message.startsWith('No results for')", self.code)
 
     def test_one_endpoint(self) -> None:

website/src/scripts/count.ts

@@ -7,8 +7,8 @@
 // Sent: the page path (never its query or #anchor), the domain of the site
 // the reader came from, and the name of an action. An action is a click on
 // a copy button, a pack or file download, a link that leaves the site, a
-// site search that found nothing (the words, when they cannot hold a
-// person's details), and the answer to "Did this page help?".
+// site search that found nothing (its words only when they are a file
+// name or a hash), and the answer to "Did this page help?".
 
 const ENDPOINT = 'https://retrobios.goatcounter.com/count';
 const HOST = 'abdess.github.io';
@@ -104,11 +104,18 @@ const onClick = (event: MouseEvent): void => {
 // Search
 
 // A search that found nothing names a missing page. Its words are sent
-// only when they cannot hold a person's details: short, no "@", no long
-// run of digits. A hash is a file asked for, and is kept.
-const isHash = (query: string): boolean => /^[0-9a-f]{8}$|^[0-9a-f]{32}$|^[0-9a-f]{40}$|^[0-9a-f]{64}$/.test(query);
-const sendable = (query: string): boolean =>
-	isHash(query) || (query.length >= 2 && query.length <= 40 && !query.includes('@') && !/\d{7,}/.test(query));
+// only in a shape no person's details take: a hash, a file name
+// ("scph5501.bin"), or one word of letters and at most four digits
+// ("scph5501", "ps3updat"). Any other search is counted with no words,
+// as "other": a name, a phone number or an address is never sent.
+const isHash = (query: string): boolean => /^(?:[0-9a-f]{8}|[0-9a-f]{32}|[0-9a-f]{40}|[0-9a-f]{64})$/.test(query);
+// The extensions of the files the collection holds: "john.smith" is no file.
+const isFileName = (query: string): boolean =>
+	/^[\w.+-]{1,60}\.(?:bin|rom|zip|7z|pup|keys|img|dat|bios|ic\d+|u\d+|sys|nvm|mec|rom1|j64|pce|col|sms|gb|gbc|gba|nds|cfg|chd|iso|txt|fw|xml|yaml|pak|pk3|grp)$/.test(query);
+const isFileStem = (query: string): boolean =>
+	/^[a-z][a-z0-9_-]{1,30}$/.test(query) && /\d/.test(query) && (query.match(/\d/g) ?? []).length <= 4;
+const searchWords = (query: string): string =>
+	isHash(query) || isFileName(query) || isFileStem(query) ? query : 'other';
 
 const watchSearch = (): void => {
 	const root = document.getElementById('starlight__search');
@@ -121,9 +128,9 @@ const watchSearch = (): void => {
 		// Pagefind's English message once a search has ended with nothing;
 		// while it runs the message reads "Searching for ...".
 		const message = root.querySelector('.pagefind-ui__message')?.textContent ?? '';
-		if (!query || !message.startsWith('No results for') || sent.has(query) || !sendable(query)) return;
+		if (!query || !message.startsWith('No results for') || sent.has(query)) return;
 		sent.add(query);
-		countEvent(`search-none/${query}`);
+		countEvent(`search-none/${searchWords(query)}`);
 	};
 	new MutationObserver(() => {
 		window.clearTimeout(timer);

wiki/privacy.md

@@ -25,8 +25,9 @@ code made from the request for 8 hours, then drops it.
 These actions are counted the same way, with the page they happened on:
 
 - a click on a copy button, a pack download or a file link;
-- a site search that found nothing, with its words, so a missing page shows up.
-  Words with an `@` or a long number are never sent;
+- a site search that found nothing, so a missing page shows up. Its words are
+  sent only when they read as a file name (`scph5501.bin`, `scph5501`) or a
+  hash. Any other search is counted as "other", without its words;
 - the answer to "Did this page help?".
 
 Not sent: your screen size, your language, a search that found results.