MDL-78525 core: fix word and character counting

lib/moodlelib.php

@@ -8381,9 +8381,10 @@ function moodle_setlocale($locale='') {
  *
  * @category string
  * @param string $string The text to be searched for words. May be HTML.
+ * @param int|null $format
  * @return int The count of words in the specified string
  */
-function count_words($string) {
+function count_words($string, $format = null) {
     // Before stripping tags, add a space after the close tag of anything that is not obviously inline.
     // Also, br is a special case because it definitely delimits a word, but has no close tag.
     $string = preg_replace('~
@@ -8400,6 +8401,11 @@ function count_words($string) {
                 <br> | <br\s*/>                 # Special cases that are not close tags.
             )
             ~x', '$1 ', $string); // Add a space after the close tag.
+    if ($format !== null && $format != FORMAT_PLAIN) {
+        // Match the usual text cleaning before display.
+        // Ideally we should apply multilang filter only here, other filters might add extra text.
+        $string = format_text($string, $format, ['filter' => false, 'noclean' => false, 'para' => false]);
+    }
     // Now remove HTML tags.
     $string = strip_tags($string);
     // Decode HTML entities.
@@ -8421,9 +8427,15 @@ function count_words($string) {
  *
  * @category string
  * @param string $string The text to be searched for letters. May be HTML.
+ * @param int|null $format
  * @return int The count of letters in the specified text.
  */
-function count_letters($string) {
+function count_letters($string, $format = null) {
+    if ($format !== null && $format != FORMAT_PLAIN) {
+        // Match the usual text cleaning before display.
+        // Ideally we should apply multilang filter only here, other filters might add extra text.
+        $string = format_text($string, $format, ['filter' => false, 'noclean' => false, 'para' => false]);
+    }
     $string = strip_tags($string); // Tags are out now.
     $string = html_entity_decode($string, ENT_COMPAT);
     $string = preg_replace('/[[:space:]]*/', '', $string); // Whitespace are out now.

lib/tests/moodlelib_test.php

@@ -3958,9 +3958,11 @@ EOF;
      * @dataProvider count_words_testcases
      * @param int $expectedcount number of words in $string.
      * @param string $string the test string to count the words of.
+     * @param int|null $format
      */
-    public function test_count_words(int $expectedcount, string $string): void {
-        $this->assertEquals($expectedcount, count_words($string));
+    public function test_count_words(int $expectedcount, string $string, $format = null): void {
+        $this->assertEquals($expectedcount, count_words($string, $format),
+            "'$string' with format '$format' does not match count $expectedcount");
     }
 
     /**
@@ -3969,6 +3971,13 @@ EOF;
      * @return array of test cases.
      */
     public function count_words_testcases(): array {
+        // Copy-pasting example from MDL-64240.
+        $copypasted = <<<EOT
+<p onclick="alert('boop');">Snoot is booped</p>
+ <script>alert('Boop the snoot');</script>
+ <img alt="Boop the Snoot." src="https://proxy.duckduckgo.com/iu/?u=http%3A%2F%2Fwww.geekfill.com%2Fwp-content%2Fuploads%2F2015%2F08%2FBoop-the-Snoot.jpg&f=1">
+EOT;
+
         // The counts here should match MS Word and Libre Office.
         return [
             [0, ''],
@@ -4005,6 +4014,16 @@ EOF;
             [1, "SO<sub>4</sub><sup>2-</sup>"],
             [6, '4+4=8 i.e. O(1) a,b,c,d I’m black&blue_really'],
             [1, '<span>a</span><span>b</span>'],
+            [1, '<span>a</span><span>b</span>', FORMAT_PLAIN],
+            [1, '<span>a</span><span>b</span>', FORMAT_HTML],
+            [1, '<span>a</span><span>b</span>', FORMAT_MOODLE],
+            [1, '<span>a</span><span>b</span>', FORMAT_MARKDOWN],
+            [1, 'aa <argh <bleh>pokus</bleh>'],
+            [2, 'aa <argh <bleh>pokus</bleh>', FORMAT_HTML],
+            [6, $copypasted],
+            [6, $copypasted, FORMAT_PLAIN],
+            [3, $copypasted, FORMAT_HTML],
+            [3, $copypasted, FORMAT_MOODLE],
         ];
     }
 
@@ -4014,9 +4033,11 @@ EOF;
      * @dataProvider count_letters_testcases
      * @param int $expectedcount number of characters in $string.
      * @param string $string the test string to count the letters of.
+     * @param int|null $format
      */
-    public function test_count_letters(int $expectedcount, string $string): void {
-        $this->assertEquals($expectedcount, count_letters($string));
+    public function test_count_letters(int $expectedcount, string $string, $format = null): void {
+        $this->assertEquals($expectedcount, count_letters($string, $format),
+            "'$string' with format '$format' does not match count $expectedcount");
     }
 
     /**
@@ -4030,6 +4051,12 @@ EOF;
             [1, 'x'],
             [1, '&amp;'],
             [4, '<p>frog</p>'],
+            [4, '<p>frog</p>', FORMAT_PLAIN],
+            [4, '<p>frog</p>', FORMAT_MOODLE],
+            [4, '<p>frog</p>', FORMAT_HTML],
+            [4, '<p>frog</p>', FORMAT_MARKDOWN],
+            [2, 'aa <argh <bleh>pokus</bleh>'],
+            [7, 'aa <argh <bleh>pokus</bleh>', FORMAT_HTML],
         ];
     }
 

mod/forum/classes/local/entities/post.php

@@ -350,8 +350,8 @@ class post {
      */
     public static function add_message_counts(\stdClass $record) : void {
         if (!empty($record->message)) {
-            $record->wordcount = count_words($record->message);
-            $record->charcount = count_letters($record->message);
+            $record->wordcount = count_words($record->message, $record->messageformat);
+            $record->charcount = count_letters($record->message, $record->messageformat);
         }
     }
 }

mod/forum/post.php

@@ -799,10 +799,9 @@ if ($mformpost->is_cancelled()) {
     // WARNING: the $fromform->message array has been overwritten, do not use it anymore!
     $fromform->messagetrust  = trusttext_trusted($modcontext);
 
-    // Clean message text, unless markdown which should be saved as it is, otherwise editing messes things up.
-    if ($fromform->messageformat != FORMAT_MARKDOWN) {
-        $fromform = trusttext_pre_edit($fromform, 'message', $modcontext);
-    }
+    // Do not clean text here, text cleaning can be done only after conversion to HTML.
+    // Word counting now uses text formatting, there is no need to abuse trusttext_pre_edit() here.
+
     if ($fromform->edit) {
         // Updating a post.
         unset($fromform->groupid);