Fix single character wildcard usage

CHANGELOG.md

@@ -57,21 +57,21 @@
   * Namespace IDs (rather than just namespace names) work for all namespaces.
 * Support for a wide range of relative timestamps (e.g., "last week", "in 2 hours", "1 month ago").
   Previously only a few relative timestamps where supported, now anything supported by PHPs `strtotime` is supported.
-- **Gallery Mode Enhancements:**  
-  - Properly supports retrieving **PageImages** when using `mode=gallery` on non-file namespace pages, if the **PageImages** extension is installed. 
-  - Now pulls directly from the **PageImages** extension instead of relying on `page_props`, allowing access to both `page_image` and `page_image_free`.
-  - Added support for `mode=gallery` when using **Intersection compatibility mode** (i.e., `<DynamicPageList>` tags).
-  - Improves support for `%IMAGE%`:  
+* **Gallery Mode Enhancements:**  
+  * Properly supports retrieving **PageImages** when using `mode=gallery` on non-file namespace pages, if the **PageImages** extension is installed. 
+  * Now pulls directly from the **PageImages** extension instead of relying on `page_props`, allowing access to both `page_image` and `page_image_free`.
+  * Added support for `mode=gallery` when using **Intersection compatibility mode** (i.e., `<DynamicPageList>` tags).
+  * Improves support for `%IMAGE%`:  
     It no longer relies on stock `/images/` directories and instead dynamically resolves images based on whatever file backend is in use.
-  - Introduces a new parameter: `gallerymode`.  
+  * Introduces two new parameters: `imagewidth` and `imageheight`, to control the size of images within galleries.
+  * Introduces another new parameter: `gallerymode`.  
     This sets the `mode=` attribute in the generated `<gallery>` tag. Defaults to **`traditional`**, but supports all standard modes:
-    - `traditional`
-    - `nolines`
-    - `packed`
-    - `packed-hover`
-    - `packed-overlay`
-    - `slideshow`
-  - Introduce two other new parameters: `imagewidth` and `imageheight`, to control the size of images within galleries.
+    * `traditional`
+    * `nolines`
+    * `packed`
+    * `packed-hover`
+    * `packed-overlay`
+    * `slideshow`
 
   For more on supported gallery modes, see:  
 https://www.mediawiki.org/wiki/Help:Images#Gallery_syntax
@@ -85,6 +85,18 @@ Previously, using `&` in a category name would incorrectly be interpreted as a l
 
 ----
 
+Previously, underscores (`_`) in LIKE-style patterns (like in `titlematch` and `nottitlematch`) were implicitly treated as single-character wildcards. This made it impossible to reliably match literal underscores (which also represent spaces), and caused patterns to match more broadly than intended.
+
+This behavior has been fixed:
+
+* A single-character wildcard must now be written explicitly as `[_]`.
+* Each `[_]` represents exactly one character (for example, `[_][_][_]` matches three characters).
+* A bare `_` is now treated as a literal underscore and will no longer act as a wildcard.
+
+**NOTE:** We do it with brackets, because brackets are disallowed title characters, so they won't cause issues in the future with some edge-cases since no title can actually contain them. For similar reasons in the future, the same or similar method may be needed for `%` in order to allow better matching of literal `%` in titles, but that is a change that will break much more uses so will be done differently, perhaps keeping existing behavior but adding some other substitution to match literal `%`. For now, matching literal `%` remains unsupported or unreliable.
+
+----
+
 The template transclusion (`Extension DPL`) has been replaced with a proper tracking category, `Pages using DynamicPageList4`. All the references and usages of `Extension DPL` has been removed, including the `CreateTemplate` maintenance script. There is now a new maintenance script, `DeleteTemplate` to delete the old template. It will auto run when updated using `update.php`.
 
 ----

includes/Query.php

@@ -465,12 +465,19 @@ private function buildRegexpExpression( string $field, string $value ): string {
 	 * @return non-empty-array<string|LikeMatch>
 	 */
 	private function splitLikePattern( string $subject ): array {
-		$segments = preg_split( '/(%)/', $subject, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY );
+		// We use underscores in brackets so that it isn't a legal title character,
+		// which may cause confusing behavior in some edge cases where you enter
+		// underscore thinking it is for spaces, like it is elsewhere.
+		$segments = preg_split( '/(%|\[_\])/', $subject, -1, PREG_SPLIT_DELIM_CAPTURE | PREG_SPLIT_NO_EMPTY );
 		$parts = array_map(
-			fn ( string $segment ): string|LikeMatch =>
-				$segment === '%' ? $this->dbr->anyString() : $segment,
+			fn ( string $segment ): string|LikeMatch => match ( $segment ) {
+				'%' => $this->dbr->anyString(),
+				'[_]' => $this->dbr->anyChar(),
+				default => $segment,
+			},
 			$segments
 		);
+
 		return $parts ?: [ '' ];
 	}
 

tests/parser/selection.txt

@@ -78,6 +78,48 @@ format=,%TITLE%,;,
 </p>
 !! end
 
+!! test
+titlematch: single-character wildcard ([_])
+!! wikitext
+<dpl>
+namespace = 0
+ordermethod=title
+titlematch = P[_][_][_]
+format=,%TITLE%,;,
+</dpl>
+!! html
+<p>Pear;Plum;
+</p>
+!! end
+
+!! test
+titlematch: single-character wildcard with fixed prefix
+!! wikitext
+<dpl>
+namespace = 0
+ordermethod=title
+titlematch = G[_][_][_][_]
+format=,%TITLE%,;,
+</dpl>
+!! html
+<p>Grape;
+</p>
+!! end
+
+!! test
+titlematch: mixed % and [_] wildcard
+!! wikitext
+<dpl>
+namespace = 0
+ordermethod=title
+titlematch = [_]p%e
+format=,%TITLE%,;,
+</dpl>
+!! html
+<p>Apple;
+</p>
+!! end
+
 !! test
 nottitlematch: exclude specific pages
 !! wikitext
@@ -106,6 +148,34 @@ format=,%TITLE%,;,
 </p>
 !! end
 
+!! test
+nottitlematch: single-character wildcard ([_])
+!! wikitext
+<dpl>
+namespace = 0
+ordermethod=title
+nottitlematch = [_]range
+format=,%TITLE%,;,
+</dpl>
+!! html
+<p>Apple;Apricot;Banana;Grape;Mango;Pear;Plum;
+</p>
+!! end
+
+!! test
+nottitlematch: mixed % and [_] wildcard
+!! wikitext
+<dpl>
+namespace = 0
+ordermethod=title
+nottitlematch = %a[_]g%
+format=,%TITLE%,;,
+</dpl>
+!! html
+<p>Apple;Apricot;Banana;Grape;Pear;Plum;
+</p>
+!! end
+
 !! test
 replaceintitle: simple string replacement
 !! wikitext