From 4b55eccea4ce8edb51cb628aa8651911ea14f3e7 Mon Sep 17 00:00:00 2001
From: Paul Pound <ppound@upei.ca>
Date: Fri, 4 Sep 2026 15:02:03 -0300
Subject: [PATCH] =?UTF-8?q?added=20ICUFoldingFilter=20so=20=C3=96konomie?=
 =?UTF-8?q?=20sorts=20with=20Okonomie?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 README.md                                     | 21 +++++++++++++++----
 .../SolrConfigSetSubscriber.php               | 17 ++++++++++-----
 2 files changed, 29 insertions(+), 9 deletions(-)

diff --git a/README.md b/README.md
index 8f16aac..bad90e0 100644
--- a/README.md
+++ b/README.md
@@ -113,10 +113,10 @@ field types do not have to precede the fields that use them.
 <fieldType name="string_sort_ignore_articles" class="solr.TextField" sortMissingLast="true" omitNorms="true">
   <analyzer>
     <tokenizer class="solr.KeywordTokenizerFactory"/>
+    <filter class="solr.ICUFoldingFilterFactory"/>
     <filter class="solr.PatternReplaceFilterFactory" pattern="^[^\p{L}\p{N}]+" replacement="" replace="all"/>
     <filter class="solr.PatternReplaceFilterFactory" pattern="^(?i)(a\s+|an\s+|the\s+)" replacement="" replace="all"/>
     <filter class="solr.PatternReplaceFilterFactory" pattern="^[^\p{L}\p{N}]+" replacement="" replace="all"/>
-    <filter class="solr.LowerCaseFilterFactory"/>
     <filter class="solr.TrimFilterFactory"/>
   </analyzer>
 </fieldType>
@@ -135,7 +135,13 @@ zip:
 Keep the comment markers — they are what makes the block findable and cleanly
 removable later.
 
-Both blocks load unchanged on Solr 8.11 and 9.x.
+Both blocks load unchanged on Solr 8.11 and 9.x. `ICUFoldingFilterFactory`
+needs the ICU analysis jar, but any core running the stock `search_api_solr`
+config set already has it — every `collated_*` field type is a
+`solr.ICUCollationField` from that same jar, and the core would not load at
+all without it. On Solr 8 the config set loads it through the
+`contrib/analysis-extras` `<lib>` directives; on Solr 9 it comes from
+`SOLR_MODULES` including `analysis-extras`.
 
 Reload the core:
 
@@ -162,8 +168,15 @@ search view.
 
 ## Verifying
 
-At `/solr/#/<core>/analysis`, field type `string_sort_ignore_articles` with
-input `The !Matrix` should produce the single token `matrix`.
+At `/solr/#/<core>/analysis`, field type `string_sort_ignore_articles` should
+reduce each of these to a single token:
+
+| Input | Token | Checks |
+| --- | --- | --- |
+| `The !Matrix` | `matrix` | article and punctuation both stripped |
+| `A Tale of Two Cities` | `tale of two cities` | one token, not one per word |
+| `Theory of Everything` | `theory of everything` | a word starting with an article is left alone |
+| `Ökonomie` | `okonomie` | diacritics folded, so it sorts under O |
 
 The field itself can be confirmed with the Schema API:
 
diff --git a/src/EventSubscriber/SolrConfigSetSubscriber.php b/src/EventSubscriber/SolrConfigSetSubscriber.php
index f585dec..b4fa213 100644
--- a/src/EventSubscriber/SolrConfigSetSubscriber.php
+++ b/src/EventSubscriber/SolrConfigSetSubscriber.php
@@ -94,15 +94,22 @@ class SolrConfigSetSubscriber implements EventSubscriberInterface {
   <analyzer>
     <!-- 1. Keep the entire string as a single token. -->
     <tokenizer class="solr.KeywordTokenizerFactory"/>
-    <!-- 2. Strip leading punctuation, symbols and spaces.
+    <!-- 2. NFKC normalise, case fold and strip diacritics, so 'Matrix',
+         'matrix' and 'Mátrix' sort together instead of the accented forms
+         landing after every ASCII letter. This also normalises a no-break
+         space to a plain one, so step 4 can see the article boundary.
+         Needs the ICU analysis jar, which is already loaded: the stock
+         search_api_solr schema types every collated_* field as
+         solr.ICUCollationField from the same jar. -->
+    <filter class="solr.ICUFoldingFilterFactory"/>
+    <!-- 3. Strip leading punctuation, symbols and spaces.
          [^\p{L}\p{N}] is "anything that is NOT a letter or a number". -->
     <filter class="solr.PatternReplaceFilterFactory" pattern="^[^\p{L}\p{N}]+" replacement="" replace="all"/>
-    <!-- 3. Remove a leading 'A ', 'An ' or 'The ' (case-insensitive). -->
+    <!-- 4. Remove a leading 'A ', 'An ' or 'The '. Step 2 has already lower
+         cased the token; (?i) is kept so this still holds on its own. -->
     <filter class="solr.PatternReplaceFilterFactory" pattern="^(?i)(a\s+|an\s+|the\s+)" replacement="" replace="all"/>
-    <!-- 4. Strip leading punctuation again, to catch "The !Matrix". -->
+    <!-- 5. Strip leading punctuation again, to catch "The !Matrix". -->
     <filter class="solr.PatternReplaceFilterFactory" pattern="^[^\p{L}\p{N}]+" replacement="" replace="all"/>
-    <!-- 5. Lowercase so 'Matrix' and 'matrix' sort together. -->
-    <filter class="solr.LowerCaseFilterFactory"/>
     <!-- 6. Trim any accidental leading/trailing whitespace. -->
     <filter class="solr.TrimFilterFactory"/>
   </analyzer>