From 16cc395f7834774ac7c518a1e2ba9584ef44a896 Mon Sep 17 00:00:00 2001
From: Sarah Hoffmann <lonvia@denofr.de>
Date: Mon, 25 Oct 2021 13:08:16 +0200
Subject: [PATCH] fix warming for ICU tokenizer

Running the warm-up search requests requires querying
the most frequent words. This must be done via the tokenizer
to honor the different formats of the word table.
---
 lib-php/admin/warm.php                 | 7 ++++++-
 lib-php/tokenizer/icu_tokenizer.php    | 9 +++++++++
 lib-php/tokenizer/legacy_tokenizer.php | 8 ++++++++
 3 files changed, 23 insertions(+), 1 deletion(-)

diff --git a/lib-php/admin/warm.php b/lib-php/admin/warm.php
index 39a37506..338ec2da 100644
--- a/lib-php/admin/warm.php
+++ b/lib-php/admin/warm.php
@@ -86,8 +86,13 @@ if (!$aResult['reverse-only']) {
     if ($bVerbose) {
         echo "\n";
     }
+
+    $oTokenizer = new \Nominatim\Tokenizer($oDB);
+
+    $aWords = $oTokenizer->mostFrequentWords(1000);
+
     $sSQL = 'SELECT word FROM word WHERE word is not null ORDER BY search_name_count DESC LIMIT 1000';
-    foreach ($oDB->getCol($sSQL) as $sWord) {
+    foreach ($aWords as $sWord) {
         if ($bVerbose) {
             echo "$sWord = ";
         }
diff --git a/lib-php/tokenizer/icu_tokenizer.php b/lib-php/tokenizer/icu_tokenizer.php
index ca224a22..f4dd3aeb 100644
--- a/lib-php/tokenizer/icu_tokenizer.php
+++ b/lib-php/tokenizer/icu_tokenizer.php
@@ -40,6 +40,15 @@ class Tokenizer
         return $this->oNormalizer->transliterate($sTerm);
     }
 
+
+    public function mostFrequentWords($iNum)
+    {
+        $sSQL = "SELECT word FROM word WHERE type = 'W'";
+        $sSQL .= "ORDER BY info->'count' DESC LIMIT ".$iNum;
+        return $this->oDB->getCol($sSQL);
+    }
+
+
     private function makeStandardWord($sTerm)
     {
         return trim($this->oTransliterator->transliterate(' '.$sTerm.' '));
diff --git a/lib-php/tokenizer/legacy_tokenizer.php b/lib-php/tokenizer/legacy_tokenizer.php
index b508d220..d5686f64 100644
--- a/lib-php/tokenizer/legacy_tokenizer.php
+++ b/lib-php/tokenizer/legacy_tokenizer.php
@@ -48,6 +48,14 @@ class Tokenizer
     }
 
 
+    public function mostFrequentWords($iNum)
+    {
+        $sSQL = 'SELECT word FROM word WHERE word is not null ';
+        $sSQL .= 'ORDER BY search_name_count DESC LIMIT '.$iNum;
+        return $this->oDB->getCol($sSQL);
+    }
+
+
     public function tokensForSpecialTerm($sTerm)
     {
         $aResults = array();
-- 
2.39.5