From aae958979581ad77bf770ca23a63436b5609117f Mon Sep 17 00:00:00 2001
From: niksedk <nikse.dk@gmail.com>
Date: Mon, 12 Oct 2026 02:59:54 +0200
Subject: [PATCH] AmericanToBritish: keep phrases ("por favor" no longer
 becomes "por favour")

Built-in <Keep text="..."/> entries in WordList.xml plus the local list's
<Ignore> entries (which may now be phrases) are masked before the rules run,
so a word inside them is left untouched while still converted elsewhere.

Fixes SubtitleEdit/subtitleedit#15946

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---
 se5/AmericanToBritish/README.md              |  4 +-
 se5/AmericanToBritish/WordListWindow.axaml   |  4 +-
 se5/Plugin-Shared/EnglishVariantConverter.cs | 97 ++++++++++++++++++++
 se5/Plugin-Shared/WordList.xml               |  3 +
 4 files changed, 105 insertions(+), 3 deletions(-)

diff --git a/se5/AmericanToBritish/README.md b/se5/AmericanToBritish/README.md
index b978244e..e3408da8 100644
--- a/se5/AmericanToBritish/README.md
+++ b/se5/AmericanToBritish/README.md
@@ -20,7 +20,9 @@ and format as the Subtitle Edit 4 plugin, so an existing list keeps working:
 ```
 
 A `Word` with the same American spelling as a built-in pair replaces it; `Ignore`
-turns a built-in pair off.
+turns a built-in pair off. An `Ignore` can also be a phrase, which is then left
+untouched wherever it occurs (`<Ignore us="por favor" />` keeps "favor" there while
+it is still converted elsewhere). A few such phrases are built in, e.g. "por favor".
 
 First Subtitle Edit 5 plugin built on top of the shared
 [`Plugin-Shared`](../Plugin-Shared/) library — Avalonia app boot, theme passing,
diff --git a/se5/AmericanToBritish/WordListWindow.axaml b/se5/AmericanToBritish/WordListWindow.axaml
index e30c2569..aa747384 100644
--- a/se5/AmericanToBritish/WordListWindow.axaml
+++ b/se5/AmericanToBritish/WordListWindow.axaml
@@ -94,13 +94,13 @@
                                Opacity="0.7"
                                TextWrapping="Wrap"
                                Margin="0,0,0,8"
-                               Text="American words from the built-in list that should never be converted." />
+                               Text="American words from the built-in list that should never be converted, or whole phrases to leave untouched (e.g. &quot;por favor&quot;)." />
                     <Grid DockPanel.Dock="Bottom"
                           ColumnDefinitions="*,Auto,Auto"
                           Margin="0,8,0,0">
                         <TextBox Grid.Column="0"
                                  x:Name="IgnoreBox"
-                                 Watermark="American word"
+                                 Watermark="American word or phrase"
                                  Margin="0,0,6,0" />
                         <Button Grid.Column="1"
                                 Content="Add"
diff --git a/se5/Plugin-Shared/EnglishVariantConverter.cs b/se5/Plugin-Shared/EnglishVariantConverter.cs
index 911a6d8f..f63eccfe 100644
--- a/se5/Plugin-Shared/EnglishVariantConverter.cs
+++ b/se5/Plugin-Shared/EnglishVariantConverter.cs
@@ -1,5 +1,7 @@
 using System;
 using System.Collections.Generic;
+using System.Linq;
+using System.Text;
 using System.Text.RegularExpressions;
 using System.Xml.Linq;
 
@@ -16,6 +18,8 @@ public enum EnglishVariantDirection
 /// (~1850 pairs), optionally extended/filtered by a <see cref="LocalWordList"/>.
 /// Each pair becomes three case-aware regexes: lowercase, UPPERCASE, and Titlecase,
 /// all matched as whole words. Picks direction via <see cref="EnglishVariantDirection"/>.
+/// Keep phrases (built-in &lt;Keep text="..."/&gt; plus the local list's ignored entries) are left
+/// untouched, e.g. the Spanish "por favor" must not become "por favour".
 /// </summary>
 public sealed class EnglishVariantConverter
 {
@@ -29,6 +33,13 @@ public sealed class EnglishVariantConverter
 
     private readonly EnglishVariantDirection _direction;
 
+    private readonly List<string> _keepPhrases = new();
+    private Regex? _keepRegex;
+
+    /// <summary>Private use area chars stand in for kept phrases while the rules run.</summary>
+    private const char PlaceholderFirst = '\uE000';
+    private const char PlaceholderLast = '\uF8FF';
+
     public EnglishVariantConverter(EnglishVariantDirection direction, LocalWordList? localWordList = null)
     {
         _direction = direction;
@@ -49,10 +60,21 @@ public EnglishVariantConverter(EnglishVariantDirection direction, LocalWordList?
             foreach (var us in localWordList.Ignored)
             {
                 ignored.Add(us);
+                _keepPhrases.Add(us);
             }
         }
 
         LoadBuiltInWordList(ignored);
+
+        if (_keepPhrases.Count > 0)
+        {
+            // Longest first so the longer of two overlapping phrases wins.
+            var alternatives = _keepPhrases
+                .Distinct(StringComparer.OrdinalIgnoreCase)
+                .OrderByDescending(p => p.Length)
+                .Select(p => Regex.Escape(p).Replace("\\ ", "\\s+"));
+            _keepRegex = new Regex("\\b(?:" + string.Join("|", alternatives) + ")\\b", RegexOptions.IgnoreCase | RegexOptions.ExplicitCapture);
+        }
     }
 
     public int RuleCount => _rules.Count;
@@ -64,6 +86,8 @@ public string Convert(string text)
             return text;
         }
 
+        var kept = MaskKeepPhrases(ref text);
+
         var candidates = CollectCandidateRules(text);
         foreach (var index in candidates)
         {
@@ -71,9 +95,73 @@ public string Convert(string text)
             text = pattern.Replace(text, replacement);
         }
 
+        text = UnmaskKeepPhrases(text, kept);
         return RevertFontColorAttribute(text);
     }
 
+    /// <summary>
+    /// Replaces each kept phrase with one private use char (not a word char, so no rule can
+    /// match it or across it). Skipped for text that already contains such chars.
+    /// </summary>
+    private List<string>? MaskKeepPhrases(ref string text)
+    {
+        if (_keepRegex == null || ContainsPlaceholderChar(text))
+        {
+            return null;
+        }
+
+        var kept = new List<string>();
+        text = _keepRegex.Replace(text, m =>
+        {
+            if (kept.Count > PlaceholderLast - PlaceholderFirst)
+            {
+                return m.Value;
+            }
+
+            kept.Add(m.Value);
+            return ((char)(PlaceholderFirst + kept.Count - 1)).ToString();
+        });
+
+        return kept.Count > 0 ? kept : null;
+    }
+
+    private static string UnmaskKeepPhrases(string text, List<string>? kept)
+    {
+        if (kept == null)
+        {
+            return text;
+        }
+
+        var sb = new StringBuilder(text.Length + 16);
+        foreach (var c in text)
+        {
+            var index = c - PlaceholderFirst;
+            if (index >= 0 && index < kept.Count)
+            {
+                sb.Append(kept[index]);
+            }
+            else
+            {
+                sb.Append(c);
+            }
+        }
+
+        return sb.ToString();
+    }
+
+    private static bool ContainsPlaceholderChar(string text)
+    {
+        foreach (var c in text)
+        {
+            if (c >= PlaceholderFirst && c <= PlaceholderLast)
+            {
+                return true;
+            }
+        }
+
+        return false;
+    }
+
     public bool TryConvert(string text, out string converted)
     {
         converted = Convert(text);
@@ -92,6 +180,15 @@ private void LoadBuiltInWordList(HashSet<string> ignored)
             return;
         }
 
+        foreach (var element in xml.Root.Elements("Keep"))
+        {
+            var keep = element.Attribute("text")?.Value.Trim();
+            if (!string.IsNullOrEmpty(keep))
+            {
+                _keepPhrases.Add(keep!);
+            }
+        }
+
         foreach (var element in xml.Root.Elements("Word"))
         {
             var us = element.Attribute("us")?.Value;
diff --git a/se5/Plugin-Shared/WordList.xml b/se5/Plugin-Shared/WordList.xml
index f3474872..7c563c02 100644
--- a/se5/Plugin-Shared/WordList.xml
+++ b/se5/Plugin-Shared/WordList.xml
@@ -1,5 +1,8 @@
 <?xml version="1.0" encoding="utf-8"?>
 <Words>
+  <!-- Keep: phrases that are never converted (foreign words that look like American spellings). -->
+  <Keep text="por favor" />
+  <Keep text="faz favor" />
   <Word us="accouterment" br="accoutrement" />
   <Word us="accouterments" br="accoutrements" />
   <Word us="aerogram" br="aerogramme" />