diff --git a/src/libcalamares/locale/cldr-extractor.py b/src/libcalamares/locale/cldr-extractor.py
new file mode 100644
index 000000000..6c4061fcc
--- /dev/null
+++ b/src/libcalamares/locale/cldr-extractor.py
@@ -0,0 +1,140 @@
+#! /usr/bin/env python3
+#
+# Python3 script to scrape some data out of ICU CLDR supplemental data.
+#
+# To use this script, you must have downloaded the CLDR data, e.g.
+# http://unicode.org/Public/cldr/35.1/, and extracted the zip file.
+# Run the script from **inside** the common/ durectory that is created
+# (or fix the hard-coded path).
+#
+# The script tries to print C++ code that compiles; if there are encoding
+# problems, it will print some kind of representation of the problematic
+# lines.
+#
+# To avoid having to cross-reference multiple XML files, the script
+# cheats: it reads the comments as well to get names. So it looks for
+# pairs of lines like this:
+#
+#    <likelySubtag from="und_BQ" to="pap_Latn_BQ"/>
+#    <!--{ ?; ?; Caribbean Netherlands } => { Papiamento; Latin; Caribbean Netherlands }-->
+#
+# It extracts the 2-character country code "BQ" from the sub-tag, and
+# parses the comment to get a language and country name (instead of looking up
+# "pap" and "BQ" in other tables). This may be considered a hack.
+#
+# A large collection of exceptions can be found in the two *_mapper tables,
+# which massage the CLDR names to Qt enum values.
+
+
+import sys
+
+# These are languages listed in CLDR that don't match
+# the enum-values in QLocale::Language.
+language_mapper = {
+    "?" : "AnyLanguage",
+    "Bangla" : "Bengali",
+    "Kalaallisut" : "Greenlandic",
+    "Haitian Creole" : "Haitian",
+    "Kyrgyz" : "Kirghiz",
+    "Norwegian Bokmål" : "NorwegianBokmal",
+    "Tokelau" : "TokelauLanguage",
+    "Tuvalu" : "TuvaluLanguage",
+    }
+
+country_mapper = {
+    "Åland Islands" : "AlandIslands",
+    "St. Barthélemy" : "SaintBarthelemy",
+    "Côte d’Ivoire" : "IvoryCoast",
+    "Curaçao" : "CuraSao",
+    "Réunion" : "Reunion",
+    "São Tomé & Príncipe" : "SaoTomeAndPrincipe",
+    "Bosnia & Herzegovina" : "BosniaAndHerzegowina",
+    "Czechia" : "CzechRepublic",
+    "St. Pierre & Miquelon" : "SaintPierreAndMiquelon",
+    "Vatican City" : "VaticanCityState",
+    "South Georgia & South Sandwich Islands" : "SouthGeorgiaAndTheSouthSandwichIslands",
+    "Timor-Leste" : "EastTimor",
+    "Wallis & Futuna" : "WallisAndFutunaIslands",
+    "Myanmar (Burma)" : "Myanmar",
+    "Svalbard & Jan Mayen" : "SvalbardAndJanMayenIslands",
+    "St. Martin" : "SaintMartin",
+    "North Macedonia" : "Macedonia",
+    "Hong Kong SAR China" : "HongKong",
+    "Macao SAR China" : "Macau",
+    "Eurozone" : "AnyCountry",  # Not likely for GeoIP
+    "Caribbean Netherlands" : "Bonaire",  # Bonaire, Saba, St.Eustatius
+    }
+    
+def extricate(l1, l2):
+    if '"und_' not in l1:
+        return
+    if '{ ?; ?;' not in l2:
+        return
+    
+    # This is extremely crude "parsing" which chops up the string
+    # by delimiter and then extracts some substring.
+    l1_parts = l1.split("und_")
+    l2_parts = l2.split(";")
+    
+    l1_first_quote = l1_parts[1].find('"')
+    l1_code = l1_parts[1][:l1_first_quote]
+    if len(l1_code) != 2:
+        return
+    
+    l2_brace = l2_parts[2].find("{")
+    l2_language = l2_parts[2][l2_brace+1:].strip()
+    l2_brace = l2_parts[2].find("}")
+    l2_country = l2_parts[2][:l2_brace-1].strip()
+    
+    # Handle mapped cases
+    l2_language = language_mapper.get(l2_language, l2_language)
+    l2_language = l2_language.replace(" ", "")
+    
+    # Handle mapped cases and then do a bunch of standard replacements.
+    l2_country = country_mapper.get(l2_country, l2_country)
+    l2_country = l2_country.replace(" ", "").replace("-", "").replace(".","").replace("&","And")
+      
+    # There shouldn't be any UTF-8 left in there.
+    try:
+        print("{!s} QLocale::Language::{!s}, QLocale::Country::{!s}, '{!s}', '{!s}' {!s},".format(
+            "{",
+            l2_language, 
+            l2_country,
+            l1_code[0], 
+            l1_code[1],
+            "}"))
+    except UnicodeEncodeError:
+        print(list(map(lambda x : '?' if x > 128 else chr(x), map(lambda x:ord(x), l2_country))))
+        raise
+
+print("""// Generated from CLDR data
+#include <QLocale>
+struct CountryData
+{
+    QLocale::Language l;
+    QLocale::Country c;
+    char cc1;
+    char cc2;
+};
+
+static const CountryData countryMap[] = {
+""")
+
+with open("supplemental/likelySubtags.xml", "rt", encoding="UTF-8") as f:
+    l1 = "a line"
+    while l1:
+        l1 = f.readline()
+        if '<likelySubtag from="und_' not in l1:
+            continue
+        l2 = f.readline()
+
+        if l1:
+            assert "likelySubtag" in l1, l1;
+            assert "<!--" in l2, l2;
+            
+            extricate(l1, l2)
+        
+print("""{ QLocale::Language::AnyLanguage, QLocale::Country::AnyCountry, 0, 0 } // Terminator
+};
+// END Generated from CLDR data
+""")