Add global "locale" config option and make key-ordering rule locale-aware

Support sorting by locale with strcoll(). Properly handle case and accents.

Note: this is a second implementation, for context see:
https://github.com/adrienverge/yamllint/pull/280
https://github.com/adrienverge/yamllint/issues/285
https://github.com/adrienverge/yamllint/pull/288

yamllint/cli.py

@@ -18,6 +18,7 @@ from __future__ import print_function
 
 import argparse
 import io
+import locale
 import os
 import platform
 import sys
@@ -175,6 +176,9 @@ def run(argv=None):
         print(e, file=sys.stderr)
         sys.exit(-1)
 
+    if conf.locale is not None:
+        locale.setlocale(locale.LC_ALL, conf.locale)
+
     max_level = 0
 
     for file in find_files_recursively(args.files, conf):

yamllint/config.py

@@ -35,6 +35,8 @@ class YamlLintConfig(object):
         self.yaml_files = pathspec.PathSpec.from_lines(
             'gitwildmatch', ['*.yaml', '*.yml', '.yamllint'])
 
+        self.locale = None
+
         if file is not None:
             with open(file) as f:
                 content = f.read()
@@ -111,6 +113,12 @@ class YamlLintConfig(object):
             self.yaml_files = pathspec.PathSpec.from_lines('gitwildmatch',
                                                            conf['yaml-files'])
 
+        if 'locale' in conf:
+            if not isinstance(conf['locale'], str):
+                raise YamlLintConfigError(
+                    'invalid config: locale should be a string')
+            self.locale = conf['locale']
+
     def validate(self):
         for id in self.rules:
             try:

yamllint/rules/key_ordering.py

@@ -16,8 +16,10 @@
 
 """
 Use this rule to enforce alphabetical ordering of keys in mappings. The sorting
-order uses the Unicode code point number. As a result, the ordering is
-case-sensitive and not accent-friendly (see examples below).
+order uses the Unicode code point number as a default. As a result, the
+ordering is case-sensitive and not accent-friendly (see examples below).
+This can be changed by setting the global ``locale`` option.  This allows to
+sort case and accents properly.
 
 .. rubric:: Examples
 
@@ -63,8 +65,24 @@ case-sensitive and not accent-friendly (see examples below).
 
     - haïr: true
       hais: true
+
+#. With global option ``locale: "en_US.UTF-8"`` and rule ``key-ordering: {}``
+
+   as opposed to before, the following code snippet would now **PASS**:
+   ::
+
+    - t-shirt: 1
+      T-shirt: 2
+      t-shirts: 3
+      T-shirts: 4
+    - hair: true
+      haïr: true
+      hais: true
+      haïssable: true
 """
 
+from locale import strcoll
+
 import yaml
 
 from yamllint.linter import LintProblem
@@ -101,7 +119,8 @@ def check(conf, token, prev, next, nextnext, context):
         # This check is done because KeyTokens can be found inside flow
         # sequences... strange, but allowed.
         if len(context['stack']) > 0 and context['stack'][-1].type == MAP:
-            if any(next.value < key for key in context['stack'][-1].keys):
+            if any(strcoll(next.value, key) < 0
+                   for key in context['stack'][-1].keys):
                 yield LintProblem(
                     next.start_mark.line + 1, next.start_mark.column + 1,
                     'wrong ordering of key "%s" in mapping' % next.value)