mformat: sort files() arguments naturally

Author: bgilbert

docs/markdown/Commands.md

@@ -473,7 +473,9 @@ The following options are recognized:
 - tab_width (int): Width of tab stops, used to compute line length
     when `indent_by` uses tab characters (default is 4).
 - sort_files (bool): When true, arguments of `files()` function are
-    sorted alphabetically (default is true).
+    sorted (default is true). *Since 1.10.0*, arguments are sorted
+    [naturally](Style-guide.md#sorting-source-paths) rather than
+    alphabetically.
 - group_arg_value (bool): When true, string argument with `--` prefix
     followed by string argument without `--` prefix are grouped on the
     same line, in multiline arguments (default is false).

docs/markdown/snippets/meson-format-sort.md

@@ -0,0 +1,5 @@
+## `meson format` sorts `files()` arguments naturally
+
+If the `sort_files` option is enabled, as it is by default, `meson format`
+now sorts `files()` arguments [naturally](Style-guide.md#sorting-source-paths)
+rather than alphabetically.

mesonbuild/mformat.py

@@ -12,7 +12,7 @@
 import sys
 
 from . import mparser
-from .mesonlib import MesonException
+from .mesonlib import MesonException, pathname_sort_key
 from .ast.postprocess import AstConditionLevel
 from .ast.printer import RawPrinter
 from .ast.visitor import FullAstVisitor
@@ -301,13 +301,12 @@ def dedent(self, value: str) -> str:
         return value
 
     def sort_arguments(self, node: mparser.ArgumentNode) -> None:
-        # TODO: natsort
         def sort_key(arg: mparser.BaseNode) -> str:
             if isinstance(arg, mparser.StringNode):
                 return arg.raw_value
             return getattr(node, 'value', '')
 
-        node.arguments.sort(key=sort_key)
+        node.arguments.sort(key=lambda arg: pathname_sort_key(sort_key(arg)))
 
     def visit_EmptyNode(self, node: mparser.EmptyNode) -> None:
         self.enter_node(node)
@@ -1064,5 +1063,5 @@ def run(options: argparse.Namespace) -> int:
 # - Option to simplify string literals
 # - Option to recognize and parse meson.build in subdirs
 # - Correctly compute line length when using tabs
-# - By default, arguments in files() are sorted alphabetically
+# - By default, arguments in files() are sorted naturally
 # - Option to group '--arg', 'value' on same line in multiline arguments

mesonbuild/rewriter.py

@@ -13,7 +13,7 @@
 from .ast.interpreter import IntrospectionBuildTarget, IntrospectionDependency, _symbol
 from .interpreterbase import UnknownValue, TV_func
 from .interpreterbase.helpers import flatten
-from mesonbuild.mesonlib import MesonException, setup_vsenv, relpath
+from mesonbuild.mesonlib import MesonException, pathname_sort_key, relpath, setup_vsenv
 from . import mlog, environment
 from functools import wraps
 from .mparser import Token, ArrayNode, ArgumentNode, ArithmeticNode, AssignmentNode, BaseNode, StringNode, BooleanNode, ElementaryNode, IdNode, FunctionNode, PlusAssignmentNode
@@ -951,15 +951,6 @@ def rel_source(src: str) -> str:
 
         # Sort files
         for i in to_sort_nodes:
-            def convert(text: str) -> T.Union[int, str]:
-                return int(text) if text.isdigit() else text.lower()
-
-            def alphanum_key(key: str) -> T.List[T.Union[int, str]]:
-                return [convert(c) for c in re.split('([0-9]+)', key)]
-
-            def path_sorter(key: str) -> T.List[T.Tuple[bool, T.List[T.Union[int, str]]]]:
-                return [(key.count('/') <= idx, alphanum_key(x)) for idx, x in enumerate(key.split('/'))]
-
             if isinstance(i, FunctionNode) and i.func_name.value in BUILD_TARGET_FUNCTIONS:
                 src_args = i.args.arguments[1:]
                 target_name = [i.args.arguments[0]]
@@ -968,7 +959,7 @@ def path_sorter(key: str) -> T.List[T.Tuple[bool, T.List[T.Union[int, str]]]]:
                 target_name = []
             unknown: T.List[BaseNode] = [x for x in src_args if not isinstance(x, StringNode)]
             sources: T.List[StringNode] = [x for x in src_args if isinstance(x, StringNode)]
-            sources = sorted(sources, key=lambda x: path_sorter(x.value))
+            sources = sorted(sources, key=lambda x: pathname_sort_key(x.value))
             i.args.arguments = target_name + unknown + T.cast(T.List[BaseNode], sources)
 
     def process(self, cmd: T.Dict[str, T.Any]) -> None:

mesonbuild/utils/universal.py

@@ -138,6 +138,7 @@ class _VerPickleLoadable(Protocol):
     'listify_array_value',
     'partition',
     'path_is_in_root',
+    'pathname_sort_key',
     'pickle_load',
     'Popen_safe',
     'Popen_safe_logged',
@@ -2494,3 +2495,17 @@ def __get__(self, instance: object, cls: T.Type) -> _T:
         value = self.__func(instance)
         setattr(instance, self.__name, value)
         return value
+
+
+def pathname_sort_key(key: str) -> tuple[tuple[bool, tuple[int | str, ...]], ...]:
+    '''Sort key for natural pathname sort, as defined in the Meson style guide.
+    Use as the key= argument to sort() or sorted().'''
+
+    def convert(text: str) -> int | str:
+        return int(text) if text.isdigit() else text.lower()
+
+    def alphanum_key(key: str) -> tuple[int | str, ...]:
+        return tuple(convert(c) for c in re.split('([0-9]+)', key))
+
+    return tuple((key.count('/') <= idx, alphanum_key(x))
+                 for idx, x in enumerate(key.split('/')))

test cases/format/1 default/meson.build

@@ -8,6 +8,7 @@ meson_files = {
     'comments': files('crazy_comments.meson'),
     'indentation': files('indentation.meson'),
     'gh13242': files('gh13242.meson'),
+    'sort': files('sort_files.meson'),
 }
 
 # Ensure empty function are formatted correctly on long lines

test cases/format/1 default/sort_files.meson

@@ -0,0 +1,19 @@
+# Based on the example in the Meson style guide
+
+sources = files(
+    'aaa/a1.c',
+    'aaa/A2.c',
+    'aaa/a3.c',
+    'bbb/subdir1/b1.c',
+    'bbb/subdir2/b2.c',
+    'bbb/subdir10/b3.c',
+    'bbb/subdir20/b4.c',
+    'bbb/b5.c',
+    'bbb/b6.c',
+    'c/a111.c',
+    'c/ab0.c',
+    'f1.c',
+    'f2.c',
+    'f10.c',
+    'f20.c',
+)