Handle Docstrings as bytes + strip all whitespace (#2037)

[etc/vim.git] / src / black / __init__.py

diff --git a/src/black/__init__.py b/src/black/__init__.py
index 2bb42a470eadb617b3cb57c4ee4b9c0c76377d40..5c9ab7516508b78ee23eebb49fbe934912548554 100644 (file)
--- a/src/black/__init__.py
+++ b/src/black/__init__.py
@@ -784,7 +784,7 @@ def reformat_many(
     except (ImportError, OSError):
         # we arrive here if the underlying system does not support multi-processing
         # like in AWS Lambda or Termux, in which case we gracefully fallback to
-        # a ThreadPollExecutor with just a single worker (more workers would not do us
+        # a ThreadPoolExecutor with just a single worker (more workers would not do us
         # any good due to the Global Interpreter Lock)
         executor = ThreadPoolExecutor(max_workers=1)
 
@@ -2153,16 +2153,35 @@ class LineGenerator(Visitor[Line]):
             # We're ignoring docstrings with backslash newline escapes because changing
             # indentation of those changes the AST representation of the code.
             prefix = get_string_prefix(leaf.value)
-            lead_len = len(prefix) + 3
-            tail_len = -3
-            indent = " " * 4 * self.current_line.depth
-            docstring = fix_docstring(leaf.value[lead_len:tail_len], indent)
+            docstring = leaf.value[len(prefix) :]  # Remove the prefix
+            quote_char = docstring[0]
+            # A natural way to remove the outer quotes is to do:
+            #   docstring = docstring.strip(quote_char)
+            # but that breaks on """""x""" (which is '""x').
+            # So we actually need to remove the first character and the next two
+            # characters but only if they are the same as the first.
+            quote_len = 1 if docstring[1] != quote_char else 3
+            docstring = docstring[quote_len:-quote_len]
+
+            if is_multiline_string(leaf):
+                indent = " " * 4 * self.current_line.depth
+                docstring = fix_docstring(docstring, indent)
+            else:
+                docstring = docstring.strip()
+
             if docstring:
-                if leaf.value[lead_len - 1] == docstring[0]:
+                # Add some padding if the docstring starts / ends with a quote mark.
+                if docstring[0] == quote_char:
                     docstring = " " + docstring
-                if leaf.value[tail_len + 1] == docstring[-1]:
+                if docstring[-1] == quote_char:
                     docstring = docstring + " "
-            leaf.value = leaf.value[0:lead_len] + docstring + leaf.value[tail_len:]
+            else:
+                # Add some padding if the docstring is empty.
+                docstring = " "
+
+            # We could enforce triple quotes at this point.
+            quote = quote_char * quote_len
+            leaf.value = prefix + quote + docstring + quote
 
         yield from self.visit_default(leaf)
 
@@ -2709,6 +2728,13 @@ def make_comment(content: str) -> str:
 
     if content[0] == "#":
         content = content[1:]
+    NON_BREAKING_SPACE = " "
+    if (
+        content
+        and content[0] == NON_BREAKING_SPACE
+        and not content.lstrip().startswith("type:")
+    ):
+        content = " " + content[1:]  # Replace NBSP by a simple space
     if content and content[0] not in " !:#'%":
         content = " " + content
     return "#" + content
@@ -6106,7 +6132,7 @@ def get_future_imports(node: Node) -> Set[str]:
 
 @lru_cache()
 def get_gitignore(root: Path) -> PathSpec:
-    """ Return a PathSpec matching gitignore content if present."""
+    """Return a PathSpec matching gitignore content if present."""
     gitignore = root / ".gitignore"
     lines: List[str] = []
     if gitignore.is_file():
@@ -6440,12 +6466,22 @@ def _stringify_ast(
             # Constant strings may be indented across newlines, if they are
             # docstrings; fold spaces after newlines when comparing. Similarly,
             # trailing and leading space may be removed.
+            # Note that when formatting Python 2 code, at least with Windows
+            # line-endings, docstrings can end up here as bytes instead of
+            # str so make sure that we handle both cases.
             if (
                 isinstance(node, ast.Constant)
                 and field == "value"
-                and isinstance(value, str)
+                and isinstance(value, (str, bytes))
             ):
-                normalized = re.sub(r" *\n[ \t]*", "\n", value).strip()
+                lineend = "\n" if isinstance(value, str) else b"\n"
+                # To normalize, we strip any leading and trailing space from
+                # each line...
+                stripped = [line.strip() for line in value.splitlines()]
+                normalized = lineend.join(stripped)  # type: ignore[attr-defined]
+                # ...and remove any blank lines at the beginning and end of
+                # the whole string
+                normalized = normalized.strip()
             else:
                 normalized = value
             yield f"{'  ' * (depth+2)}{normalized!r},  # {value.__class__.__name__}"
@@ -6946,11 +6982,6 @@ def patched_main() -> None:
 
 
 def is_docstring(leaf: Leaf) -> bool:
-    if not is_multiline_string(leaf):
-        # For the purposes of docstring re-indentation, we don't need to do anything
-        # with single-line docstrings.
-        return False
-
     if prev_siblings_are(
         leaf.parent, [None, token.NEWLINE, token.INDENT, syms.simple_stmt]
     ):