]> git.madduck.net Git - etc/vim.git/blobdiff - black.py

madduck's git repository

Every one of the projects in this repository is available at the canonical URL git://git.madduck.net/madduck/pub/<projectpath> — see each project's metadata for the exact URL.

All patches and comments are welcome. Please squash your changes to logical commits before using git-format-patch and git-send-email to patches@git.madduck.net. If you'd read over the Git project's submission guidelines and adhered to them, I'd be especially grateful.

SSH access, as well as push access can be individually arranged.

If you use my repositories frequently, consider adding the following snippet to ~/.gitconfig and using the third clone URL listed for each project:

[url "git://git.madduck.net/madduck/"]
  insteadOf = madduck:

Revert `format_file_in_place()` and `format_stdin_to_stdout()` to return bools
[etc/vim.git] / black.py
index 774d91dc2a7f64ec2dbc252a55e0fe051d5c429c..e09b72380fbd60950dcc94253c61a6c8cd526d15 100644 (file)
--- a/black.py
+++ b/black.py
@@ -1,17 +1,40 @@
 #!/usr/bin/env python3
+
 import asyncio
+import pickle
 from asyncio.base_events import BaseEventLoop
 from concurrent.futures import Executor, ProcessPoolExecutor
-from functools import partial
+from enum import Enum
+from functools import partial, wraps
 import keyword
+import logging
+from multiprocessing import Manager
 import os
 from pathlib import Path
+import re
 import tokenize
+import signal
+import sys
 from typing import (
-    Dict, Generic, Iterable, Iterator, List, Optional, Set, Tuple, TypeVar, Union
+    Any,
+    Callable,
+    Collection,
+    Dict,
+    Generic,
+    Iterable,
+    Iterator,
+    List,
+    Optional,
+    Pattern,
+    Set,
+    Tuple,
+    Type,
+    TypeVar,
+    Union,
 )
 
-from attr import attrib, dataclass, Factory
+from appdirs import user_cache_dir
+from attr import dataclass, Factory
 import click
 
 # lib2to3 fork
@@ -20,7 +43,7 @@ from blib2to3 import pygram, pytree
 from blib2to3.pgen2 import driver, token
 from blib2to3.pgen2.parse import ParseError
 
-__version__ = "18.3a1"
+__version__ = "18.4a2"
 DEFAULT_LINE_LENGTH = 88
 # types
 syms = pygram.python_symbols
@@ -30,54 +53,123 @@ Depth = int
 NodeType = int
 LeafID = int
 Priority = int
+Index = int
 LN = Union[Leaf, Node]
+SplitFunc = Callable[["Line", bool], Iterator["Line"]]
+Timestamp = float
+FileSize = int
+CacheInfo = Tuple[Timestamp, FileSize]
+Cache = Dict[Path, CacheInfo]
 out = partial(click.secho, bold=True, err=True)
-err = partial(click.secho, fg='red', err=True)
+err = partial(click.secho, fg="red", err=True)
 
 
 class NothingChanged(UserWarning):
-    """Raised by `format_file` when the reformatted code is the same as source."""
+    """Raised by :func:`format_file` when reformatted code is the same as source."""
 
 
 class CannotSplit(Exception):
     """A readable split that fits the allotted line length is impossible.
 
-    Raised by `left_hand_split()` and `right_hand_split()`.
+    Raised by :func:`left_hand_split`, :func:`right_hand_split`, and
+    :func:`delimiter_split`.
+    """
+
+
+class FormatError(Exception):
+    """Base exception for `# fmt: on` and `# fmt: off` handling.
+
+    It holds the number of bytes of the prefix consumed before the format
+    control comment appeared.
     """
 
+    def __init__(self, consumed: int) -> None:
+        super().__init__(consumed)
+        self.consumed = consumed
+
+    def trim_prefix(self, leaf: Leaf) -> None:
+        leaf.prefix = leaf.prefix[self.consumed:]
+
+    def leaf_from_consumed(self, leaf: Leaf) -> Leaf:
+        """Returns a new Leaf from the consumed part of the prefix."""
+        unformatted_prefix = leaf.prefix[:self.consumed]
+        return Leaf(token.NEWLINE, unformatted_prefix)
+
+
+class FormatOn(FormatError):
+    """Found a comment like `# fmt: on` in the file."""
+
+
+class FormatOff(FormatError):
+    """Found a comment like `# fmt: off` in the file."""
+
+
+class WriteBack(Enum):
+    NO = 0
+    YES = 1
+    DIFF = 2
+
+
+class Changed(Enum):
+    NO = 0
+    CACHED = 1
+    YES = 2
+
 
 @click.command()
 @click.option(
-    '-l',
-    '--line-length',
+    "-l",
+    "--line-length",
     type=int,
     default=DEFAULT_LINE_LENGTH,
-    help='How many character per line to allow.',
+    help="How many character per line to allow.",
     show_default=True,
 )
 @click.option(
-    '--check',
+    "--check",
     is_flag=True,
     help=(
-        "Don't write back the files, just return the status.  Return code 0 "
-        "means nothing changed.  Return code 1 means some files were "
+        "Don't write the files back, just return the status.  Return code 0 "
+        "means nothing would change.  Return code 1 means some files would be "
         "reformatted.  Return code 123 means there was an internal error."
     ),
 )
 @click.option(
-    '--fast/--safe',
+    "--diff",
+    is_flag=True,
+    help="Don't write the files back, just output a diff for each file on stdout.",
+)
+@click.option(
+    "--fast/--safe",
+    is_flag=True,
+    help="If --fast given, skip temporary sanity checks. [default: --safe]",
+)
+@click.option(
+    "-q",
+    "--quiet",
     is_flag=True,
-    help='If --fast given, skip temporary sanity checks. [default: --safe]',
+    help=(
+        "Don't emit non-error messages to stderr. Errors are still emitted, "
+        "silence those with 2>/dev/null."
+    ),
 )
 @click.version_option(version=__version__)
 @click.argument(
-    'src',
+    "src",
     nargs=-1,
-    type=click.Path(exists=True, file_okay=True, dir_okay=True, readable=True),
+    type=click.Path(
+        exists=True, file_okay=True, dir_okay=True, readable=True, allow_dash=True
+    ),
 )
 @click.pass_context
 def main(
-    ctx: click.Context, line_length: int, check: bool, fast: bool, src: List[str]
+    ctx: click.Context,
+    line_length: int,
+    check: bool,
+    diff: bool,
+    fast: bool,
+    quiet: bool,
+    src: List[str],
 ) -> None:
     """The uncompromising code formatter."""
     sources: List[Path] = []
@@ -88,110 +180,258 @@ def main(
         elif p.is_file():
             # if a file was explicitly given, we don't care about its extension
             sources.append(p)
+        elif s == "-":
+            sources.append(Path("-"))
         else:
-            err(f'invalid path: {s}')
+            err(f"invalid path: {s}")
+    if check and diff:
+        exc = click.ClickException("Options --check and --diff are mutually exclusive")
+        exc.exit_code = 2
+        raise exc
+
+    if check:
+        write_back = WriteBack.NO
+    elif diff:
+        write_back = WriteBack.DIFF
+    else:
+        write_back = WriteBack.YES
     if len(sources) == 0:
         ctx.exit(0)
+        return
+
     elif len(sources) == 1:
-        p = sources[0]
-        report = Report()
-        try:
-            changed = format_file_in_place(
-                p, line_length=line_length, fast=fast, write_back=not check
-            )
-            report.done(p, changed)
-        except Exception as exc:
-            report.failed(p, str(exc))
-        ctx.exit(report.return_code)
+        return_code = run_single_file_mode(
+            line_length, check, fast, quiet, write_back, sources[0]
+        )
     else:
-        loop = asyncio.get_event_loop()
-        executor = ProcessPoolExecutor(max_workers=os.cpu_count())
-        return_code = 1
-        try:
-            return_code = loop.run_until_complete(
-                schedule_formatting(
-                    sources, line_length, not check, fast, loop, executor
+        return_code = run_multi_file_mode(line_length, fast, quiet, write_back, sources)
+    ctx.exit(return_code)
+
+
+def run_single_file_mode(
+    line_length: int,
+    check: bool,
+    fast: bool,
+    quiet: bool,
+    write_back: WriteBack,
+    src: Path,
+) -> int:
+    report = Report(check=check, quiet=quiet)
+    try:
+        changed = Changed.NO
+        if not src.is_file() and str(src) == "-":
+            if format_stdin_to_stdout(
+                line_length=line_length, fast=fast, write_back=write_back
+            ):
+                changed = Changed.YES
+        else:
+            cache: Cache = {}
+            if write_back != WriteBack.DIFF:
+                cache = read_cache()
+                src = src.resolve()
+                if src in cache and cache[src] == get_cache_info(src):
+                    changed = Changed.CACHED
+            if (
+                changed is not Changed.CACHED
+                and format_file_in_place(
+                    src, line_length=line_length, fast=fast, write_back=write_back
                 )
+            ):
+                changed = Changed.YES
+            if write_back != WriteBack.DIFF and changed is not Changed.NO:
+                write_cache(cache, [src])
+        report.done(src, changed)
+    except Exception as exc:
+        report.failed(src, str(exc))
+    return report.return_code
+
+
+def run_multi_file_mode(
+    line_length: int,
+    fast: bool,
+    quiet: bool,
+    write_back: WriteBack,
+    sources: List[Path],
+) -> int:
+    loop = asyncio.get_event_loop()
+    executor = ProcessPoolExecutor(max_workers=os.cpu_count())
+    return_code = 1
+    try:
+        return_code = loop.run_until_complete(
+            schedule_formatting(
+                sources, line_length, write_back, fast, quiet, loop, executor
             )
-        finally:
-            loop.close()
-            ctx.exit(return_code)
+        )
+    finally:
+        shutdown(loop)
+        return return_code
 
 
 async def schedule_formatting(
     sources: List[Path],
     line_length: int,
-    write_back: bool,
+    write_back: WriteBack,
     fast: bool,
+    quiet: bool,
     loop: BaseEventLoop,
     executor: Executor,
 ) -> int:
-    tasks = {
-        src: loop.run_in_executor(
-            executor, format_file_in_place, src, line_length, fast, write_back
-        )
-        for src in sources
-    }
-    await asyncio.wait(tasks.values())
+    """Run formatting of `sources` in parallel using the provided `executor`.
+
+    (Use ProcessPoolExecutors for actual parallelism.)
+
+    `line_length`, `write_back`, and `fast` options are passed to
+    :func:`format_file_in_place`.
+    """
+    report = Report(check=write_back is WriteBack.NO, quiet=quiet)
+    cache: Cache = {}
+    if write_back != WriteBack.DIFF:
+        cache = read_cache()
+        sources, cached = filter_cached(cache, sources)
+        for src in cached:
+            report.done(src, Changed.CACHED)
     cancelled = []
-    report = Report()
-    for src, task in tasks.items():
-        if not task.done():
-            report.failed(src, 'timed out, cancelling')
-            task.cancel()
-            cancelled.append(task)
-        elif task.exception():
-            report.failed(src, str(task.exception()))
-        else:
-            report.done(src, task.result())
+    formatted = []
+    if sources:
+        lock = None
+        if write_back == WriteBack.DIFF:
+            # For diff output, we need locks to ensure we don't interleave output
+            # from different processes.
+            manager = Manager()
+            lock = manager.Lock()
+        tasks = {
+            src: loop.run_in_executor(
+                executor, format_file_in_place, src, line_length, fast, write_back, lock
+            )
+            for src in sources
+        }
+        _task_values = list(tasks.values())
+        loop.add_signal_handler(signal.SIGINT, cancel, _task_values)
+        loop.add_signal_handler(signal.SIGTERM, cancel, _task_values)
+        await asyncio.wait(_task_values)
+        for src, task in tasks.items():
+            if not task.done():
+                report.failed(src, "timed out, cancelling")
+                task.cancel()
+                cancelled.append(task)
+            elif task.cancelled():
+                cancelled.append(task)
+            elif task.exception():
+                report.failed(src, str(task.exception()))
+            else:
+                formatted.append(src)
+                report.done(src, Changed.YES if task.result() else Changed.NO)
+
     if cancelled:
-        await asyncio.wait(cancelled, timeout=2)
-    out('All done! ✨ 🍰 ✨')
-    click.echo(str(report))
+        await asyncio.gather(*cancelled, loop=loop, return_exceptions=True)
+    elif not quiet:
+        out("All done! ✨ 🍰 ✨")
+    if not quiet:
+        click.echo(str(report))
+
+    if write_back != WriteBack.DIFF and formatted:
+        write_cache(cache, formatted)
+
     return report.return_code
 
 
 def format_file_in_place(
-    src: Path, line_length: int, fast: bool, write_back: bool = False
+    src: Path,
+    line_length: int,
+    fast: bool,
+    write_back: WriteBack = WriteBack.NO,
+    lock: Any = None,  # multiprocessing.Manager().Lock() is some crazy proxy
 ) -> bool:
-    """Format the file and rewrite if changed. Return True if changed."""
+    """Format file under `src` path. Return True if changed.
+
+    If `write_back` is True, write reformatted code back to stdout.
+    `line_length` and `fast` options are passed to :func:`format_file_contents`.
+    """
+
+    with tokenize.open(src) as src_buffer:
+        src_contents = src_buffer.read()
     try:
-        contents, encoding = format_file(src, line_length=line_length, fast=fast)
+        dst_contents = format_file_contents(
+            src_contents, line_length=line_length, fast=fast
+        )
     except NothingChanged:
         return False
 
-    if write_back:
-        with open(src, "w", encoding=encoding) as f:
-            f.write(contents)
+    if write_back == write_back.YES:
+        with open(src, "w", encoding=src_buffer.encoding) as f:
+            f.write(dst_contents)
+    elif write_back == write_back.DIFF:
+        src_name = f"{src.name}  (original)"
+        dst_name = f"{src.name}  (formatted)"
+        diff_contents = diff(src_contents, dst_contents, src_name, dst_name)
+        if lock:
+            lock.acquire()
+        try:
+            sys.stdout.write(diff_contents)
+        finally:
+            if lock:
+                lock.release()
     return True
 
 
-def format_file(
-    src: Path, line_length: int, fast: bool
-) -> Tuple[FileContent, Encoding]:
-    """Reformats a file and returns its contents and encoding."""
-    with tokenize.open(src) as src_buffer:
-        src_contents = src_buffer.read()
-    if src_contents.strip() == '':
-        raise NothingChanged(src)
+def format_stdin_to_stdout(
+    line_length: int, fast: bool, write_back: WriteBack = WriteBack.NO
+) -> bool:
+    """Format file on stdin. Return True if changed.
+
+    If `write_back` is True, write reformatted code back to stdout.
+    `line_length` and `fast` arguments are passed to :func:`format_file_contents`.
+    """
+    src = sys.stdin.read()
+    dst = src
+    try:
+        dst = format_file_contents(src, line_length=line_length, fast=fast)
+        return True
+
+    except NothingChanged:
+        return False
+
+    finally:
+        if write_back == WriteBack.YES:
+            sys.stdout.write(dst)
+        elif write_back == WriteBack.DIFF:
+            src_name = "<stdin>  (original)"
+            dst_name = "<stdin>  (formatted)"
+            sys.stdout.write(diff(src, dst, src_name, dst_name))
+
+
+def format_file_contents(
+    src_contents: str, line_length: int, fast: bool
+) -> FileContent:
+    """Reformat contents a file and return new contents.
+
+    If `fast` is False, additionally confirm that the reformatted code is
+    valid by calling :func:`assert_equivalent` and :func:`assert_stable` on it.
+    `line_length` is passed to :func:`format_str`.
+    """
+    if src_contents.strip() == "":
+        raise NothingChanged
 
     dst_contents = format_str(src_contents, line_length=line_length)
     if src_contents == dst_contents:
-        raise NothingChanged(src)
+        raise NothingChanged
 
     if not fast:
         assert_equivalent(src_contents, dst_contents)
         assert_stable(src_contents, dst_contents, line_length=line_length)
-    return dst_contents, src_buffer.encoding
+    return dst_contents
 
 
 def format_str(src_contents: str, line_length: int) -> FileContent:
-    """Reformats a string and returns new contents."""
+    """Reformat a string and return new contents.
+
+    `line_length` determines how many characters per line are allowed.
+    """
     src_node = lib2to3_parse(src_contents)
     dst_contents = ""
-    comments: List[Line] = []
     lines = LineGenerator()
     elt = EmptyLineTracker()
+    py36 = is_python36(src_node)
     empty_line = Line()
     after = 0
     for current_line in lines.visit(src_node):
@@ -200,36 +440,41 @@ def format_str(src_contents: str, line_length: int) -> FileContent:
         before, after = elt.maybe_empty_lines(current_line)
         for _ in range(before):
             dst_contents += str(empty_line)
-        if not current_line.is_comment:
-            for comment in comments:
-                dst_contents += str(comment)
-            comments = []
-            for line in split_line(current_line, line_length=line_length):
-                dst_contents += str(line)
-        else:
-            comments.append(current_line)
-    for comment in comments:
-        dst_contents += str(comment)
+        for line in split_line(current_line, line_length=line_length, py36=py36):
+            dst_contents += str(line)
     return dst_contents
 
 
+GRAMMARS = [
+    pygram.python_grammar_no_print_statement_no_exec_statement,
+    pygram.python_grammar_no_print_statement,
+    pygram.python_grammar_no_exec_statement,
+    pygram.python_grammar,
+]
+
+
 def lib2to3_parse(src_txt: str) -> Node:
     """Given a string with source, return the lib2to3 Node."""
     grammar = pygram.python_grammar_no_print_statement
-    drv = driver.Driver(grammar, pytree.convert)
-    if src_txt[-1] != '\n':
-        nl = '\r\n' if '\r\n' in src_txt[:1024] else '\n'
+    if src_txt[-1] != "\n":
+        nl = "\r\n" if "\r\n" in src_txt[:1024] else "\n"
         src_txt += nl
-    try:
-        result = drv.parse_string(src_txt, True)
-    except ParseError as pe:
-        lineno, column = pe.context[1]
-        lines = src_txt.splitlines()
+    for grammar in GRAMMARS:
+        drv = driver.Driver(grammar, pytree.convert)
         try:
-            faulty_line = lines[lineno - 1]
-        except IndexError:
-            faulty_line = "<line number missing in source>"
-        raise ValueError(f"Cannot parse: {lineno}:{column}: {faulty_line}") from None
+            result = drv.parse_string(src_txt, True)
+            break
+
+        except ParseError as pe:
+            lineno, column = pe.context[1]
+            lines = src_txt.splitlines()
+            try:
+                faulty_line = lines[lineno - 1]
+            except IndexError:
+                faulty_line = "<line number missing in source>"
+            exc = ValueError(f"Cannot parse: {lineno}:{column}: {faulty_line}")
+    else:
+        raise exc from None
 
     if isinstance(result, Leaf):
         result = Node(syms.file_input, [result])
@@ -242,20 +487,30 @@ def lib2to3_unparse(node: Node) -> str:
     return code
 
 
-T = TypeVar('T')
+T = TypeVar("T")
 
 
 class Visitor(Generic[T]):
-    """Basic lib2to3 visitor that yields things on visiting."""
+    """Basic lib2to3 visitor that yields things of type `T` on `visit()`."""
 
     def visit(self, node: LN) -> Iterator[T]:
+        """Main method to visit `node` and its children.
+
+        It tries to find a `visit_*()` method for the given `node.type`, like
+        `visit_simple_stmt` for Node objects or `visit_INDENT` for Leaf objects.
+        If no dedicated `visit_*()` method is found, chooses `visit_default()`
+        instead.
+
+        Then yields objects of type `T` from the selected visitor.
+        """
         if node.type < 256:
             name = token.tok_name[node.type]
         else:
             name = type_repr(node.type)
-        yield from getattr(self, f'visit_{name}', self.visit_default)(node)
+        yield from getattr(self, f"visit_{name}", self.visit_default)(node)
 
     def visit_default(self, node: LN) -> Iterator[T]:
+        """Default `visit_*()` implementation. Recurses to children of `node`."""
         if isinstance(node, Node):
             for child in node.children:
                 yield from self.visit(child)
@@ -263,32 +518,41 @@ class Visitor(Generic[T]):
 
 @dataclass
 class DebugVisitor(Visitor[T]):
-    tree_depth: int = attrib(default=0)
+    tree_depth: int = 0
 
     def visit_default(self, node: LN) -> Iterator[T]:
-        indent = ' ' * (2 * self.tree_depth)
+        indent = " " * (2 * self.tree_depth)
         if isinstance(node, Node):
             _type = type_repr(node.type)
-            out(f'{indent}{_type}', fg='yellow')
+            out(f"{indent}{_type}", fg="yellow")
             self.tree_depth += 1
             for child in node.children:
                 yield from self.visit(child)
 
             self.tree_depth -= 1
-            out(f'{indent}/{_type}', fg='yellow', bold=False)
+            out(f"{indent}/{_type}", fg="yellow", bold=False)
         else:
             _type = token.tok_name.get(node.type, str(node.type))
-            out(f'{indent}{_type}', fg='blue', nl=False)
+            out(f"{indent}{_type}", fg="blue", nl=False)
             if node.prefix:
                 # We don't have to handle prefixes for `Node` objects since
                 # that delegates to the first child anyway.
-                out(f' {node.prefix!r}', fg='green', bold=False, nl=False)
-            out(f' {node.value!r}', fg='blue', bold=False)
+                out(f" {node.prefix!r}", fg="green", bold=False, nl=False)
+            out(f" {node.value!r}", fg="blue", bold=False)
+
+    @classmethod
+    def show(cls, code: str) -> None:
+        """Pretty-print the lib2to3 AST of a given string of `code`.
+
+        Convenience method for debugging.
+        """
+        v: DebugVisitor[None] = DebugVisitor()
+        list(v.visit(lib2to3_parse(code)))
 
 
 KEYWORDS = set(keyword.kwlist)
 WHITESPACE = {token.DEDENT, token.INDENT, token.NEWLINE}
-FLOW_CONTROL = {'return', 'raise', 'break', 'continue'}
+FLOW_CONTROL = {"return", "raise", "break", "continue"}
 STATEMENT = {
     syms.if_stmt,
     syms.while_stmt,
@@ -300,7 +564,7 @@ STATEMENT = {
     syms.classdef,
 }
 STANDALONE_COMMENT = 153
-LOGIC_OPERATORS = {'and', 'or'}
+LOGIC_OPERATORS = {"and", "or"}
 COMPARATORS = {
     token.LESS,
     token.GREATER,
@@ -318,11 +582,26 @@ MATH_OPERATORS = {
     token.AMPER,
     token.PERCENT,
     token.CIRCUMFLEX,
+    token.TILDE,
     token.LEFTSHIFT,
     token.RIGHTSHIFT,
     token.DOUBLESTAR,
     token.DOUBLESLASH,
 }
+STARS = {token.STAR, token.DOUBLESTAR}
+VARARGS_PARENTS = {
+    syms.arglist,
+    syms.argument,  # double star in arglist
+    syms.trailer,  # single argument to call
+    syms.typedargslist,
+    syms.varargslist,  # lambdas
+}
+UNPACKING_PARENTS = {
+    syms.atom,  # single element of a list or set literal
+    syms.dictsetmaker,
+    syms.listmaker,
+    syms.testlist_gexp,
+}
 COMPREHENSION_PRIORITY = 20
 COMMA_PRIORITY = 10
 LOGIC_PRIORITY = 5
@@ -333,70 +612,85 @@ MATH_PRIORITY = 1
 
 @dataclass
 class BracketTracker:
-    depth: int = attrib(default=0)
-    bracket_match: Dict[Tuple[Depth, NodeType], Leaf] = attrib(default=Factory(dict))
-    delimiters: Dict[LeafID, Priority] = attrib(default=Factory(dict))
-    previous: Optional[Leaf] = attrib(default=None)
+    """Keeps track of brackets on a line."""
+
+    depth: int = 0
+    bracket_match: Dict[Tuple[Depth, NodeType], Leaf] = Factory(dict)
+    delimiters: Dict[LeafID, Priority] = Factory(dict)
+    previous: Optional[Leaf] = None
 
     def mark(self, leaf: Leaf) -> None:
+        """Mark `leaf` with bracket-related metadata. Keep track of delimiters.
+
+        All leaves receive an int `bracket_depth` field that stores how deep
+        within brackets a given leaf is. 0 means there are no enclosing brackets
+        that started on this line.
+
+        If a leaf is itself a closing bracket, it receives an `opening_bracket`
+        field that it forms a pair with. This is a one-directional link to
+        avoid reference cycles.
+
+        If a leaf is a delimiter (a token on which Black can split the line if
+        needed) and it's on depth 0, its `id()` is stored in the tracker's
+        `delimiters` field.
+        """
         if leaf.type == token.COMMENT:
             return
 
         if leaf.type in CLOSING_BRACKETS:
             self.depth -= 1
             opening_bracket = self.bracket_match.pop((self.depth, leaf.type))
-            leaf.opening_bracket = opening_bracket  # type: ignore
-        leaf.bracket_depth = self.depth  # type: ignore
+            leaf.opening_bracket = opening_bracket
+        leaf.bracket_depth = self.depth
         if self.depth == 0:
-            delim = is_delimiter(leaf)
-            if delim:
-                self.delimiters[id(leaf)] = delim
-            elif self.previous is not None:
-                if leaf.type == token.STRING and self.previous.type == token.STRING:
-                    self.delimiters[id(self.previous)] = STRING_PRIORITY
-                elif (
-                    leaf.type == token.NAME and
-                    leaf.value == 'for' and
-                    leaf.parent and
-                    leaf.parent.type in {syms.comp_for, syms.old_comp_for}
-                ):
-                    self.delimiters[id(self.previous)] = COMPREHENSION_PRIORITY
-                elif (
-                    leaf.type == token.NAME and
-                    leaf.value == 'if' and
-                    leaf.parent and
-                    leaf.parent.type in {syms.comp_if, syms.old_comp_if}
-                ):
-                    self.delimiters[id(self.previous)] = COMPREHENSION_PRIORITY
+            delim = is_split_before_delimiter(leaf, self.previous)
+            if delim and self.previous is not None:
+                self.delimiters[id(self.previous)] = delim
+            else:
+                delim = is_split_after_delimiter(leaf, self.previous)
+                if delim:
+                    self.delimiters[id(leaf)] = delim
         if leaf.type in OPENING_BRACKETS:
             self.bracket_match[self.depth, BRACKET[leaf.type]] = leaf
             self.depth += 1
         self.previous = leaf
 
     def any_open_brackets(self) -> bool:
-        """Returns True if there is an yet unmatched open bracket on the line."""
+        """Return True if there is an yet unmatched open bracket on the line."""
         return bool(self.bracket_match)
 
-    def max_priority(self, exclude: Iterable[LeafID] =()) -> int:
-        """Returns the highest priority of a delimiter found on the line.
+    def max_delimiter_priority(self, exclude: Iterable[LeafID] = ()) -> int:
+        """Return the highest priority of a delimiter found on the line.
 
         Values are consistent with what `is_delimiter()` returns.
+        Raises ValueError on no delimiters.
         """
         return max(v for k, v in self.delimiters.items() if k not in exclude)
 
 
 @dataclass
 class Line:
-    depth: int = attrib(default=0)
-    leaves: List[Leaf] = attrib(default=Factory(list))
-    comments: Dict[LeafID, Leaf] = attrib(default=Factory(dict))
-    bracket_tracker: BracketTracker = attrib(default=Factory(BracketTracker))
-    inside_brackets: bool = attrib(default=False)
-    has_for: bool = attrib(default=False)
-    _for_loop_variable: bool = attrib(default=False, init=False)
+    """Holds leaves and comments. Can be printed with `str(line)`."""
+
+    depth: int = 0
+    leaves: List[Leaf] = Factory(list)
+    comments: List[Tuple[Index, Leaf]] = Factory(list)
+    bracket_tracker: BracketTracker = Factory(BracketTracker)
+    inside_brackets: bool = False
+    has_for: bool = False
+    _for_loop_variable: bool = False
 
     def append(self, leaf: Leaf, preformatted: bool = False) -> None:
-        has_value = leaf.value.strip()
+        """Add a new `leaf` to the end of the line.
+
+        Unless `preformatted` is True, the `leaf` will receive a new consistent
+        whitespace prefix and metadata applied by :class:`BracketTracker`.
+        Trailing commas are maybe removed, unpacked for loop variables are
+        demoted from being delimiters.
+
+        Inline comments are put aside.
+        """
+        has_value = leaf.type in BRACKETS or bool(leaf.value.strip())
         if not has_value:
             return
 
@@ -409,35 +703,54 @@ class Line:
             self.bracket_tracker.mark(leaf)
             self.maybe_remove_trailing_comma(leaf)
             self.maybe_increment_for_loop_variable(leaf)
-            if self.maybe_adapt_standalone_comment(leaf):
-                return
 
         if not self.append_comment(leaf):
             self.leaves.append(leaf)
 
+    def append_safe(self, leaf: Leaf, preformatted: bool = False) -> None:
+        """Like :func:`append()` but disallow invalid standalone comment structure.
+
+        Raises ValueError when any `leaf` is appended after a standalone comment
+        or when a standalone comment is not the first leaf on the line.
+        """
+        if self.bracket_tracker.depth == 0:
+            if self.is_comment:
+                raise ValueError("cannot append to standalone comments")
+
+            if self.leaves and leaf.type == STANDALONE_COMMENT:
+                raise ValueError(
+                    "cannot append standalone comments to a populated line"
+                )
+
+        self.append(leaf, preformatted=preformatted)
+
     @property
     def is_comment(self) -> bool:
-        return bool(self) and self.leaves[0].type == STANDALONE_COMMENT
+        """Is this line a standalone comment?"""
+        return len(self.leaves) == 1 and self.leaves[0].type == STANDALONE_COMMENT
 
     @property
     def is_decorator(self) -> bool:
+        """Is this line a decorator?"""
         return bool(self) and self.leaves[0].type == token.AT
 
     @property
     def is_import(self) -> bool:
+        """Is this an import line?"""
         return bool(self) and is_import(self.leaves[0])
 
     @property
     def is_class(self) -> bool:
+        """Is this line a class definition?"""
         return (
-            bool(self) and
-            self.leaves[0].type == token.NAME and
-            self.leaves[0].value == 'class'
+            bool(self)
+            and self.leaves[0].type == token.NAME
+            and self.leaves[0].value == "class"
         )
 
     @property
     def is_def(self) -> bool:
-        """Also returns True for async defs."""
+        """Is this a function definition? (Also returns True for async defs.)"""
         try:
             first_leaf = self.leaves[0]
         except IndexError:
@@ -448,50 +761,70 @@ class Line:
         except IndexError:
             second_leaf = None
         return (
-            (first_leaf.type == token.NAME and first_leaf.value == 'def') or
-            (
-                first_leaf.type == token.NAME and
-                first_leaf.value == 'async' and
-                second_leaf is not None and
-                second_leaf.type == token.NAME and
-                second_leaf.value == 'def'
+            (first_leaf.type == token.NAME and first_leaf.value == "def")
+            or (
+                first_leaf.type == token.ASYNC
+                and second_leaf is not None
+                and second_leaf.type == token.NAME
+                and second_leaf.value == "def"
             )
         )
 
     @property
     def is_flow_control(self) -> bool:
+        """Is this line a flow control statement?
+
+        Those are `return`, `raise`, `break`, and `continue`.
+        """
         return (
-            bool(self) and
-            self.leaves[0].type == token.NAME and
-            self.leaves[0].value in FLOW_CONTROL
+            bool(self)
+            and self.leaves[0].type == token.NAME
+            and self.leaves[0].value in FLOW_CONTROL
         )
 
     @property
     def is_yield(self) -> bool:
+        """Is this line a yield statement?"""
         return (
-            bool(self) and
-            self.leaves[0].type == token.NAME and
-            self.leaves[0].value == 'yield'
+            bool(self)
+            and self.leaves[0].type == token.NAME
+            and self.leaves[0].value == "yield"
         )
 
+    def contains_standalone_comments(self, depth_limit: int = sys.maxsize) -> bool:
+        """If so, needs to be split before emitting."""
+        for leaf in self.leaves:
+            if leaf.type == STANDALONE_COMMENT:
+                if leaf.bracket_depth <= depth_limit:
+                    return True
+
+        return False
+
     def maybe_remove_trailing_comma(self, closing: Leaf) -> bool:
+        """Remove trailing comma if there is one and it's safe."""
         if not (
-            self.leaves and
-            self.leaves[-1].type == token.COMMA and
-            closing.type in CLOSING_BRACKETS
+            self.leaves
+            and self.leaves[-1].type == token.COMMA
+            and closing.type in CLOSING_BRACKETS
         ):
             return False
 
-        if closing.type == token.RSQB or closing.type == token.RBRACE:
-            self.leaves.pop()
+        if closing.type == token.RBRACE:
+            self.remove_trailing_comma()
             return True
 
+        if closing.type == token.RSQB:
+            comma = self.leaves[-1]
+            if comma.parent and comma.parent.type == syms.listmaker:
+                self.remove_trailing_comma()
+                return True
+
         # For parens let's check if it's safe to remove the comma.  If the
         # trailing one is the only one, we might mistakenly change a tuple
         # into a different type by removing the comma.
-        depth = closing.bracket_depth + 1  # type: ignore
+        depth = closing.bracket_depth + 1
         commas = 0
-        opening = closing.opening_bracket  # type: ignore
+        opening = closing.opening_bracket
         for _opening_index, leaf in enumerate(self.leaves):
             if leaf is opening:
                 break
@@ -503,11 +836,15 @@ class Line:
             if leaf is closing:
                 break
 
-            bracket_depth = leaf.bracket_depth  # type: ignore
+            bracket_depth = leaf.bracket_depth
             if bracket_depth == depth and leaf.type == token.COMMA:
                 commas += 1
+                if leaf.parent and leaf.parent.type == syms.arglist:
+                    commas += 1
+                    break
+
         if commas > 1:
-            self.leaves.pop()
+            self.remove_trailing_comma()
             return True
 
         return False
@@ -515,10 +852,10 @@ class Line:
     def maybe_increment_for_loop_variable(self, leaf: Leaf) -> bool:
         """In a for loop, or comprehension, the variables are often unpacks.
 
-        To avoid splitting on the comma in this situation, we will increase
-        the depth of tokens between `for` and `in`.
+        To avoid splitting on the comma in this situation, increase the depth of
+        tokens between `for` and `in`.
         """
-        if leaf.type == token.NAME and leaf.value == 'for':
+        if leaf.type == token.NAME and leaf.value == "for":
             self.has_for = True
             self.bracket_tracker.depth += 1
             self._for_loop_variable = True
@@ -527,106 +864,172 @@ class Line:
         return False
 
     def maybe_decrement_after_for_loop_variable(self, leaf: Leaf) -> bool:
-        # See `maybe_increment_for_loop_variable` above for explanation.
-        if self._for_loop_variable and leaf.type == token.NAME and leaf.value == 'in':
+        """See `maybe_increment_for_loop_variable` above for explanation."""
+        if self._for_loop_variable and leaf.type == token.NAME and leaf.value == "in":
             self.bracket_tracker.depth -= 1
             self._for_loop_variable = False
             return True
 
         return False
 
-    def maybe_adapt_standalone_comment(self, comment: Leaf) -> bool:
-        """Hack a standalone comment to act as a trailing comment for line splitting.
-
-        If this line has brackets and a standalone `comment`, we need to adapt
-        it to be able to still reformat the line.
-
-        This is not perfect, the line to which the standalone comment gets
-        appended will appear "too long" when splitting.
-        """
-        if not (
-            comment.type == STANDALONE_COMMENT and
-            self.bracket_tracker.any_open_brackets()
+    def append_comment(self, comment: Leaf) -> bool:
+        """Add an inline or standalone comment to the line."""
+        if (
+            comment.type == STANDALONE_COMMENT
+            and self.bracket_tracker.any_open_brackets()
         ):
+            comment.prefix = ""
             return False
 
-        comment.type = token.COMMENT
-        comment.prefix = '\n' + '    ' * (self.depth + 1)
-        return self.append_comment(comment)
-
-    def append_comment(self, comment: Leaf) -> bool:
         if comment.type != token.COMMENT:
             return False
 
-        try:
-            after = id(self.last_non_delimiter())
-        except LookupError:
+        after = len(self.leaves) - 1
+        if after == -1:
             comment.type = STANDALONE_COMMENT
-            comment.prefix = ''
+            comment.prefix = ""
             return False
 
         else:
-            if after in self.comments:
-                self.comments[after].value += str(comment)
-            else:
-                self.comments[after] = comment
+            self.comments.append((after, comment))
             return True
 
-    def last_non_delimiter(self) -> Leaf:
-        for i in range(len(self.leaves)):
-            last = self.leaves[-i - 1]
-            if not is_delimiter(last):
-                return last
+    def comments_after(self, leaf: Leaf) -> Iterator[Leaf]:
+        """Generate comments that should appear directly after `leaf`."""
+        for _leaf_index, _leaf in enumerate(self.leaves):
+            if leaf is _leaf:
+                break
 
-        raise LookupError("No non-delimiters found")
+        else:
+            return
+
+        for index, comment_after in self.comments:
+            if _leaf_index == index:
+                yield comment_after
+
+    def remove_trailing_comma(self) -> None:
+        """Remove the trailing comma and moves the comments attached to it."""
+        comma_index = len(self.leaves) - 1
+        for i in range(len(self.comments)):
+            comment_index, comment = self.comments[i]
+            if comment_index == comma_index:
+                self.comments[i] = (comma_index - 1, comment)
+        self.leaves.pop()
 
     def __str__(self) -> str:
+        """Render the line."""
         if not self:
-            return '\n'
+            return "\n"
 
-        indent = '    ' * self.depth
+        indent = "    " * self.depth
         leaves = iter(self.leaves)
         first = next(leaves)
-        res = f'{first.prefix}{indent}{first.value}'
+        res = f"{first.prefix}{indent}{first.value}"
         for leaf in leaves:
             res += str(leaf)
-        for comment in self.comments.values():
+        for _, comment in self.comments:
             res += str(comment)
-        return res + '\n'
+        return res + "\n"
 
     def __bool__(self) -> bool:
+        """Return True if the line has leaves or comments."""
         return bool(self.leaves or self.comments)
 
 
+class UnformattedLines(Line):
+    """Just like :class:`Line` but stores lines which aren't reformatted."""
+
+    def append(self, leaf: Leaf, preformatted: bool = True) -> None:
+        """Just add a new `leaf` to the end of the lines.
+
+        The `preformatted` argument is ignored.
+
+        Keeps track of indentation `depth`, which is useful when the user
+        says `# fmt: on`. Otherwise, doesn't do anything with the `leaf`.
+        """
+        try:
+            list(generate_comments(leaf))
+        except FormatOn as f_on:
+            self.leaves.append(f_on.leaf_from_consumed(leaf))
+            raise
+
+        self.leaves.append(leaf)
+        if leaf.type == token.INDENT:
+            self.depth += 1
+        elif leaf.type == token.DEDENT:
+            self.depth -= 1
+
+    def __str__(self) -> str:
+        """Render unformatted lines from leaves which were added with `append()`.
+
+        `depth` is not used for indentation in this case.
+        """
+        if not self:
+            return "\n"
+
+        res = ""
+        for leaf in self.leaves:
+            res += str(leaf)
+        return res
+
+    def append_comment(self, comment: Leaf) -> bool:
+        """Not implemented in this class. Raises `NotImplementedError`."""
+        raise NotImplementedError("Unformatted lines don't store comments separately.")
+
+    def maybe_remove_trailing_comma(self, closing: Leaf) -> bool:
+        """Does nothing and returns False."""
+        return False
+
+    def maybe_increment_for_loop_variable(self, leaf: Leaf) -> bool:
+        """Does nothing and returns False."""
+        return False
+
+
 @dataclass
 class EmptyLineTracker:
     """Provides a stateful method that returns the number of potential extra
     empty lines needed before and after the currently processed line.
 
-    Note: this tracker works on lines that haven't been split yet.
+    Note: this tracker works on lines that haven't been split yet.  It assumes
+    the prefix of the first leaf consists of optional newlines.  Those newlines
+    are consumed by `maybe_empty_lines()` and included in the computation.
     """
-    previous_line: Optional[Line] = attrib(default=None)
-    previous_after: int = attrib(default=0)
-    previous_defs: List[int] = attrib(default=Factory(list))
+    previous_line: Optional[Line] = None
+    previous_after: int = 0
+    previous_defs: List[int] = Factory(list)
 
     def maybe_empty_lines(self, current_line: Line) -> Tuple[int, int]:
-        """Returns the number of extra empty lines before and after the `current_line`.
+        """Return the number of extra empty lines before and after the `current_line`.
 
-        This is for separating `def`, `async def` and `class` with extra empty lines
-        (two on module-level), as well as providing an extra empty line after flow
-        control keywords to make them more prominent.
+        This is for separating `def`, `async def` and `class` with extra empty
+        lines (two on module-level), as well as providing an extra empty line
+        after flow control keywords to make them more prominent.
         """
+        if isinstance(current_line, UnformattedLines):
+            return 0, 0
+
         before, after = self._maybe_empty_lines(current_line)
+        before -= self.previous_after
         self.previous_after = after
         self.previous_line = current_line
         return before, after
 
     def _maybe_empty_lines(self, current_line: Line) -> Tuple[int, int]:
-        before = 0
+        max_allowed = 1
+        if current_line.depth == 0:
+            max_allowed = 2
+        if current_line.leaves:
+            # Consume the first leaf's extra newlines.
+            first_leaf = current_line.leaves[0]
+            before = first_leaf.prefix.count("\n")
+            before = min(before, max_allowed)
+            first_leaf.prefix = ""
+        else:
+            before = 0
         depth = current_line.depth
         while self.previous_defs and self.previous_defs[-1] >= depth:
             self.previous_defs.pop()
-            before = (1 if depth else 2) - self.previous_after
+            before = 1 if depth else 2
         is_decorator = current_line.is_decorator
         if is_decorator or current_line.is_def or current_line.is_class:
             if not is_decorator:
@@ -642,24 +1045,23 @@ class EmptyLineTracker:
             newlines = 2
             if current_line.depth:
                 newlines -= 1
-            newlines -= self.previous_after
             return newlines, 0
 
         if current_line.is_flow_control:
             return before, 1
 
         if (
-            self.previous_line and
-            self.previous_line.is_import and
-            not current_line.is_import and
-            depth == self.previous_line.depth
+            self.previous_line
+            and self.previous_line.is_import
+            and not current_line.is_import
+            and depth == self.previous_line.depth
         ):
             return (before or 1), 0
 
         if (
-            self.previous_line and
-            self.previous_line.is_yield and
-            (not current_line.is_yield or depth != self.previous_line.depth)
+            self.previous_line
+            and self.previous_line.is_yield
+            and (not current_line.is_yield or depth != self.previous_line.depth)
         ):
             return (before or 1), 0
 
@@ -673,10 +1075,9 @@ class LineGenerator(Visitor[Line]):
     Note: destroys the tree it's visiting by mutating prefixes of its leaves
     in ways that will no longer stringify to valid Python code on the tree.
     """
-    current_line: Line = attrib(default=Factory(Line))
-    standalone_comments: List[Leaf] = attrib(default=Factory(list))
+    current_line: Line = Factory(Line)
 
-    def line(self, indent: int = 0) -> Iterator[Line]:
+    def line(self, indent: int = 0, type: Type[Line] = Line) -> Iterator[Line]:
         """Generate a line.
 
         If the line is empty, only emit if it makes sense.
@@ -685,60 +1086,95 @@ class LineGenerator(Visitor[Line]):
         If any lines were generated, set up a new current_line.
         """
         if not self.current_line:
-            self.current_line.depth += indent
+            if self.current_line.__class__ == type:
+                self.current_line.depth += indent
+            else:
+                self.current_line = type(depth=self.current_line.depth + indent)
             return  # Line is empty, don't emit. Creating a new one unnecessary.
 
         complete_line = self.current_line
-        self.current_line = Line(depth=complete_line.depth + indent)
+        self.current_line = type(depth=complete_line.depth + indent)
         yield complete_line
 
+    def visit(self, node: LN) -> Iterator[Line]:
+        """Main method to visit `node` and its children.
+
+        Yields :class:`Line` objects.
+        """
+        if isinstance(self.current_line, UnformattedLines):
+            # File contained `# fmt: off`
+            yield from self.visit_unformatted(node)
+
+        else:
+            yield from super().visit(node)
+
     def visit_default(self, node: LN) -> Iterator[Line]:
+        """Default `visit_*()` implementation. Recurses to children of `node`."""
         if isinstance(node, Leaf):
-            for comment in generate_comments(node):
-                if self.current_line.bracket_tracker.any_open_brackets():
-                    # any comment within brackets is subject to splitting
-                    self.current_line.append(comment)
-                elif comment.type == token.COMMENT:
-                    # regular trailing comment
-                    self.current_line.append(comment)
-                    yield from self.line()
+            any_open_brackets = self.current_line.bracket_tracker.any_open_brackets()
+            try:
+                for comment in generate_comments(node):
+                    if any_open_brackets:
+                        # any comment within brackets is subject to splitting
+                        self.current_line.append(comment)
+                    elif comment.type == token.COMMENT:
+                        # regular trailing comment
+                        self.current_line.append(comment)
+                        yield from self.line()
+
+                    else:
+                        # regular standalone comment
+                        yield from self.line()
+
+                        self.current_line.append(comment)
+                        yield from self.line()
+
+            except FormatOff as f_off:
+                f_off.trim_prefix(node)
+                yield from self.line(type=UnformattedLines)
+                yield from self.visit(node)
+
+            except FormatOn as f_on:
+                # This only happens here if somebody says "fmt: on" multiple
+                # times in a row.
+                f_on.trim_prefix(node)
+                yield from self.visit_default(node)
 
-                else:
-                    # regular standalone comment, to be processed later (see
-                    # docstring in `generate_comments()`
-                    self.standalone_comments.append(comment)
-            normalize_prefix(node)
-            if node.type not in WHITESPACE:
-                for comment in self.standalone_comments:
-                    yield from self.line()
-
-                    self.current_line.append(comment)
-                    yield from self.line()
-
-                self.standalone_comments = []
-                self.current_line.append(node)
+            else:
+                normalize_prefix(node, inside_brackets=any_open_brackets)
+                if node.type == token.STRING:
+                    normalize_string_quotes(node)
+                if node.type not in WHITESPACE:
+                    self.current_line.append(node)
         yield from super().visit_default(node)
 
-    def visit_suite(self, node: Node) -> Iterator[Line]:
-        """Body of a statement after a colon."""
-        children = iter(node.children)
-        # Process newline before indenting.  It might contain an inline
-        # comment that should go right after the colon.
-        newline = next(children)
-        yield from self.visit(newline)
+    def visit_INDENT(self, node: Node) -> Iterator[Line]:
+        """Increase indentation level, maybe yield a line."""
+        # In blib2to3 INDENT never holds comments.
         yield from self.line(+1)
+        yield from self.visit_default(node)
 
-        for child in children:
-            yield from self.visit(child)
-
+    def visit_DEDENT(self, node: Node) -> Iterator[Line]:
+        """Decrease indentation level, maybe yield a line."""
+        # DEDENT has no value. Additionally, in blib2to3 it never holds comments.
         yield from self.line(-1)
 
-    def visit_stmt(self, node: Node, keywords: Set[str]) -> Iterator[Line]:
+    def visit_stmt(
+        self, node: Node, keywords: Set[str], parens: Set[str]
+    ) -> Iterator[Line]:
         """Visit a statement.
 
-        The relevant Python language keywords for this statement are NAME leaves
-        within it.
+        This implementation is shared for `if`, `while`, `for`, `try`, `except`,
+        `def`, `with`, `class`, and `assert`.
+
+        The relevant Python language `keywords` for a given statement will be
+        NAME leaves within it. This methods puts those on a separate line.
+
+        `parens` holds pairs of nodes where invisible parentheses should be put.
+        Keys hold nodes after which opening parentheses should be put, values
+        hold nodes before which closing parentheses should be put.
         """
+        normalize_invisible_parens(node, parens_after=parens)
         for child in node.children:
             if child.type == token.NAME and child.value in keywords:  # type: ignore
                 yield from self.line()
@@ -746,7 +1182,7 @@ class LineGenerator(Visitor[Line]):
             yield from self.visit(child)
 
     def visit_simple_stmt(self, node: Node) -> Iterator[Line]:
-        """A statement without nested statements."""
+        """Visit a statement without nested statements."""
         is_suite_like = node.parent and node.parent.type in STATEMENT
         if is_suite_like:
             yield from self.line(+1)
@@ -758,13 +1194,14 @@ class LineGenerator(Visitor[Line]):
             yield from self.visit_default(node)
 
     def visit_async_stmt(self, node: Node) -> Iterator[Line]:
+        """Visit `async def`, `async for`, `async with`."""
         yield from self.line()
 
         children = iter(node.children)
         for child in children:
             yield from self.visit(child)
 
-            if child.type == token.NAME and child.value == 'async':  # type: ignore
+            if child.type == token.ASYNC:
                 break
 
         internal_stmt = next(children)
@@ -772,96 +1209,155 @@ class LineGenerator(Visitor[Line]):
             yield from self.visit(child)
 
     def visit_decorators(self, node: Node) -> Iterator[Line]:
+        """Visit decorators."""
         for child in node.children:
             yield from self.line()
             yield from self.visit(child)
 
+    def visit_import_from(self, node: Node) -> Iterator[Line]:
+        """Visit import_from and maybe put invisible parentheses.
+
+        This is separate from `visit_stmt` because import statements don't
+        support arbitrary atoms and thus handling of parentheses is custom.
+        """
+        check_lpar = False
+        for index, child in enumerate(node.children):
+            if check_lpar:
+                if child.type == token.LPAR:
+                    # make parentheses invisible
+                    child.value = ""  # type: ignore
+                    node.children[-1].value = ""  # type: ignore
+                else:
+                    # insert invisible parentheses
+                    node.insert_child(index, Leaf(token.LPAR, ""))
+                    node.append_child(Leaf(token.RPAR, ""))
+                break
+
+            check_lpar = (
+                child.type == token.NAME and child.value == "import"  # type: ignore
+            )
+
+        for child in node.children:
+            yield from self.visit(child)
+
     def visit_SEMI(self, leaf: Leaf) -> Iterator[Line]:
+        """Remove a semicolon and put the other statement on a separate line."""
         yield from self.line()
 
     def visit_ENDMARKER(self, leaf: Leaf) -> Iterator[Line]:
+        """End of file. Process outstanding comments and end with a newline."""
         yield from self.visit_default(leaf)
         yield from self.line()
 
+    def visit_unformatted(self, node: LN) -> Iterator[Line]:
+        """Used when file contained a `# fmt: off`."""
+        if isinstance(node, Node):
+            for child in node.children:
+                yield from self.visit(child)
+
+        else:
+            try:
+                self.current_line.append(node)
+            except FormatOn as f_on:
+                f_on.trim_prefix(node)
+                yield from self.line()
+                yield from self.visit(node)
+
+            if node.type == token.ENDMARKER:
+                # somebody decided not to put a final `# fmt: on`
+                yield from self.line()
+
     def __attrs_post_init__(self) -> None:
         """You are in a twisty little maze of passages."""
         v = self.visit_stmt
-        self.visit_if_stmt = partial(v, keywords={'if', 'else', 'elif'})
-        self.visit_while_stmt = partial(v, keywords={'while', 'else'})
-        self.visit_for_stmt = partial(v, keywords={'for', 'else'})
-        self.visit_try_stmt = partial(v, keywords={'try', 'except', 'else', 'finally'})
-        self.visit_except_clause = partial(v, keywords={'except'})
-        self.visit_funcdef = partial(v, keywords={'def'})
-        self.visit_with_stmt = partial(v, keywords={'with'})
-        self.visit_classdef = partial(v, keywords={'class'})
+        Ø: Set[str] = set()
+        self.visit_assert_stmt = partial(v, keywords={"assert"}, parens={"assert", ","})
+        self.visit_if_stmt = partial(v, keywords={"if", "else", "elif"}, parens={"if"})
+        self.visit_while_stmt = partial(v, keywords={"while", "else"}, parens={"while"})
+        self.visit_for_stmt = partial(v, keywords={"for", "else"}, parens={"for", "in"})
+        self.visit_try_stmt = partial(
+            v, keywords={"try", "except", "else", "finally"}, parens=Ø
+        )
+        self.visit_except_clause = partial(v, keywords={"except"}, parens=Ø)
+        self.visit_with_stmt = partial(v, keywords={"with"}, parens=Ø)
+        self.visit_funcdef = partial(v, keywords={"def"}, parens=Ø)
+        self.visit_classdef = partial(v, keywords={"class"}, parens=Ø)
         self.visit_async_funcdef = self.visit_async_stmt
         self.visit_decorated = self.visit_decorators
 
 
+IMPLICIT_TUPLE = {syms.testlist, syms.testlist_star_expr, syms.exprlist}
 BRACKET = {token.LPAR: token.RPAR, token.LSQB: token.RSQB, token.LBRACE: token.RBRACE}
 OPENING_BRACKETS = set(BRACKET.keys())
 CLOSING_BRACKETS = set(BRACKET.values())
 BRACKETS = OPENING_BRACKETS | CLOSING_BRACKETS
+ALWAYS_NO_SPACE = CLOSING_BRACKETS | {token.COMMA, STANDALONE_COMMENT}
 
 
-def whitespace(leaf: Leaf) -> str:
+def whitespace(leaf: Leaf) -> str:  # noqa C901
     """Return whitespace prefix if needed for the given `leaf`."""
-    NO = ''
-    SPACE = ' '
-    DOUBLESPACE = '  '
+    NO = ""
+    SPACE = " "
+    DOUBLESPACE = "  "
     t = leaf.type
     p = leaf.parent
     v = leaf.value
-    if t == token.COLON:
-        return NO
-
-    if t == token.COMMA:
-        return NO
-
-    if t == token.RPAR:
+    if t in ALWAYS_NO_SPACE:
         return NO
 
     if t == token.COMMENT:
         return DOUBLESPACE
 
-    if t == STANDALONE_COMMENT:
-        return NO
-
-    if t in CLOSING_BRACKETS:
+    assert p is not None, f"INTERNAL ERROR: hand-made leaf without parent: {leaf!r}"
+    if t == token.COLON and p.type not in {syms.subscript, syms.subscriptlist}:
         return NO
 
-    assert p is not None, f"INTERNAL ERROR: hand-made leaf without parent: {leaf!r}"
     prev = leaf.prev_sibling
     if not prev:
         prevp = preceding_leaf(p)
         if not prevp or prevp.type in OPENING_BRACKETS:
             return NO
 
+        if t == token.COLON:
+            return SPACE if prevp.type == token.COMMA else NO
+
         if prevp.type == token.EQUAL:
-            if prevp.parent and prevp.parent.type in {
-                syms.typedargslist,
-                syms.varargslist,
-                syms.parameters,
-                syms.arglist,
-                syms.argument,
-            }:
-                return NO
+            if prevp.parent:
+                if prevp.parent.type in {
+                    syms.arglist, syms.argument, syms.parameters, syms.varargslist
+                }:
+                    return NO
+
+                elif prevp.parent.type == syms.typedargslist:
+                    # A bit hacky: if the equal sign has whitespace, it means we
+                    # previously found it's a typed argument.  So, we're using
+                    # that, too.
+                    return prevp.prefix
 
-        elif prevp.type == token.DOUBLESTAR:
-            if prevp.parent and prevp.parent.type in {
-                syms.typedargslist,
-                syms.varargslist,
-                syms.parameters,
-                syms.arglist,
-                syms.dictsetmaker,
-            }:
+        elif prevp.type in STARS:
+            if is_vararg(prevp, within=VARARGS_PARENTS | UNPACKING_PARENTS):
                 return NO
 
         elif prevp.type == token.COLON:
-            if prevp.parent and prevp.parent.type == syms.subscript:
+            if prevp.parent and prevp.parent.type in {syms.subscript, syms.sliceop}:
                 return NO
 
-        elif prevp.parent and prevp.parent.type == syms.factor:
+        elif (
+            prevp.parent
+            and prevp.parent.type == syms.factor
+            and prevp.type in MATH_OPERATORS
+        ):
+            return NO
+
+        elif (
+            prevp.type == token.RIGHTSHIFT
+            and prevp.parent
+            and prevp.parent.type == syms.shift_expr
+            and prevp.prev_sibling
+            and prevp.prev_sibling.type == token.NAME
+            and prevp.prev_sibling.value == "print"  # type: ignore
+        ):
+            # Python 2 print chevron
             return NO
 
     elif prev.type in OPENING_BRACKETS:
@@ -875,7 +1371,7 @@ def whitespace(leaf: Leaf) -> str:
         if not prev or prev.type != token.COMMA:
             return NO
 
-    if p.type == syms.varargslist:
+    elif p.type == syms.varargslist:
         # lambdas
         if t == token.RPAR:
             return NO
@@ -934,7 +1430,7 @@ def whitespace(leaf: Leaf) -> str:
             if not prevp or prevp.type == token.LPAR:
                 return NO
 
-        elif prev.type == token.EQUAL or prev.type == token.DOUBLESTAR:
+        elif prev.type in {token.EQUAL} | STARS:
             return NO
 
     elif p.type == syms.decorator:
@@ -958,7 +1454,14 @@ def whitespace(leaf: Leaf) -> str:
 
     elif p.type == syms.subscript:
         # indexing
-        if not prev or prev.type == token.COLON:
+        if not prev:
+            assert p.parent is not None, "subscripts are always parented"
+            if p.parent.type == syms.subscriptlist:
+                return SPACE
+
+            return NO
+
+        else:
             return NO
 
     elif p.type == syms.atom:
@@ -967,9 +1470,9 @@ def whitespace(leaf: Leaf) -> str:
             return NO
 
     elif (
-        p.type == syms.listmaker or
-        p.type == syms.testlist_gexp or
-        p.type == syms.subscriptlist
+        p.type == syms.listmaker
+        or p.type == syms.testlist_gexp
+        or p.type == syms.subscriptlist
     ):
         # list interior, including unpacking
         if not prev:
@@ -983,7 +1486,7 @@ def whitespace(leaf: Leaf) -> str:
         if prev.type == token.DOUBLESTAR:
             return NO
 
-    elif p.type == syms.factor or p.type == syms.star_expr:
+    elif p.type in {syms.factor, syms.star_expr}:
         # unary ops
         if not prev:
             prevp = preceding_leaf(p)
@@ -992,9 +1495,10 @@ def whitespace(leaf: Leaf) -> str:
 
             prevp_parent = prevp.parent
             assert prevp_parent is not None
-            if prevp.type == token.COLON and prevp_parent.type in {
-                syms.subscript, syms.sliceop
-            }:
+            if (
+                prevp.type == token.COLON
+                and prevp_parent.type in {syms.subscript, syms.sliceop}
+            ):
                 return NO
 
             elif prevp.type == token.EQUAL and prevp_parent.type == syms.argument:
@@ -1009,7 +1513,7 @@ def whitespace(leaf: Leaf) -> str:
                 return NO
 
         elif t == token.NAME:
-            if v == 'import':
+            if v == "import":
                 return SPACE
 
             if prev and prev.type == token.DOT:
@@ -1022,7 +1526,7 @@ def whitespace(leaf: Leaf) -> str:
 
 
 def preceding_leaf(node: Optional[LN]) -> Optional[Leaf]:
-    """Returns the first leaf that precedes `node`, if any."""
+    """Return the first leaf that precedes `node`, if any."""
     while node:
         res = node.prev_sibling
         if res:
@@ -1039,32 +1543,85 @@ def preceding_leaf(node: Optional[LN]) -> Optional[Leaf]:
     return None
 
 
-def is_delimiter(leaf: Leaf) -> int:
-    """Returns the priority of the `leaf` delimiter. Returns 0 if not delimiter.
+def is_split_after_delimiter(leaf: Leaf, previous: Leaf = None) -> int:
+    """Return the priority of the `leaf` delimiter, given a line break after it.
+
+    The delimiter priorities returned here are from those delimiters that would
+    cause a line break after themselves.
 
     Higher numbers are higher priority.
     """
     if leaf.type == token.COMMA:
         return COMMA_PRIORITY
 
-    if leaf.type == token.NAME and leaf.value in LOGIC_OPERATORS:
-        return LOGIC_PRIORITY
+    return 0
+
+
+def is_split_before_delimiter(leaf: Leaf, previous: Leaf = None) -> int:
+    """Return the priority of the `leaf` delimiter, given a line before after it.
+
+    The delimiter priorities returned here are from those delimiters that would
+    cause a line break before themselves.
+
+    Higher numbers are higher priority.
+    """
+    if is_vararg(leaf, within=VARARGS_PARENTS | UNPACKING_PARENTS):
+        # * and ** might also be MATH_OPERATORS but in this case they are not.
+        # Don't treat them as a delimiter.
+        return 0
+
+    if (
+        leaf.type in MATH_OPERATORS
+        and leaf.parent
+        and leaf.parent.type not in {syms.factor, syms.star_expr}
+    ):
+        return MATH_PRIORITY
 
     if leaf.type in COMPARATORS:
         return COMPARATOR_PRIORITY
 
     if (
-        leaf.type in MATH_OPERATORS and
-        leaf.parent and
-        leaf.parent.type not in {syms.factor, syms.star_expr}
+        leaf.type == token.STRING
+        and previous is not None
+        and previous.type == token.STRING
     ):
-        return MATH_PRIORITY
+        return STRING_PRIORITY
+
+    if (
+        leaf.type == token.NAME
+        and leaf.value == "for"
+        and leaf.parent
+        and leaf.parent.type in {syms.comp_for, syms.old_comp_for}
+    ):
+        return COMPREHENSION_PRIORITY
+
+    if (
+        leaf.type == token.NAME
+        and leaf.value == "if"
+        and leaf.parent
+        and leaf.parent.type in {syms.comp_if, syms.old_comp_if}
+    ):
+        return COMPREHENSION_PRIORITY
+
+    if leaf.type == token.NAME and leaf.value in LOGIC_OPERATORS and leaf.parent:
+        return LOGIC_PRIORITY
 
     return 0
 
 
+def is_delimiter(leaf: Leaf, previous: Leaf = None) -> int:
+    """Return the priority of the `leaf` delimiter. Return 0 if not delimiter.
+
+    Higher numbers are higher priority.
+    """
+    return max(
+        is_split_before_delimiter(leaf, previous),
+        is_split_after_delimiter(leaf, previous),
+    )
+
+
 def generate_comments(leaf: Leaf) -> Iterator[Leaf]:
-    """Cleans the prefix of the `leaf` and generates comments from it, if any.
+    """Clean the prefix of the `leaf` and generate comments from it, if any.
 
     Comments in lib2to3 are shoved into the whitespace prefix.  This happens
     in `pgen2/driver.py:Driver.parse_tokens()`.  This was a brilliant implementation
@@ -1082,53 +1639,94 @@ def generate_comments(leaf: Leaf) -> Iterator[Leaf]:
     Inline comments are emitted as regular token.COMMENT leaves.  Standalone
     are emitted with a fake STANDALONE_COMMENT token identifier.
     """
-    if not leaf.prefix:
+    p = leaf.prefix
+    if not p:
         return
 
-    if '#' not in leaf.prefix:
+    if "#" not in p:
         return
 
-    before_comment, content = leaf.prefix.split('#', 1)
-    content = content.rstrip()
-    if content and (content[0] not in {' ', '!', '#'}):
-        content = ' ' + content
-    is_standalone_comment = (
-        '\n' in before_comment or '\n' in content or leaf.type == token.DEDENT
-    )
-    if not is_standalone_comment:
-        # simple trailing comment
-        yield Leaf(token.COMMENT, value='#' + content)
-        return
-
-    for line in ('#' + content).split('\n'):
+    consumed = 0
+    nlines = 0
+    for index, line in enumerate(p.split("\n")):
+        consumed += len(line) + 1  # adding the length of the split '\n'
         line = line.lstrip()
-        if not line.startswith('#'):
+        if not line:
+            nlines += 1
+        if not line.startswith("#"):
             continue
 
-        yield Leaf(STANDALONE_COMMENT, line)
+        if index == 0 and leaf.type != token.ENDMARKER:
+            comment_type = token.COMMENT  # simple trailing comment
+        else:
+            comment_type = STANDALONE_COMMENT
+        comment = make_comment(line)
+        yield Leaf(comment_type, comment, prefix="\n" * nlines)
+
+        if comment in {"# fmt: on", "# yapf: enable"}:
+            raise FormatOn(consumed)
+
+        if comment in {"# fmt: off", "# yapf: disable"}:
+            if comment_type == STANDALONE_COMMENT:
+                raise FormatOff(consumed)
+
+            prev = preceding_leaf(leaf)
+            if not prev or prev.type in WHITESPACE:  # standalone comment in disguise
+                raise FormatOff(consumed)
+
+        nlines = 0
+
+
+def make_comment(content: str) -> str:
+    """Return a consistently formatted comment from the given `content` string.
+
+    All comments (except for "##", "#!", "#:") should have a single space between
+    the hash sign and the content.
+
+    If `content` didn't start with a hash sign, one is provided.
+    """
+    content = content.rstrip()
+    if not content:
+        return "#"
+
+    if content[0] == "#":
+        content = content[1:]
+    if content and content[0] not in " !:#":
+        content = " " + content
+    return "#" + content
 
 
-def split_line(line: Line, line_length: int, inner: bool = False) -> Iterator[Line]:
-    """Splits a `line` into potentially many lines.
+def split_line(
+    line: Line, line_length: int, inner: bool = False, py36: bool = False
+) -> Iterator[Line]:
+    """Split a `line` into potentially many lines.
 
     They should fit in the allotted `line_length` but might not be able to.
     `inner` signifies that there were a pair of brackets somewhere around the
     current `line`, possibly transitively. This means we can fallback to splitting
     by delimiters if the LHS/RHS don't yield any results.
+
+    If `py36` is True, splitting may generate syntax that is only compatible
+    with Python 3.6 and later.
     """
-    line_str = str(line).strip('\n')
-    if len(line_str) <= line_length and '\n' not in line_str:
+    if isinstance(line, UnformattedLines) or line.is_comment:
         yield line
         return
 
+    line_str = str(line).strip("\n")
+    if (
+        len(line_str) <= line_length
+        and "\n" not in line_str  # multiline strings
+        and not line.contains_standalone_comments()
+    ):
+        yield line
+        return
+
+    split_funcs: List[SplitFunc]
     if line.is_def:
         split_funcs = [left_hand_split]
     elif line.inside_brackets:
-        split_funcs = [delimiter_split]
-        if '\n' not in line_str:
-            # Only attempt RHS if we don't have multiline strings or comments
-            # on this line.
-            split_funcs.append(right_hand_split)
+        split_funcs = [delimiter_split, standalone_comment_split, right_hand_split]
     else:
         split_funcs = [right_hand_split]
     for split_func in split_funcs:
@@ -1137,11 +1735,13 @@ def split_line(line: Line, line_length: int, inner: bool = False) -> Iterator[Li
         # split altogether.
         result: List[Line] = []
         try:
-            for l in split_func(line):
-                if str(l).strip('\n') == line_str:
+            for l in split_func(line, py36):
+                if str(l).strip("\n") == line_str:
                     raise CannotSplit("Split function returned an unchanged result")
 
-                result.extend(split_line(l, line_length=line_length, inner=True))
+                result.extend(
+                    split_line(l, line_length=line_length, inner=True, py36=py36)
+                )
         except CannotSplit as cs:
             continue
 
@@ -1153,7 +1753,7 @@ def split_line(line: Line, line_length: int, inner: bool = False) -> Iterator[Li
         yield line
 
 
-def left_hand_split(line: Line) -> Iterator[Line]:
+def left_hand_split(line: Line, py36: bool = False) -> Iterator[Line]:
     """Split line into many lines, starting with the first matching bracket pair.
 
     Note: this usually looks weird, only use this for function definitions.
@@ -1169,11 +1769,11 @@ def left_hand_split(line: Line) -> Iterator[Line]:
     matching_bracket = None
     for leaf in line.leaves:
         if (
-            current_leaves is body_leaves and
-            leaf.type in CLOSING_BRACKETS and
-            leaf.opening_bracket is matching_bracket  # type: ignore
+            current_leaves is body_leaves
+            and leaf.type in CLOSING_BRACKETS
+            and leaf.opening_bracket is matching_bracket
         ):
-            current_leaves = tail_leaves
+            current_leaves = tail_leaves if body_leaves else head_leaves
         current_leaves.append(leaf)
         if current_leaves is head_leaves:
             if leaf.type in OPENING_BRACKETS:
@@ -1181,34 +1781,22 @@ def left_hand_split(line: Line) -> Iterator[Line]:
                 current_leaves = body_leaves
     # Since body is a new indent level, remove spurious leading whitespace.
     if body_leaves:
-        normalize_prefix(body_leaves[0])
+        normalize_prefix(body_leaves[0], inside_brackets=True)
     # Build the new lines.
-    for result, leaves in (
-        (head, head_leaves), (body, body_leaves), (tail, tail_leaves)
-    ):
+    for result, leaves in (head, head_leaves), (body, body_leaves), (tail, tail_leaves):
         for leaf in leaves:
             result.append(leaf, preformatted=True)
-            comment_after = line.comments.get(id(leaf))
-            if comment_after:
+            for comment_after in line.comments_after(leaf):
                 result.append(comment_after, preformatted=True)
-    # Check if the split succeeded.
-    tail_len = len(str(tail))
-    if not body:
-        if tail_len == 0:
-            raise CannotSplit("Splitting brackets produced the same line")
-
-        elif tail_len < 3:
-            raise CannotSplit(
-                f"Splitting brackets on an empty body to save "
-                f"{tail_len} characters is not worth it"
-            )
-
+    bracket_split_succeeded_or_raise(head, body, tail)
     for result in (head, body, tail):
         if result:
             yield result
 
 
-def right_hand_split(line: Line) -> Iterator[Line]:
+def right_hand_split(
+    line: Line, py36: bool = False, omit: Collection[LeafID] = ()
+) -> Iterator[Line]:
     """Split line into many lines, starting with the last matching bracket pair."""
     head = Line(depth=line.depth)
     body = Line(depth=line.depth + 1, inside_brackets=True)
@@ -1218,32 +1806,73 @@ def right_hand_split(line: Line) -> Iterator[Line]:
     head_leaves: List[Leaf] = []
     current_leaves = tail_leaves
     opening_bracket = None
+    closing_bracket = None
     for leaf in reversed(line.leaves):
         if current_leaves is body_leaves:
             if leaf is opening_bracket:
-                current_leaves = head_leaves
+                current_leaves = head_leaves if body_leaves else tail_leaves
         current_leaves.append(leaf)
         if current_leaves is tail_leaves:
-            if leaf.type in CLOSING_BRACKETS:
-                opening_bracket = leaf.opening_bracket  # type: ignore
+            if leaf.type in CLOSING_BRACKETS and id(leaf) not in omit:
+                opening_bracket = leaf.opening_bracket
+                closing_bracket = leaf
                 current_leaves = body_leaves
     tail_leaves.reverse()
     body_leaves.reverse()
     head_leaves.reverse()
     # Since body is a new indent level, remove spurious leading whitespace.
     if body_leaves:
-        normalize_prefix(body_leaves[0])
+        normalize_prefix(body_leaves[0], inside_brackets=True)
+    elif not head_leaves:
+        # No `head` and no `body` means the split failed. `tail` has all content.
+        raise CannotSplit("No brackets found")
+
     # Build the new lines.
-    for result, leaves in (
-        (head, head_leaves), (body, body_leaves), (tail, tail_leaves)
-    ):
+    for result, leaves in (head, head_leaves), (body, body_leaves), (tail, tail_leaves):
         for leaf in leaves:
             result.append(leaf, preformatted=True)
-            comment_after = line.comments.get(id(leaf))
-            if comment_after:
+            for comment_after in line.comments_after(leaf):
                 result.append(comment_after, preformatted=True)
-    # Check if the split succeeded.
-    tail_len = len(str(tail).strip('\n'))
+    bracket_split_succeeded_or_raise(head, body, tail)
+    assert opening_bracket and closing_bracket
+    if (
+        opening_bracket.type == token.LPAR
+        and not opening_bracket.value
+        and closing_bracket.type == token.RPAR
+        and not closing_bracket.value
+    ):
+        # These parens were optional. If there aren't any delimiters or standalone
+        # comments in the body, they were unnecessary and another split without
+        # them should be attempted.
+        if not (
+            body.bracket_tracker.delimiters or line.contains_standalone_comments(0)
+        ):
+            omit = {id(closing_bracket), *omit}
+            yield from right_hand_split(line, py36=py36, omit=omit)
+            return
+
+    ensure_visible(opening_bracket)
+    ensure_visible(closing_bracket)
+    for result in (head, body, tail):
+        if result:
+            yield result
+
+
+def bracket_split_succeeded_or_raise(head: Line, body: Line, tail: Line) -> None:
+    """Raise :exc:`CannotSplit` if the last left- or right-hand split failed.
+
+    Do nothing otherwise.
+
+    A left- or right-hand split is based on a pair of brackets. Content before
+    (and including) the opening bracket is left on one line, content inside the
+    brackets is put on a separate line, and finally content starting with and
+    following the closing bracket is put on a separate line.
+
+    Those are called `head`, `body`, and `tail`, respectively. If the split
+    produced the same line (all content in `head`) or ended up with an empty `body`
+    and the `tail` is just the closing bracket, then it's considered failed.
+    """
+    tail_len = len(str(tail).strip())
     if not body:
         if tail_len == 0:
             raise CannotSplit("Splitting brackets produced the same line")
@@ -1254,15 +1883,28 @@ def right_hand_split(line: Line) -> Iterator[Line]:
                 f"{tail_len} characters is not worth it"
             )
 
-    for result in (head, body, tail):
-        if result:
-            yield result
 
+def dont_increase_indentation(split_func: SplitFunc) -> SplitFunc:
+    """Normalize prefix of the first leaf in every line returned by `split_func`.
 
-def delimiter_split(line: Line) -> Iterator[Line]:
+    This is a decorator over relevant split functions.
+    """
+
+    @wraps(split_func)
+    def split_wrapper(line: Line, py36: bool = False) -> Iterator[Line]:
+        for l in split_func(line, py36):
+            normalize_prefix(l.leaves[0], inside_brackets=True)
+            yield l
+
+    return split_wrapper
+
+
+@dont_increase_indentation
+def delimiter_split(line: Line, py36: bool = False) -> Iterator[Line]:
     """Split according to delimiters of the highest priority.
 
-    This kind of split doesn't increase indentation.
+    If `py36` is True, the split will add trailing commas also in function
+    signatures that contain `*` and `**`.
     """
     try:
         last_leaf = line.leaves[-1]
@@ -1271,67 +1913,340 @@ def delimiter_split(line: Line) -> Iterator[Line]:
 
     delimiters = line.bracket_tracker.delimiters
     try:
-        delimiter_priority = line.bracket_tracker.max_priority(exclude={id(last_leaf)})
+        delimiter_priority = line.bracket_tracker.max_delimiter_priority(
+            exclude={id(last_leaf)}
+        )
     except ValueError:
         raise CannotSplit("No delimiters found")
 
     current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
+    lowest_depth = sys.maxsize
+    trailing_comma_safe = True
+
+    def append_to_line(leaf: Leaf) -> Iterator[Line]:
+        """Append `leaf` to current line or to new line if appending impossible."""
+        nonlocal current_line
+        try:
+            current_line.append_safe(leaf, preformatted=True)
+        except ValueError as ve:
+            yield current_line
+
+            current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
+            current_line.append(leaf)
+
     for leaf in line.leaves:
-        current_line.append(leaf, preformatted=True)
-        comment_after = line.comments.get(id(leaf))
-        if comment_after:
-            current_line.append(comment_after, preformatted=True)
+        yield from append_to_line(leaf)
+
+        for comment_after in line.comments_after(leaf):
+            yield from append_to_line(comment_after)
+
+        lowest_depth = min(lowest_depth, leaf.bracket_depth)
+        if (
+            leaf.bracket_depth == lowest_depth
+            and is_vararg(leaf, within=VARARGS_PARENTS)
+        ):
+            trailing_comma_safe = trailing_comma_safe and py36
         leaf_priority = delimiters.get(id(leaf))
         if leaf_priority == delimiter_priority:
-            normalize_prefix(current_line.leaves[0])
             yield current_line
 
             current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
     if current_line:
         if (
-            delimiter_priority == COMMA_PRIORITY and
-            current_line.leaves[-1].type != token.COMMA
+            trailing_comma_safe
+            and delimiter_priority == COMMA_PRIORITY
+            and current_line.leaves[-1].type != token.COMMA
+            and current_line.leaves[-1].type != STANDALONE_COMMENT
         ):
-            current_line.append(Leaf(token.COMMA, ','))
-        normalize_prefix(current_line.leaves[0])
+            current_line.append(Leaf(token.COMMA, ","))
+        yield current_line
+
+
+@dont_increase_indentation
+def standalone_comment_split(line: Line, py36: bool = False) -> Iterator[Line]:
+    """Split standalone comments from the rest of the line."""
+    if not line.contains_standalone_comments(0):
+        raise CannotSplit("Line does not have any standalone comments")
+
+    current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
+
+    def append_to_line(leaf: Leaf) -> Iterator[Line]:
+        """Append `leaf` to current line or to new line if appending impossible."""
+        nonlocal current_line
+        try:
+            current_line.append_safe(leaf, preformatted=True)
+        except ValueError as ve:
+            yield current_line
+
+            current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
+            current_line.append(leaf)
+
+    for leaf in line.leaves:
+        yield from append_to_line(leaf)
+
+        for comment_after in line.comments_after(leaf):
+            yield from append_to_line(comment_after)
+
+    if current_line:
         yield current_line
 
 
 def is_import(leaf: Leaf) -> bool:
-    """Returns True if the given leaf starts an import statement."""
+    """Return True if the given leaf starts an import statement."""
     p = leaf.parent
     t = leaf.type
     v = leaf.value
     return bool(
-        t == token.NAME and
-        (
-            (v == 'import' and p and p.type == syms.import_name) or
-            (v == 'from' and p and p.type == syms.import_from)
+        t == token.NAME
+        and (
+            (v == "import" and p and p.type == syms.import_name)
+            or (v == "from" and p and p.type == syms.import_from)
         )
     )
 
 
-def normalize_prefix(leaf: Leaf) -> None:
-    """Leave existing extra newlines for imports.  Remove everything else."""
-    if is_import(leaf):
-        spl = leaf.prefix.split('#', 1)
-        nl_count = spl[0].count('\n')
-        if len(spl) > 1:
-            # Skip one newline since it was for a standalone comment.
-            nl_count -= 1
-        leaf.prefix = '\n' * nl_count
+def normalize_prefix(leaf: Leaf, *, inside_brackets: bool) -> None:
+    """Leave existing extra newlines if not `inside_brackets`. Remove everything
+    else.
+
+    Note: don't use backslashes for formatting or you'll lose your voting rights.
+    """
+    if not inside_brackets:
+        spl = leaf.prefix.split("#")
+        if "\\" not in spl[0]:
+            nl_count = spl[-1].count("\n")
+            if len(spl) > 1:
+                nl_count -= 1
+            leaf.prefix = "\n" * nl_count
+            return
+
+    leaf.prefix = ""
+
+
+def normalize_string_quotes(leaf: Leaf) -> None:
+    """Prefer double quotes but only if it doesn't cause more escaping.
+
+    Adds or removes backslashes as appropriate. Doesn't parse and fix
+    strings nested in f-strings (yet).
+
+    Note: Mutates its argument.
+    """
+    value = leaf.value.lstrip("furbFURB")
+    if value[:3] == '"""':
         return
 
-    leaf.prefix = ''
+    elif value[:3] == "'''":
+        orig_quote = "'''"
+        new_quote = '"""'
+    elif value[0] == '"':
+        orig_quote = '"'
+        new_quote = "'"
+    else:
+        orig_quote = "'"
+        new_quote = '"'
+    first_quote_pos = leaf.value.find(orig_quote)
+    if first_quote_pos == -1:
+        return  # There's an internal error
+
+    prefix = leaf.value[:first_quote_pos]
+    unescaped_new_quote = re.compile(rf"(([^\\]|^)(\\\\)*){new_quote}")
+    escaped_new_quote = re.compile(rf"([^\\]|^)\\(\\\\)*{new_quote}")
+    escaped_orig_quote = re.compile(rf"([^\\]|^)\\(\\\\)*{orig_quote}")
+    body = leaf.value[first_quote_pos + len(orig_quote):-len(orig_quote)]
+    if "r" in prefix.casefold():
+        if unescaped_new_quote.search(body):
+            # There's at least one unescaped new_quote in this raw string
+            # so converting is impossible
+            return
+
+        # Do not introduce or remove backslashes in raw strings
+        new_body = body
+    else:
+        # remove unnecessary quotes
+        new_body = sub_twice(escaped_new_quote, rf"\1\2{new_quote}", body)
+        if body != new_body:
+            # Consider the string without unnecessary quotes as the original
+            body = new_body
+            leaf.value = f"{prefix}{orig_quote}{body}{orig_quote}"
+        new_body = sub_twice(escaped_orig_quote, rf"\1\2{orig_quote}", new_body)
+        new_body = sub_twice(unescaped_new_quote, rf"\1\\{new_quote}", new_body)
+    if new_quote == '"""' and new_body[-1] == '"':
+        # edge case:
+        new_body = new_body[:-1] + '\\"'
+    orig_escape_count = body.count("\\")
+    new_escape_count = new_body.count("\\")
+    if new_escape_count > orig_escape_count:
+        return  # Do not introduce more escaping
+
+    if new_escape_count == orig_escape_count and orig_quote == '"':
+        return  # Prefer double quotes
+
+    leaf.value = f"{prefix}{new_quote}{new_body}{new_quote}"
+
+
+def normalize_invisible_parens(node: Node, parens_after: Set[str]) -> None:
+    """Make existing optional parentheses invisible or create new ones.
+
+    Standardizes on visible parentheses for single-element tuples, and keeps
+    existing visible parentheses for other tuples and generator expressions.
+    """
+    check_lpar = False
+    for child in list(node.children):
+        if check_lpar:
+            if child.type == syms.atom:
+                if not (
+                    is_empty_tuple(child)
+                    or is_one_tuple(child)
+                    or max_delimiter_priority_in_atom(child) >= COMMA_PRIORITY
+                ):
+                    first = child.children[0]
+                    last = child.children[-1]
+                    if first.type == token.LPAR and last.type == token.RPAR:
+                        # make parentheses invisible
+                        first.value = ""  # type: ignore
+                        last.value = ""  # type: ignore
+            elif is_one_tuple(child):
+                # wrap child in visible parentheses
+                lpar = Leaf(token.LPAR, "(")
+                rpar = Leaf(token.RPAR, ")")
+                index = child.remove() or 0
+                node.insert_child(index, Node(syms.atom, [lpar, child, rpar]))
+            else:
+                # wrap child in invisible parentheses
+                lpar = Leaf(token.LPAR, "")
+                rpar = Leaf(token.RPAR, "")
+                index = child.remove() or 0
+                node.insert_child(index, Node(syms.atom, [lpar, child, rpar]))
+
+        check_lpar = isinstance(child, Leaf) and child.value in parens_after
+
+
+def is_empty_tuple(node: LN) -> bool:
+    """Return True if `node` holds an empty tuple."""
+    return (
+        node.type == syms.atom
+        and len(node.children) == 2
+        and node.children[0].type == token.LPAR
+        and node.children[1].type == token.RPAR
+    )
+
 
+def is_one_tuple(node: LN) -> bool:
+    """Return True if `node` holds a tuple with one element, with or without parens."""
+    if node.type == syms.atom:
+        if len(node.children) != 3:
+            return False
+
+        lpar, gexp, rpar = node.children
+        if not (
+            lpar.type == token.LPAR
+            and gexp.type == syms.testlist_gexp
+            and rpar.type == token.RPAR
+        ):
+            return False
+
+        return len(gexp.children) == 2 and gexp.children[1].type == token.COMMA
+
+    return (
+        node.type in IMPLICIT_TUPLE
+        and len(node.children) == 2
+        and node.children[1].type == token.COMMA
+    )
+
+
+def is_vararg(leaf: Leaf, within: Set[NodeType]) -> bool:
+    """Return True if `leaf` is a star or double star in a vararg or kwarg.
+
+    If `within` includes VARARGS_PARENTS, this applies to function signatures.
+    If `within` includes COLLECTION_LIBERALS_PARENTS, it applies to right
+    hand-side extended iterable unpacking (PEP 3132) and additional unpacking
+    generalizations (PEP 448).
+    """
+    if leaf.type not in STARS or not leaf.parent:
+        return False
+
+    p = leaf.parent
+    if p.type == syms.star_expr:
+        # Star expressions are also used as assignment targets in extended
+        # iterable unpacking (PEP 3132).  See what its parent is instead.
+        if not p.parent:
+            return False
+
+        p = p.parent
+
+    return p.type in within
+
+
+def max_delimiter_priority_in_atom(node: LN) -> int:
+    if node.type != syms.atom:
+        return 0
+
+    first = node.children[0]
+    last = node.children[-1]
+    if not (first.type == token.LPAR and last.type == token.RPAR):
+        return 0
+
+    bt = BracketTracker()
+    for c in node.children[1:-1]:
+        if isinstance(c, Leaf):
+            bt.mark(c)
+        else:
+            for leaf in c.leaves():
+                bt.mark(leaf)
+    try:
+        return bt.max_delimiter_priority()
+
+    except ValueError:
+        return 0
 
-PYTHON_EXTENSIONS = {'.py'}
+
+def ensure_visible(leaf: Leaf) -> None:
+    """Make sure parentheses are visible.
+
+    They could be invisible as part of some statements (see
+    :func:`normalize_invible_parens` and :func:`visit_import_from`).
+    """
+    if leaf.type == token.LPAR:
+        leaf.value = "("
+    elif leaf.type == token.RPAR:
+        leaf.value = ")"
+
+
+def is_python36(node: Node) -> bool:
+    """Return True if the current file is using Python 3.6+ features.
+
+    Currently looking for:
+    - f-strings; and
+    - trailing commas after * or ** in function signatures.
+    """
+    for n in node.pre_order():
+        if n.type == token.STRING:
+            value_head = n.value[:2]  # type: ignore
+            if value_head in {'f"', 'F"', "f'", "F'", "rf", "fr", "RF", "FR"}:
+                return True
+
+        elif (
+            n.type == syms.typedargslist
+            and n.children
+            and n.children[-1].type == token.COMMA
+        ):
+            for ch in n.children:
+                if ch.type in STARS:
+                    return True
+
+    return False
+
+
+PYTHON_EXTENSIONS = {".py"}
 BLACKLISTED_DIRECTORIES = {
-    'build', 'buck-out', 'dist', '_build', '.git', '.hg', '.mypy_cache', '.tox', '.venv'
+    "build", "buck-out", "dist", "_build", ".git", ".hg", ".mypy_cache", ".tox", ".venv"
 }
 
 
 def gen_python_files_in_dir(path: Path) -> Iterator[Path]:
+    """Generate all files under `path` which aren't under BLACKLISTED_DIRECTORIES
+    and have one of the PYTHON_EXTENSIONS.
+    """
     for child in path.iterdir():
         if child.is_dir():
             if child.name in BLACKLISTED_DIRECTORIES:
@@ -1345,67 +2260,85 @@ def gen_python_files_in_dir(path: Path) -> Iterator[Path]:
 
 @dataclass
 class Report:
-    """Provides a reformatting counter."""
-    change_count: int = attrib(default=0)
-    same_count: int = attrib(default=0)
-    failure_count: int = attrib(default=0)
-
-    def done(self, src: Path, changed: bool) -> None:
+    """Provides a reformatting counter. Can be rendered with `str(report)`."""
+    check: bool = False
+    quiet: bool = False
+    change_count: int = 0
+    same_count: int = 0
+    failure_count: int = 0
+
+    def done(self, src: Path, changed: Changed) -> None:
         """Increment the counter for successful reformatting. Write out a message."""
-        if changed:
-            out(f'reformatted {src}')
+        if changed is Changed.YES:
+            reformatted = "would reformat" if self.check else "reformatted"
+            if not self.quiet:
+                out(f"{reformatted} {src}")
             self.change_count += 1
         else:
-            out(f'{src} already well formatted, good job.', bold=False)
+            if not self.quiet:
+                if changed is Changed.NO:
+                    msg = f"{src} already well formatted, good job."
+                else:
+                    msg = f"{src} wasn't modified on disk since last run."
+                out(msg, bold=False)
             self.same_count += 1
 
     def failed(self, src: Path, message: str) -> None:
         """Increment the counter for failed reformatting. Write out a message."""
-        err(f'error: cannot format {src}: {message}')
+        err(f"error: cannot format {src}: {message}")
         self.failure_count += 1
 
     @property
     def return_code(self) -> int:
-        """Which return code should the app use considering the current state."""
+        """Return the exit code that the app should use.
+
+        This considers the current state of changed files and failures:
+        - if there were any failures, return 123;
+        - if any files were changed and --check is being used, return 1;
+        - otherwise return 0.
+        """
         # According to http://tldp.org/LDP/abs/html/exitcodes.html starting with
         # 126 we have special returncodes reserved by the shell.
         if self.failure_count:
             return 123
 
-        elif self.change_count:
+        elif self.change_count and self.check:
             return 1
 
         return 0
 
     def __str__(self) -> str:
-        """A color report of the current state.
+        """Render a color report of the current state.
 
         Use `click.unstyle` to remove colors.
         """
+        if self.check:
+            reformatted = "would be reformatted"
+            unchanged = "would be left unchanged"
+            failed = "would fail to reformat"
+        else:
+            reformatted = "reformatted"
+            unchanged = "left unchanged"
+            failed = "failed to reformat"
         report = []
         if self.change_count:
-            s = 's' if self.change_count > 1 else ''
+            s = "s" if self.change_count > 1 else ""
             report.append(
-                click.style(f'{self.change_count} file{s} reformatted', bold=True)
+                click.style(f"{self.change_count} file{s} {reformatted}", bold=True)
             )
         if self.same_count:
-            s = 's' if self.same_count > 1 else ''
-            report.append(f'{self.same_count} file{s} left unchanged')
+            s = "s" if self.same_count > 1 else ""
+            report.append(f"{self.same_count} file{s} {unchanged}")
         if self.failure_count:
-            s = 's' if self.failure_count > 1 else ''
+            s = "s" if self.failure_count > 1 else ""
             report.append(
-                click.style(
-                    f'{self.failure_count} file{s} failed to reformat', fg='red'
-                )
+                click.style(f"{self.failure_count} file{s} {failed}", fg="red")
             )
-        return ', '.join(report) + '.'
+        return ", ".join(report) + "."
 
 
 def assert_equivalent(src: str, dst: str) -> None:
-    """Raises AssertionError if `src` and `dst` aren't equivalent.
-
-    This is a temporary sanity check until Black becomes stable.
-    """
+    """Raise AssertionError if `src` and `dst` aren't equivalent."""
 
     import ast
     import traceback
@@ -1438,72 +2371,170 @@ def assert_equivalent(src: str, dst: str) -> None:
     try:
         src_ast = ast.parse(src)
     except Exception as exc:
-        raise AssertionError(f"cannot parse source: {exc}") from None
+        major, minor = sys.version_info[:2]
+        raise AssertionError(
+            f"cannot use --safe with this file; failed to parse source file "
+            f"with Python {major}.{minor}'s builtin AST. Re-run with --fast "
+            f"or stop using deprecated Python 2 syntax. AST error message: {exc}"
+        )
 
     try:
         dst_ast = ast.parse(dst)
     except Exception as exc:
-        log = dump_to_file(''.join(traceback.format_tb(exc.__traceback__)), dst)
+        log = dump_to_file("".join(traceback.format_tb(exc.__traceback__)), dst)
         raise AssertionError(
             f"INTERNAL ERROR: Black produced invalid code: {exc}. "
             f"Please report a bug on https://github.com/ambv/black/issues.  "
-            f"This invalid output might be helpful: {log}",
+            f"This invalid output might be helpful: {log}"
         ) from None
 
-    src_ast_str = '\n'.join(_v(src_ast))
-    dst_ast_str = '\n'.join(_v(dst_ast))
+    src_ast_str = "\n".join(_v(src_ast))
+    dst_ast_str = "\n".join(_v(dst_ast))
     if src_ast_str != dst_ast_str:
-        log = dump_to_file(diff(src_ast_str, dst_ast_str, 'src', 'dst'))
+        log = dump_to_file(diff(src_ast_str, dst_ast_str, "src", "dst"))
         raise AssertionError(
             f"INTERNAL ERROR: Black produced code that is not equivalent to "
             f"the source.  "
             f"Please report a bug on https://github.com/ambv/black/issues.  "
-            f"This diff might be helpful: {log}",
+            f"This diff might be helpful: {log}"
         ) from None
 
 
 def assert_stable(src: str, dst: str, line_length: int) -> None:
-    """Raises AssertionError if `dst` reformats differently the second time.
-
-    This is a temporary sanity check until Black becomes stable.
-    """
+    """Raise AssertionError if `dst` reformats differently the second time."""
     newdst = format_str(dst, line_length=line_length)
     if dst != newdst:
         log = dump_to_file(
-            diff(src, dst, 'source', 'first pass'),
-            diff(dst, newdst, 'first pass', 'second pass'),
+            diff(src, dst, "source", "first pass"),
+            diff(dst, newdst, "first pass", "second pass"),
         )
         raise AssertionError(
             f"INTERNAL ERROR: Black produced different code on the second pass "
             f"of the formatter.  "
             f"Please report a bug on https://github.com/ambv/black/issues.  "
-            f"This diff might be helpful: {log}",
+            f"This diff might be helpful: {log}"
         ) from None
 
 
 def dump_to_file(*output: str) -> str:
-    """Dumps `output` to a temporary file. Returns path to the file."""
+    """Dump `output` to a temporary file. Return path to the file."""
     import tempfile
 
     with tempfile.NamedTemporaryFile(
-        mode='w', prefix='blk_', suffix='.log', delete=False
+        mode="w", prefix="blk_", suffix=".log", delete=False, encoding="utf8"
     ) as f:
         for lines in output:
             f.write(lines)
-            f.write('\n')
+            if lines and lines[-1] != "\n":
+                f.write("\n")
     return f.name
 
 
 def diff(a: str, b: str, a_name: str, b_name: str) -> str:
-    """Returns a udiff string between strings `a` and `b`."""
+    """Return a unified diff string between strings `a` and `b`."""
     import difflib
 
-    a_lines = [line + '\n' for line in a.split('\n')]
-    b_lines = [line + '\n' for line in b.split('\n')]
-    return ''.join(
+    a_lines = [line + "\n" for line in a.split("\n")]
+    b_lines = [line + "\n" for line in b.split("\n")]
+    return "".join(
         difflib.unified_diff(a_lines, b_lines, fromfile=a_name, tofile=b_name, n=5)
     )
 
 
-if __name__ == '__main__':
+def cancel(tasks: List[asyncio.Task]) -> None:
+    """asyncio signal handler that cancels all `tasks` and reports to stderr."""
+    err("Aborted!")
+    for task in tasks:
+        task.cancel()
+
+
+def shutdown(loop: BaseEventLoop) -> None:
+    """Cancel all pending tasks on `loop`, wait for them, and close the loop."""
+    try:
+        # This part is borrowed from asyncio/runners.py in Python 3.7b2.
+        to_cancel = [task for task in asyncio.Task.all_tasks(loop) if not task.done()]
+        if not to_cancel:
+            return
+
+        for task in to_cancel:
+            task.cancel()
+        loop.run_until_complete(
+            asyncio.gather(*to_cancel, loop=loop, return_exceptions=True)
+        )
+    finally:
+        # `concurrent.futures.Future` objects cannot be cancelled once they
+        # are already running. There might be some when the `shutdown()` happened.
+        # Silence their logger's spew about the event loop being closed.
+        cf_logger = logging.getLogger("concurrent.futures")
+        cf_logger.setLevel(logging.CRITICAL)
+        loop.close()
+
+
+def sub_twice(regex: Pattern[str], replacement: str, original: str) -> str:
+    """Replace `regex` with `replacement` twice on `original`.
+
+    This is used by string normalization to perform replaces on
+    overlapping matches.
+    """
+    return regex.sub(replacement, regex.sub(replacement, original))
+
+
+CACHE_DIR = Path(user_cache_dir("black", version=__version__))
+CACHE_FILE = CACHE_DIR / "cache.pickle"
+
+
+def read_cache() -> Cache:
+    """Read the cache if it exists and is well formed.
+
+    If it is not well formed, the call to write_cache later should resolve the issue.
+    """
+    if not CACHE_FILE.exists():
+        return {}
+
+    with CACHE_FILE.open("rb") as fobj:
+        try:
+            cache: Cache = pickle.load(fobj)
+        except pickle.UnpicklingError:
+            return {}
+
+    return cache
+
+
+def get_cache_info(path: Path) -> CacheInfo:
+    """Return the information used to check if a file is already formatted or not."""
+    stat = path.stat()
+    return stat.st_mtime, stat.st_size
+
+
+def filter_cached(
+    cache: Cache, sources: Iterable[Path]
+) -> Tuple[List[Path], List[Path]]:
+    """Split a list of paths into two.
+
+    The first list contains paths of files that modified on disk or are not in the
+    cache. The other list contains paths to non-modified files.
+    """
+    todo, done = [], []
+    for src in sources:
+        src = src.resolve()
+        if cache.get(src) != get_cache_info(src):
+            todo.append(src)
+        else:
+            done.append(src)
+    return todo, done
+
+
+def write_cache(cache: Cache, sources: List[Path]) -> None:
+    """Update the cache file."""
+    try:
+        if not CACHE_DIR.exists():
+            CACHE_DIR.mkdir(parents=True)
+        new_cache = {**cache, **{src.resolve(): get_cache_info(src) for src in sources}}
+        with CACHE_FILE.open("wb") as fobj:
+            pickle.dump(new_cache, fobj, protocol=pickle.HIGHEST_PROTOCOL)
+    except OSError:
+        pass
+
+
+if __name__ == "__main__":
     main()