blurb.py

#!/usr/bin/env python3
"""Command-line tool to manage CPython Misc/NEWS.d entries."""
__version__ = "1.1.0"

##
## blurb version 1.0
## Part of the blurb package.
## Copyright 2015-2018 by Larry Hastings
##
## Redistribution and use in source and binary forms, with or without
## modification, are permitted provided that the following conditions are
## met:
##
## 1. Redistributions of source code must retain the above copyright
## notice, this list of conditions and the following disclaimer.
##
## 2. Redistributions in binary form must reproduce the above copyright
## notice, this list of conditions and the following disclaimer in the
## documentation and/or other materials provided with the distribution.
##
## 3. Neither the name of the copyright holder nor the names of its
## contributors may be used to endorse or promote products derived from
## this software without specific prior written permission.
##
## THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS
## IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
## TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
## PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
## HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
## SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED
## TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
## PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
## LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
## NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
## SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
##
##
## Licensed to the Python Software Foundation under a contributor agreement.
##


# TODO
#
# automatic git adds and removes

import atexit
import base64
import builtins
import glob
import hashlib
import io
import inspect
import itertools
import math
import os
from pathlib import Path
import re
import shlex
import shutil
import subprocess
import sys
import tempfile
import textwrap
import time
import unittest


#
# This template is the canonical list of acceptable section names!
# It's parsed internally into the "sections" set.
#

template = """

#
# Please enter the relevant GitHub issue number here:
#
.. gh-issue:

#
# Uncomment one of these "section:" lines to specify which section
# this entry should go in in Misc/NEWS.d.
#
#.. section: Security
#.. section: Core and Builtins
#.. section: Library
#.. section: Documentation
#.. section: Tests
#.. section: Build
#.. section: Windows
#.. section: macOS
#.. section: IDLE
#.. section: Tools/Demos
#.. section: C API

# Write your Misc/NEWS.d entry below.  It should be a simple ReST paragraph.
# Don't start with "- Issue #<n>: " or "- gh-issue-<n>: " or that sort of stuff.
###########################################################################


""".lstrip()

root = None
original_dir = None
sections = []

for line in template.split('\n'):
    line = line.strip()
    prefix, found, section = line.partition("#.. section: ")
    if found and not prefix:
        sections.append(section.strip())


def sanitize_section(section):
    """
Cleans up a section string, making it viable as a directory name.
    """
    return section.replace("/", "-")

_unsanitize_section = {
    "Tools-Demos": "Tools/Demos",
    }

def unsanitize_section(section):
    return _unsanitize_section.get(section, section)


def textwrap_body(body, *, subsequent_indent=''):
    """
    Accepts either a string or an iterable of strings.
    (Iterable is assumed to be individual lines.)
    Returns a string.
    """
    if isinstance(body, str):
        text = body
    else:
        text = "\n".join(body).rstrip()

    # textwrap merges paragraphs, ARGH

    # step 1: remove trailing whitespace from individual lines
    #   (this means that empty lines will just have \n, no invisible whitespace)
    lines = []
    for line in text.split("\n"):
        lines.append(line.rstrip())
    text = "\n".join(lines)
    # step 2: break into paragraphs and wrap those
    paragraphs = text.split("\n\n")
    paragraphs2 = []
    kwargs = {'break_long_words': False, 'break_on_hyphens': False}
    if subsequent_indent:
        kwargs['subsequent_indent'] = subsequent_indent
    dont_reflow = False
    for paragraph in paragraphs:
        # don't reflow bulleted / numbered lists
        dont_reflow = dont_reflow or paragraph.startswith(("* ", "1. ", "#. "))
        if dont_reflow:
            initial = kwargs.get("initial_indent", "")
            subsequent = kwargs.get("subsequent_indent", "")
            if initial or subsequent:
                lines = [line.rstrip() for line in paragraph.split("\n")]
                indents = itertools.chain(
                    itertools.repeat(initial, 1),
                    itertools.repeat(subsequent),
                    )
                lines = [indent + line for indent, line in zip(indents, lines)]
                paragraph = "\n".join(lines)
            paragraphs2.append(paragraph)
        else:
            # Why do we reflow the text twice?  Because it can actually change
            # between the first and second reflows, and we want the text to
            # be stable.  The problem is that textwrap.wrap is deliberately
            # dumb about how many spaces follow a period in prose.
            #
            # We're reflowing at 76 columns, but let's pretend it's 30 for
            # illustration purposes.  If we give textwrap.wrap the following
            # text--ignore the line of 30 dashes, that's just to help you
            # with visualization:
            #
            #  ------------------------------
            #  xxxx xxxx xxxx xxxx xxxx.  xxxx
            #
            # The first textwrap.wrap will return this:
            #  "xxxx xxxx xxxx xxxx xxxx.\nxxxx"
            #
            # If we reflow it again, textwrap will rejoin the lines, but
            # only with one space after the period!  So this time it'll
            # all fit on one line, behold:
            #  ------------------------------
            #  xxxx xxxx xxxx xxxx xxxx. xxxx
            # and so it now returns:
            #  "xxxx xxxx xxxx xxxx xxxx. xxxx"
            #
            # textwrap.wrap supports trying to add two spaces after a peroid:
            #    https://docs.python.org/3/library/textwrap.html#textwrap.TextWrapper.fix_sentence_endings
            # But it doesn't work all that well, because it's not smart enough
            # to do a really good job.
            #
            # Since blurbs are eventually turned into ReST and rendered anyway,
            # and since the Zen says "In the face of ambiguity, refuse the
            # temptation to guess", I don't sweat it.  I run textwrap.wrap
            # twice, so it's stable, and this means occasionally it'll
            # convert two spaces to one space, no big deal.

            paragraph = "\n".join(textwrap.wrap(paragraph.strip(), width=76, **kwargs)).rstrip()
            paragraph = "\n".join(textwrap.wrap(paragraph.strip(), width=76, **kwargs)).rstrip()
            paragraphs2.append(paragraph)
        # don't reflow literal code blocks (I hope)
        dont_reflow = paragraph.endswith("::")
        if subsequent_indent:
            kwargs['initial_indent'] = subsequent_indent
    text = "\n\n".join(paragraphs2).rstrip()
    if not text.endswith("\n"):
        text += "\n"
    return text


def current_date():
    return time.strftime("%Y-%m-%d", time.localtime())

def sortable_datetime():
    return time.strftime("%Y-%m-%d-%H-%M-%S", time.localtime())


def prompt(prompt):
    return input(f"[{prompt}> ")

def require_ok(prompt):
    prompt = f"[{prompt}> "
    while True:
        s = input(prompt).strip()
        if s == 'ok':
            return s

class pushd:
    def __init__(self, path):
        self.path = path

    def __enter__(self):
        self.previous_cwd = os.getcwd()
        os.chdir(self.path)

    def __exit__(self, *args):
        os.chdir(self.previous_cwd)


def safe_mkdir(path):
    if not os.path.exists(path):
        os.makedirs(path)


def which(cmd, path="PATH"):
    """Find cmd on PATH."""
    if os.path.exists(cmd):
        return cmd
    if cmd[0] == '/':
        return None
    for segment in os.getenv(path).split(":"):
        program = os.path.normpath(os.path.join(segment, cmd))
        if os.path.exists(program):
            return program
    return None


def strip_whitespace_lines(lines):
    # strip from head
    while lines:
        if lines[0]:
            break
        lines.pop(0)

    # strip from tail
    while lines:
        if lines[-1]:
            return
        lines.pop()


def longest_line(lines):
    longest = 0
    for line in lines:
        longest = max(longest, len(line))
    return longest


def version_key(element):
    fields = list(element.split("."))
    if len(fields) == 1:
        return element

    # in sorted order,
    # 3.5.0a1 < 3.5.0b1 < 3.5.0rc1 < 3.5.0
    # so for sorting purposes we transform
    # "3.5." and "3.5.0" into "3.5.0zz0"
    last = fields.pop()
    for s in ("a", "b", "rc"):
        if s in last:
            last, stage, stage_version = last.partition(s)
            break
    else:
        stage = 'zz'
        stage_version = "0"

    fields.append(last)
    while len(fields) < 3:
        fields.append("0")

    fields.extend([stage, stage_version])
    fields = [s.rjust(6, "0") for s in fields]

    return ".".join(fields)


def nonceify(body):
    digest = hashlib.md5(body.encode("utf-8")).digest()
    return base64.urlsafe_b64encode(digest)[0:6].decode('ascii')


def glob_versions():
    with pushd("Misc/NEWS.d"):
        versions = []
        for wildcard in ("2.*.rst", "3.*.rst", "next"):
            files = [x.partition(".rst")[0] for x in glob.glob(wildcard)]
            versions.extend(files)
    xform = [version_key(x) for x in versions]
    xform.sort(reverse=True)
    versions = sorted(versions, key=version_key, reverse=True)
    return versions


def glob_blurbs(version):
    filenames = []
    base = os.path.join("Misc", "NEWS.d", version)
    if version != "next":
        wildcard = base + ".rst"
        filenames.extend(glob.glob(wildcard))
    else:
        for section in sections:
            wildcard = os.path.join(base, sanitize_section(section), "*.rst")
            entries = glob.glob(wildcard)
            entries.sort(reverse=True)
            deletables = [x for x in entries if x.endswith("/README.rst")]
            for filename in deletables:
                entries.remove(filename)
            filenames.extend(entries)
    return filenames


def printable_version(version):
    if version == "next":
        return version
    if "a" in version:
        return version.replace("a", " alpha ")
    if "b" in version:
        return version.replace("b", " beta ")
    if "rc" in version:
        return version.replace("rc", " release candidate ")
    return version + " final"


class BlurbError(RuntimeError):
    pass

"""

The format of a blurb file:

    ENTRY
    [ENTRY2
    ENTRY3
    ...]

In other words, you may have one or more ENTRYs (entries) in a blurb file.

The format of an ENTRY:

    METADATA
    BODY

The METADATA section is optional.
The BODY section is mandatory and must be non-empty.

Format of the METADATA section:

  * Lines starting with ".." are metadata lines of the format:
        .. name: value
  * Lines starting with "#" are comments:
        # comment line
  * Empty and whitespace-only lines are ignored.
  * Trailing whitespace is removed.  Leading whitespace is not removed
    or ignored.

The first nonblank line that doesn't start with ".." or "#" automatically
terminates the METADATA section and is the first line of the BODY.

Format of the BODY section:

  * The BODY section should be a single paragraph of English text
    in ReST format.  It should not use the following ReST markup
    features:
      * section headers
      * comments
      * directives, citations, or footnotes
      * Any features that require significant line breaks,
        like lists, definition lists, quoted paragraphs, line blocks,
        literal code blocks, and tables.
    Note that this is not (currently) enforced.
  * Trailing whitespace is stripped.  Leading whitespace is preserved.
  * Empty lines between non-empty lines are preserved.
    Trailing empty lines are stripped.
  * The BODY mustn't start with "Issue #", "gh-", or "- ".
    (This formatting will be inserted when rendering the final output.)
  * Lines longer than 76 characters will be wordwrapped.
      * In the final output, the first line will have
        "- gh-issue-<gh-issue-number>: " inserted at the front,
        and subsequent lines will have two spaces inserted
        at the front.

To terminate an ENTRY, specify a line containing only "..".  End of file
also terminates the last ENTRY.

-----------------------------------------------------------------------------

The format of a "next" file is exactly the same, except that we're storing
four pieces of metadata in the filename instead of in the metadata section.
Those four pieces of metadata are: section, gh-issue, date, and nonce.

-----------------------------------------------------------------------------

In addition to the four conventional metadata (section, gh-issue, date, and nonce),
there are two additional metadata used per-version: "release date" and
"no changes".  These may only be present in the metadata block in the *first*
blurb in a blurb file.
  * "release date" is the day a particular version of Python was released.
  * "no changes", if present, notes that there were no actual changes
    for this version.  When used, there are two more things that must be
    true about the the blurb file:
      * There should only be one entry inside the blurb file.
      * That entry's gh-issue number must be 0.

"""

class Blurbs(list):

    def parse(self, text, *, metadata=None, filename="input"):
        """
        Parses a string.  Appends a list of blurb ENTRIES to self, as tuples:
          (metadata, body)
        metadata is a dict.  body is a string.
        """

        metadata = metadata or {}
        body = []
        in_metadata = True

        line_number = None

        def throw(s):
            raise BlurbError(f"Error in {filename}:{line_number}:\n{s}")

        def finish_entry():
            nonlocal body
            nonlocal in_metadata
            nonlocal metadata
            nonlocal self

            if not body:
                throw("Blurb 'body' text must not be empty!")
            text = textwrap_body(body)
            for naughty_prefix in ("- ", "Issue #", "bpo-", "gh-", "gh-issue-"):
                if re.match(naughty_prefix, text, re.I):
                    throw("Blurb 'body' can't start with " + repr(naughty_prefix) + "!")

            no_changes = metadata.get('no changes')
            section = metadata.get('section')

            if not no_changes:
                if not section:
                    throw("No 'section' specified.  You must provide one!")
                elif section not in sections:
                    throw("Invalid 'section'!  You must use one of the predefined sections.")

            issue_number = None

            if metadata.get("gh-issue") is not None:
                try:
                    issue_number = int(metadata.get('gh-issue'))
                except (TypeError, ValueError):
                    throw("Invalid GitHub issue number! (" + repr(issue_number) + ")")
            elif metadata.get("bpo") is not None:
                try:
                    issue_number = int(metadata.get('bpo'))
                except (TypeError, ValueError):
                    throw("Invalid bpo issue number! (" + repr(issue_number) + ")")


            self.append((metadata, text))
            metadata = {}
            body = []
            in_metadata = True

        for line_number, line in enumerate(text.split("\n")):
            line = line.rstrip()
            if in_metadata:
                if line.startswith('..'):
                    line = line[2:].strip()
                    name, colon, value = line.partition(":")
                    assert colon
                    name = name.lower().strip()
                    value = value.strip()
                    if name in metadata:
                        throw("Blurb metadata sets " + repr(name) + " twice!")
                    metadata[name] = value
                    continue
                if line.startswith("#") or not line:
                    continue
                in_metadata = False

            if line == "..":
                finish_entry()
                continue
            body.append(line)

        finish_entry()

    def load(self, filename, *, metadata=None):
        """
Read a blurb file.

Broadly equivalent to blurb.parse(open(filename).read()).
        """
        with open(filename, encoding="utf-8") as file:
            text = file.read()
        self.parse(text, metadata=metadata, filename=filename)

    def __str__(self):
        output = []
        add = output.append
        add_separator = False
        for metadata, body in self:
            if add_separator:
                add("\n..\n\n")
            else:
                add_separator = True
            if metadata:
                for name, value in sorted(metadata.items()):
                    add(f".. {name}: {value}\n")
                add("\n")
            add(textwrap_body(body))
        return "".join(output)

    def save(self, path):
        dirname = os.path.dirname(path)
        safe_mkdir(dirname)

        text = str(self)
        with open(path, "wt", encoding="utf-8") as file:
            file.write(text)

    @staticmethod
    def _parse_next_filename(filename):
        """
Parses a "next" filename into its equivalent blurb metadata.
Returns a dict.
        """
        components = filename.split(os.sep)
        section, filename = components[-2:]
        section = unsanitize_section(section)
        assert section in sections, f"Unknown section {section}"

        fields = [x.strip() for x in filename.split(".")]
        assert len(fields) >= 4, f"Can't parse 'next' filename! filename {filename!r} fields {fields}"
        assert fields[-1] == "rst"

        metadata = {"date": fields[0], "nonce": fields[-2], "section": section}

        for field in fields[1:-2]:
            for name in ("gh-issue","bpo"):
                _, got, value = field.partition(name + "-")
                if got:
                    metadata[name] = value.strip()
                    break
            else:
                assert False, "Found unparsable field in 'next' filename: " + repr(field)

        return metadata

    def load_next(self, filename):
        metadata = self._parse_next_filename(filename)
        o = type(self)()
        o.load(filename, metadata=metadata)
        assert len(o) == 1
        self.extend(o)

    def ensure_metadata(self):
        metadata, body = self[-1]
        assert 'section' in metadata
        for name, default in (
            ("gh-issue", "0"),
            ("bpo", "0"),
            ("date", sortable_datetime()),
            ("nonce", nonceify(body)),
            ):
            if name not in metadata:
                metadata[name] = default

    def _extract_next_filename(self):
        """
        changes metadata!
        """
        self.ensure_metadata()
        metadata, body = self[-1]
        metadata['section'] = sanitize_section(metadata['section'])
        metadata['root'] = root
        if int(metadata["gh-issue"]) > 0 :
            path = "{root}/Misc/NEWS.d/next/{section}/{date}.gh-issue-{gh-issue}.{nonce}.rst".format_map(metadata)
        elif int(metadata["bpo"]) > 0:
            # assume it's a GH issue number
            path = "{root}/Misc/NEWS.d/next/{section}/{date}.bpo-{bpo}.{nonce}.rst".format_map(metadata)
        for name in "root section date gh-issue bpo nonce".split():
            del metadata[name]
        return path


    def save_next(self):
        assert len(self) == 1
        blurb = type(self)()
        metadata, body = self[0]
        metadata = dict(metadata)
        blurb.append((metadata, body))
        filename = blurb._extract_next_filename()
        blurb.save(filename)
        return filename

    def save_split_next(self):
        """
        Save out blurbs created from "blurb split".
        They don't have dates, so we have to get creative.
        """
        filenames = []
        # the "date" MUST have a leading zero.
        # this ensures these files sort after all
        # newly created blurbs.
        width = int(math.ceil(math.log(len(self), 10))) + 1
        i = 1
        blurb = Blurbs()
        while self:
            metadata, body = self.pop()
            metadata['date'] = str(i).rjust(width, '0')
            if 'release date' in metadata:
                del metadata['release date']
            blurb.append((metadata, body))
            filename = blurb._extract_next_filename()
            blurb.save(filename)
            blurb.clear()
            filenames.append(filename)
            i += 1
        return filenames


tests_run = 0

class TestParserPasses(unittest.TestCase):
    directory = "blurb/tests/pass"

    def filename_test(self, filename):
        b = Blurbs()
        b.load(filename)
        self.assertTrue(b)
        if os.path.exists(filename + '.res'):
            with open(filename + '.res', encoding='utf-8') as file:
                expected = file.read()
            self.assertEqual(str(b), expected)

    def test_files(self):
        global tests_run
        with pushd(self.directory):
            for filename in glob.glob("*"):
                if filename[-4:] == '.res':
                    self.assertTrue(os.path.exists(filename[:-4]), filename)
                    continue
                self.filename_test(filename)
                print(".", end="")
                sys.stdout.flush()
                tests_run += 1


class TestParserFailures(TestParserPasses):
    directory = "blurb/tests/fail"

    def filename_test(self, filename):
        b = Blurbs()
        with self.assertRaises(Exception):
            b.load(filename)



def run(s):
    process = subprocess.run(s.split(), capture_output=True)
    process.check_returncode()
    return process.stdout.decode('ascii')


readme_re = re.compile(r"This is Python version [23]\.\d").match

def chdir_to_repo_root():
    global root

    # find the root of the local CPython repo
    # note that we can't ask git, because we might
    # be in an exported directory tree!

    # we intentionally start in a (probably nonexistant) subtree
    # the first thing the while loop does is .., basically
    path = os.path.abspath("garglemox")
    while True:
        next_path = os.path.dirname(path)
        if next_path == path:
            sys.exit('You\'re not inside a CPython repo right now!')
        path = next_path

        os.chdir(path)

        def test_first_line(filename, test):
            if not os.path.exists(filename):
                return False
            with open(filename) as file:
                lines = file.read().split('\n')
                if not (lines and test(lines[0])):
                    return False
            return True

        if not (test_first_line("README", readme_re)
            or test_first_line("README.rst", readme_re)):
            continue

        if not test_first_line("LICENSE",  "A. HISTORY OF THE SOFTWARE".__eq__):
            continue
        if not os.path.exists("Include/Python.h"):
            continue
        if not os.path.exists("Python/ceval.c"):
            continue

        break

    root = path
    return root


def error(*a):
    s = " ".join(str(x) for x in a)
    sys.exit("Error: " + s)


subcommands = {}

def subcommand(fn):
    global subcommands
    name = fn.__name__
    subcommands[name] = fn
    return fn

def get_subcommand(subcommand):
    fn = subcommands.get(subcommand)
    if not fn:
        error(f"Unknown subcommand: {subcommand}\nRun 'blurb help' for help.")
    return fn



@subcommand
def help(subcommand=None):
    """
Print help for subcommands.

Prints the help text for the specified subcommand.
If subcommand is not specified, prints one-line summaries for every command.
    """

    if not subcommand:
        print("blurb version", __version__)
        print()
        print("Management tool for CPython Misc/NEWS and Misc/NEWS.d entries.")
        print()
        print("Usage:")
        print("    blurb [subcommand] [options...]")
        print()

        # print list of subcommands
        summaries = []
        longest_name_len = -1
        for name, fn in subcommands.items():
            if name.startswith('-'):
                continue
            longest_name_len = max(longest_name_len, len(name))
            if not fn.__doc__:
                error("help is broken, no docstring for " + fn.__name__)
            fields = fn.__doc__.lstrip().split("\n")
            if not fields:
                first_line = "(no help available)"
            else:
                first_line = fields[0]
            summaries.append((name, first_line))
        summaries.sort()

        print("Available subcommands:")
        print()
        for name, summary in summaries:
            print(" ", name.ljust(longest_name_len), " ", summary)

        print()
        print("If blurb is run without any arguments, this is equivalent to 'blurb add'.")

        sys.exit(0)

    fn = get_subcommand(subcommand)
    doc = fn.__doc__.strip()
    if not doc:
        error("help is broken, no docstring for " + subcommand)

    options = []
    positionals = []

    nesting = 0
    for name, p in inspect.signature(fn).parameters.items():
        if p.kind == inspect.Parameter.KEYWORD_ONLY:
            short_option = name[0]
            options.append(f" [-{short_option}|--{name}]")
        elif p.kind == inspect.Parameter.POSITIONAL_OR_KEYWORD:
            positionals.append(" ")
            has_default = (p.default != inspect._empty)
            if has_default:
                positionals.append("[")
                nesting += 1
            positionals.append(f"<{name}>")
    positionals.append("]" * nesting)


    parameters = "".join(options + positionals)
    print(f"blurb {subcommand}{parameters}")
    print()
    print(doc)
    sys.exit(0)

# Make "blurb --help" work.
subcommands["--help"] = help


@subcommand
def test(*args):
    """
Run unit tests.  Only works inside source repo, not when installed.
    """
    # unittest.main doesn't work because this isn't a module
    # so we'll do it ourselves

    print("-" * 79)

    for clsname, cls in sorted(globals().items()):
        if clsname.startswith("Test") and isinstance(cls, type):
            o = cls()
            for fnname in sorted(dir(o)):
                if fnname.startswith("test"):
                    fn = getattr(o, fnname)
                    if callable(fn):
                        fn()
    print()
    print(tests_run, "tests passed.")


def find_editor():
    for var in 'GIT_EDITOR', 'EDITOR':
        editor = os.environ.get(var)
        if editor is not None:
            return editor
    if sys.platform == 'win32':
        fallbacks = ['notepad.exe']
    else:
        fallbacks = ['/etc/alternatives/editor', 'nano']
    for fallback in fallbacks:
        if os.path.isabs(fallback):
            found_path = fallback
        else:
            found_path = shutil.which(fallback)
        if found_path and os.path.exists(found_path):
            return found_path
    error('Could not find an editor! Set the EDITOR environment variable.')


@subcommand
def add():
    """
Add a blurb (a Misc/NEWS.d/next entry) to the current CPython repo.
    """

    editor = find_editor()

    handle, tmp_path = tempfile.mkstemp(".rst")
    os.close(handle)
    atexit.register(lambda : os.unlink(tmp_path))

    def init_tmp_with_template():
        with open(tmp_path, "wt", encoding="utf-8") as file:
            # hack:
            # my editor likes to strip trailing whitespace from lines.
            # normally this is a good idea.  but in the case of the template
            # it's unhelpful.
            # so, manually ensure there's a space at the end of the gh-issue line.
            text = template

            issue_line = ".. gh-issue:"
            without_space = "\n" + issue_line + "\n"
            with_space = "\n" + issue_line + " \n"
            if without_space not in text:
                sys.exit("Can't find gh-issue line to ensure there's a space on the end!")
            text = text.replace(without_space, with_space)
            file.write(text)

    init_tmp_with_template()

    # We need to be clever about EDITOR.
    # On the one hand, it might be a legitimate path to an
    #   executable containing spaces.
    # On the other hand, it might be a partial command-line
    #   with options.
    if shutil.which(editor):
        args = [editor]
    else:
        args = list(shlex.split(editor))
        if not shutil.which(args[0]):
            sys.exit(f"Invalid GIT_EDITOR / EDITOR value: {editor}")
    args.append(tmp_path)

    while True:
        subprocess.run(args)

        failure = None
        blurb = Blurbs()
        try:
            blurb.load(tmp_path)
        except BlurbError as e:
            failure = str(e)

        if not failure:
            assert len(blurb) # if parse_blurb succeeds, we should always have a body
            if len(blurb) > 1:
                failure = "Too many entries!  Don't specify '..' on a line by itself."

        if failure:
            print()
            print(f"Error: {failure}")
            print()
            try:
                prompt("Hit return to retry (or Ctrl-C to abort)")
            except KeyboardInterrupt:
                print()
                return
            print()
            continue
        break

    path = blurb.save_next()
    git_add_files.append(path)
    flush_git_add_files()
    print("Ready for commit.")



@subcommand
def release(version):
    """
Move all new blurbs to a single blurb file for the release.

This is used by the release manager when cutting a new release.
    """
    if version == ".":
        # harvest version number from dirname of repo
        # I remind you, we're in the Misc subdir right now
        version = os.path.basename(root)

    existing_filenames = glob_blurbs(version)
    if existing_filenames:
        error("Sorry, can't handle appending 'next' files to an existing version (yet).")

    output = f"Misc/NEWS.d/{version}.rst"
    filenames = glob_blurbs("next")
    blurbs = Blurbs()
    date = current_date()

    if not filenames:
        print(f"No blurbs found.  Setting {version} as having no changes.")
        body = f"There were no new changes in version {version}.\n"
        metadata = {"no changes": "True", "gh-issue": "0", "section": "Library", "date": date, "nonce": nonceify(body)}
        blurbs.append((metadata, body))
    else:
        no_changes = None
        count = len(filenames)
        print(f'Merging {count} blurbs to "{output}".')

        for filename in filenames:
            if not filename.endswith(".rst"):
                continue
            blurbs.load_next(filename)

        metadata = blurbs[0][0]

    metadata['release date'] = date
    print("Saving.")

    blurbs.save(output)
    git_add_files.append(output)
    flush_git_add_files()

    how_many = len(filenames)
    print(f"Removing {how_many} 'next' files from git.")
    git_rm_files.extend(filenames)
    flush_git_rm_files()

    # sanity check: ensuring that saving/reloading the merged blurb file works.
    blurbs2 = Blurbs()
    blurbs2.load(output)
    assert blurbs2 == blurbs, f"Reloading {output} isn't reproducible?!"

    print()
    print("Ready for commit.")



@subcommand
def merge(output=None, *, forced=False):
    """
Merge all blurbs together into a single Misc/NEWS file.

Optional output argument specifies where to write to.
Default is <cpython-root>/Misc/NEWS.

If overwriting, blurb merge will prompt you to make sure it's okay.
To force it to overwrite, use -f.
    """
    if output:
        output = os.path.join(original_dir, output)
    else:
        output = "Misc/NEWS"

    versions = glob_versions()
    if not versions:
        sys.exit("You literally don't have ANY blurbs to merge together!")

    if os.path.exists(output) and not forced:
        builtins.print("You already have a", repr(output), "file.")
        require_ok("Type ok to overwrite")

    write_news(output, versions=versions)


def write_news(output, *, versions):
    buff = io.StringIO()

    def print(*a, sep=" "):
        s = sep.join(str(x) for x in a)
        return builtins.print(s, file=buff)

    print ("""
+++++++++++
Python News
+++++++++++

""".strip())

    for version in versions:
        filenames = glob_blurbs(version)

        blurbs = Blurbs()
        if version == "next":
            for filename in filenames:
                if os.path.basename(filename) == "README.rst":
                    continue
                blurbs.load_next(filename)
            if not blurbs:
                continue
            metadata = blurbs[0][0]
            metadata['release date'] = "XXXX-XX-XX"
        else:
            assert len(filenames) == 1
            blurbs.load(filenames[0])

        header = "What's New in Python " + printable_version(version) + "?"
        print()
        print(header)
        print("=" * len(header))
        print()


        metadata, body = blurbs[0]
        release_date = metadata["release date"]

        print(f"*Release date: {release_date}*")
        print()

        if "no changes" in metadata:
            print(body)
            print()
            continue

        last_section = None
        for metadata, body in blurbs:
            section = metadata['section']
            if last_section != section:
                last_section = section
                print(section)
                print("-" * len(section))
                print()
            if metadata.get("gh-issue"):
                issue_number = metadata['gh-issue']
                if int(issue_number):
                    body = "gh-" + issue_number + ": " + body
            elif metadata.get("bpo"):
                issue_number = metadata['bpo']
                if int(issue_number):
                    body = "bpo-" + issue_number + ": " + body

            body = "- " + body
            text = textwrap_body(body, subsequent_indent='  ')
            print(text)
    print()
    print("**(For information about older versions, consult the HISTORY file.)**")


    new_contents = buff.getvalue()

    # Only write in `output` if the contents are different
    # This speeds up subsequent Sphinx builds
    try:
        previous_contents = Path(output).read_text(encoding="UTF-8")
    except (FileNotFoundError, UnicodeError):
        previous_contents = None
    if new_contents != previous_contents:
        Path(output).write_text(new_contents, encoding="UTF-8")
    else:
        builtins.print(output, "is already up to date")


git_add_files = []
def flush_git_add_files():
    if git_add_files:
        subprocess.run(["git", "add", "--force", *git_add_files]).check_returncode()
        git_add_files.clear()

git_rm_files = []
def flush_git_rm_files():
    if git_rm_files:
        try:
            subprocess.run(["git", "rm", "--quiet", "--force", *git_rm_files]).check_returncode()
        except subprocess.CalledProcessError:
            pass

        # clean up
        for path in git_rm_files:
            try:
                os.unlink(path)
            except FileNotFoundError:
                pass

        git_rm_files.clear()


# @subcommand
# def noop():
#     "Do-nothing command.  Used for blurb smoke-testing."
#     pass


@subcommand
def populate():
    """
Creates and populates the Misc/NEWS.d directory tree.
    """
    os.chdir("Misc")
    safe_mkdir("NEWS.d/next")

    for section in sections:
        dir_name = sanitize_section(section)
        dir_path = f"NEWS.d/next/{dir_name}"
        safe_mkdir(dir_path)
        readme_path = f"NEWS.d/next/{dir_name}/README.rst"
        with open(readme_path, "wt", encoding="utf-8") as readme:
            readme.write(f"Put news entry ``blurb`` files for the *{section}* section in this directory.\n")
        git_add_files.append(dir_path)
        git_add_files.append(readme_path)
    flush_git_add_files()


@subcommand
def export():
    """
Removes blurb data files, for building release tarballs/installers.
    """
    os.chdir("Misc")
    shutil.rmtree("NEWS.d", ignore_errors=True)



# @subcommand
# def arg(*, boolean=False, option=True):
#    """
#    Test function for blurb command-line processing.
#    """
#    print(f"arg: boolean {boolean} option {option}")


def main():
    global original_dir

    args = sys.argv[1:]

    if not args:
        args = ["add"]
    elif args[0] == "-h":
        # slight hack
        args[0] = "help"

    subcommand = args[0]
    args = args[1:]

    fn = get_subcommand(subcommand)

    # hack
    if fn in (test, help):
        sys.exit(fn(*args))

    try:
        original_dir = os.getcwd()
        chdir_to_repo_root()

        # map keyword arguments to options
        # we only handle boolean options
        # and they must have default values
        short_options = {}
        long_options = {}
        kwargs = {}
        for name, p in inspect.signature(fn).parameters.items():
            if p.kind == inspect.Parameter.KEYWORD_ONLY:
                assert isinstance(p.default, bool), "blurb command-line processing only handles boolean options"
                kwargs[name] = p.default
                short_options[name[0]] = name
                long_options[name] = name

        filtered_args = []
        done_with_options = False

        def handle_option(s, dict):
            name = dict.get(s, None)
            if not name:
                sys.exit(f'blurb: Unknown option for {subcommand}: "{s}"')
            kwargs[name] = not kwargs[name]

        # print(f"short_options {short_options} long_options {long_options}")
        for a in args:
            if done_with_options:
                filtered_args.append(a)
                continue
            if a.startswith('-'):
                if a == "--":
                    done_with_options = True
                elif a.startswith("--"):
                    handle_option(a[2:], long_options)
                else:
                    for s in a[1:]:
                        handle_option(s, short_options)
                continue
            filtered_args.append(a)


        sys.exit(fn(*filtered_args, **kwargs))
    except TypeError as e:
        # almost certainly wrong number of arguments.
        # count arguments of function and print appropriate error message.
        specified = len(args)
        required = optional = 0
        for p in inspect.signature(fn).parameters.values():
            if p.default == inspect._empty:
                required += 1
            else:
                optional += 1
        total = required + optional

        if required <= specified <= total:
            # whoops, must be a real type error, reraise
            raise e

        how_many = f"{specified} argument"
        if specified != 1:
            how_many += "s"

        if total == 0:
            middle = "accepts no arguments"
        else:
            if total == required:
                middle = "requires"
            else:
                plural = "" if required == 1 else "s"
                middle = f"requires at least {required} argument{plural} and at most"
            middle += f" {total} argument"
            if total != 1:
                middle += "s"

        print(f'Error: Wrong number of arguments!\n\nblurb {subcommand} {middle},\nand you specified {how_many}.')
        print()
        print("usage: ", end="")
        help(subcommand)


if __name__ == '__main__':
    main()