view hgdemandimport/demandimportpy3.py @ 35665:1ad1e59b405e

lfs: control tracked file selection via a tracked file

Since the lfs tracking policy can dramatically affect the repository, it makes
more sense to have the policy file checked in, than to rely on all developers
configuring their .hgrc properly.  The inspiration for this is the .hgeol file.
The configuration lives under '[track]', so that other things can be added in
the future.  Eventually, the config option should be limited to `convert` only.

If the file can't be parsed for any reason (including unrecognized elements of
the minifileset language), the commit will abort until the problem is corrected.
This seems more useful than the warning that hgeol emits, and has no effect on
reading the data, so there's no compatibility concerns.

My initial thought was to read the file and change each "key = value" line into
"((key) & (value))", so that each line could be ORed together, and make a single
pass at compiling.  Unfortunately, that prevents exclusions if there's a
catchall rule.  Consider what happens to a large *.c file here:

  [track]
  **.c = none()
  ** = size('>1MB')
  # ((**.c) & (none())) | ((**) & (size('>1MB'))) => anything > 1MB

I also thought about having separate [include] and [exclude] sections.  But that
just seems to open things up to user mistakes.  Consider:

  [include]
  **.zip = all()
  **.php = size('>10MB')

  [exclude]
  **.zip = all()  # Who wins?
  **.php = none() # Effectively 'all()' (i.e. nothing excluded), or >10MB ?

Therefore, it just compiles each key and value separately, and walks until the
key matches something.  I'm not sure how to enforce just file patterns on LHS
without leaking knowledge about the minifileset here.  That means this will
allow odd looking lines like this:

  [track]
  **.c | **.txt = none()

But that's also fewer lines to compile, so slightly more efficient?  Some things
like 'none()' won't work as expected on LHS though, because that won't match, so
that line is skipped.  For now, these quirks are not mentioned in the
documentation.

Jun previously expressed concern about efficiency when scaling to large repos,
so I tried avoiding 'repo[None]'.  (localrepo.commit() gets repo[None] already,
but doesn't tie it to the workingcommitctx used here.)  Therefore, I looked at
the passed context for 'AMR' status.  But that doesn't help with the normal case
where the policy file is tracked, but clean.  That requires looking up p1() to
read the file.  I don't see any way to get the content of one file without first
creating the full parent context.

# demandimportpy3 - global demand-loading of modules for Mercurial
#
# Copyright 2017 Facebook Inc.
#
# This software may be used and distributed according to the terms of the
# GNU General Public License version 2 or any later version.

"""Lazy loading for Python 3.6 and above.

This uses the new importlib finder/loader functionality available in Python 3.5
and up. The code reuses most of the mechanics implemented inside importlib.util,
but with a few additions:

* Allow excluding certain modules from lazy imports.
* Expose an interface that's substantially the same as demandimport for
  Python 2.

This also has some limitations compared to the Python 2 implementation:

* Much of the logic is per-package, not per-module, so any packages loaded
  before demandimport is enabled will not be lazily imported in the future. In
  practice, we only expect builtins to be loaded before demandimport is
  enabled.
"""

# This line is unnecessary, but it satisfies test-check-py3-compat.t.
from __future__ import absolute_import

import contextlib
import importlib.abc
import importlib.machinery
import importlib.util
import sys

_deactivated = False

class _lazyloaderex(importlib.util.LazyLoader):
    """This is a LazyLoader except it also follows the _deactivated global and
    the ignore list.
    """
    def exec_module(self, module):
        """Make the module load lazily."""
        if _deactivated or module.__name__ in ignore:
            self.loader.exec_module(module)
        else:
            super().exec_module(module)

# This is 3.6+ because with Python 3.5 it isn't possible to lazily load
# extensions. See the discussion in https://bugs.python.org/issue26186 for more.
_extensions_loader = _lazyloaderex.factory(
    importlib.machinery.ExtensionFileLoader)
_bytecode_loader = _lazyloaderex.factory(
    importlib.machinery.SourcelessFileLoader)
_source_loader = _lazyloaderex.factory(importlib.machinery.SourceFileLoader)

def _makefinder(path):
    return importlib.machinery.FileFinder(
        path,
        # This is the order in which loaders are passed in in core Python.
        (_extensions_loader, importlib.machinery.EXTENSION_SUFFIXES),
        (_source_loader, importlib.machinery.SOURCE_SUFFIXES),
        (_bytecode_loader, importlib.machinery.BYTECODE_SUFFIXES),
    )

ignore = []

def init(ignorelist):
    global ignore
    ignore = ignorelist

def isenabled():
    return _makefinder in sys.path_hooks and not _deactivated

def disable():
    try:
        while True:
            sys.path_hooks.remove(_makefinder)
    except ValueError:
        pass

def enable():
    sys.path_hooks.insert(0, _makefinder)

@contextlib.contextmanager
def deactivated():
    # This implementation is a bit different from Python 2's. Python 3
    # maintains a per-package finder cache in sys.path_importer_cache (see
    # PEP 302). This means that we can't just call disable + enable.
    # If we do that, in situations like:
    #
    #   demandimport.enable()
    #   ...
    #   from foo.bar import mod1
    #   with demandimport.deactivated():
    #       from foo.bar import mod2
    #
    # mod2 will be imported lazily. (The converse also holds -- whatever finder
    # first gets cached will be used.)
    #
    # Instead, have a global flag the LazyLoader can use.
    global _deactivated
    demandenabled = isenabled()
    if demandenabled:
        _deactivated = True
    try:
        yield
    finally:
        if demandenabled:
            _deactivated = False