AlkantarClanX12
Current Path : /opt/hc_python/lib/python3.8/site-packages/aenum/ |
Current File : //opt/hc_python/lib/python3.8/site-packages/aenum/_tuple.py |
from ._common import * from ._constant import NamedConstant import sys as _sys __all__ = [ 'TupleSize', 'NamedTuple', ] # NamedTuple class NamedTupleDict(OrderedDict): """Track field order and ensure field names are not reused. NamedTupleMeta will use the names found in self._field_names to translate to indices. """ def __init__(self, *args, **kwds): self._field_names = [] super(NamedTupleDict, self).__init__(*args, **kwds) def __setitem__(self, key, value): """Records anything not dundered or not a descriptor. If a field name is used twice, an error is raised. Single underscore (sunder) names are reserved. """ if is_sunder(key): if key not in ('_size_', '_order_', '_fields_', '_review_'): raise ValueError( '_sunder_ names, such as %r, are reserved for future NamedTuple use' % (key, ) ) elif is_dunder(key): if key == '__order__': key = '_order_' elif key in self._field_names: # overwriting a field? raise TypeError('attempt to reuse field name: %r' % (key, )) elif not is_descriptor(value): if key in self: # field overwriting a descriptor? raise TypeError('%s already defined as: %r' % (key, self[key])) self._field_names.append(key) super(NamedTupleDict, self).__setitem__(key, value) class _TupleAttributeAtIndex(object): def __init__(self, name, index, doc, default): self.name = name self.index = index if doc is undefined: doc = None self.__doc__ = doc self.default = default def __get__(self, instance, owner): if instance is None: return self if len(instance) <= self.index: raise AttributeError('%s instance has no value for %s' % (instance.__class__.__name__, self.name)) return instance[self.index] def __repr__(self): return '%s(%d)' % (self.__class__.__name__, self.index) class undefined(object): def __repr__(self): return 'undefined' def __bool__(self): return False __nonzero__ = __bool__ undefined = undefined() class TupleSize(NamedConstant): fixed = constant('fixed', 'tuple length is static') minimum = constant('minimum', 'tuple must be at least x long (x is calculated during creation') variable = constant('variable', 'tuple length can be anything') class NamedTupleMeta(type): """Metaclass for NamedTuple""" @classmethod def __prepare__(metacls, cls, bases, size=undefined, **kwds): return NamedTupleDict() def __init__(cls, *args , **kwds): super(NamedTupleMeta, cls).__init__(*args) def __new__(metacls, cls, bases, clsdict, size=undefined, **kwds): if bases == (object, ): bases = (tuple, object) elif tuple not in bases: if object in bases: index = bases.index(object) bases = bases[:index] + (tuple, ) + bases[index:] else: bases = bases + (tuple, ) # include any fields from base classes base_dict = NamedTupleDict() namedtuple_bases = [] for base in bases: if isinstance(base, NamedTupleMeta): namedtuple_bases.append(base) i = 0 if namedtuple_bases: for name, index, doc, default in metacls._convert_fields(*namedtuple_bases): base_dict[name] = index, doc, default i = max(i, index) # construct properly ordered dict with normalized indexes for k, v in clsdict.items(): base_dict[k] = v original_dict = base_dict if size is not undefined and '_size_' in original_dict: raise TypeError('_size_ cannot be set if "size" is passed in header') add_order = isinstance(clsdict, NamedTupleDict) clsdict = NamedTupleDict() clsdict.setdefault('_size_', size or TupleSize.fixed) unnumbered = OrderedDict() numbered = OrderedDict() _order_ = original_dict.pop('_order_', []) if _order_ : _order_ = _order_.replace(',',' ').split() add_order = False # and process this class for k, v in original_dict.items(): if k not in original_dict._field_names: clsdict[k] = v else: # TODO:normalize v here if isinstance(v, baseinteger): # assume an offset v = v, undefined, undefined i = v[0] + 1 target = numbered elif isinstance(v, basestring): # assume a docstring if add_order: v = i, v, undefined i += 1 target = numbered else: v = undefined, v, undefined target = unnumbered elif isinstance(v, tuple) and len(v) in (2, 3) and isinstance(v[0], baseinteger) and isinstance(v[1], (basestring, NoneType)): # assume an offset, a docstring, and (maybe) a default if len(v) == 2: v = v + (undefined, ) v = v i = v[0] + 1 target = numbered elif isinstance(v, tuple) and len(v) in (1, 2) and isinstance(v[0], (basestring, NoneType)): # assume a docstring, and (maybe) a default if len(v) == 1: v = v + (undefined, ) if add_order: v = (i, ) + v i += 1 target = numbered else: v = (undefined, ) + v target = unnumbered else: # refuse to guess further raise ValueError('not sure what to do with %s=%r (should be OFFSET [, DOC [, DEFAULT]])' % (k, v)) target[k] = v # all index values have been normalized # deal with _order_ (or lack thereof) fields = [] aliases = [] seen = set() max_len = 0 if not _order_: if unnumbered: raise ValueError("_order_ not specified and OFFSETs not declared for %r" % (unnumbered.keys(), )) for name, (index, doc, default) in sorted(numbered.items(), key=lambda nv: (nv[1][0], nv[0])): if index in seen: aliases.append(name) else: fields.append(name) seen.add(index) max_len = max(max_len, index + 1) offsets = numbered else: # check if any unnumbered not in _order_ missing = set(unnumbered) - set(_order_) if missing: raise ValueError("unable to order fields: %s (use _order_ or specify OFFSET" % missing) offsets = OrderedDict() # if any unnumbered, number them from their position in _order_ i = 0 for k in _order_: try: index, doc, default = unnumbered.pop(k, None) or numbered.pop(k) except IndexError: raise ValueError('%s (from _order_) not found in %s' % (k, cls)) if index is not undefined: i = index if i in seen: aliases.append(k) else: fields.append(k) seen.add(i) offsets[k] = i, doc, default i += 1 max_len = max(max_len, i) # now handle anything in numbered for k, (index, doc, default) in sorted(numbered.items(), key=lambda nv: (nv[1][0], nv[0])): if index in seen: aliases.append(k) else: fields.append(k) seen.add(index) offsets[k] = index, doc, default max_len = max(max_len, index+1) # at this point fields and aliases should be ordered lists, offsets should be an # OrdededDict with each value an int, str or None or undefined, default or None or undefined assert len(fields) + len(aliases) == len(offsets), "number of fields + aliases != number of offsets" assert set(fields) & set(offsets) == set(fields), "some fields are not in offsets: %s" % set(fields) & set(offsets) assert set(aliases) & set(offsets) == set(aliases), "some aliases are not in offsets: %s" % set(aliases) & set(offsets) for name, (index, doc, default) in offsets.items(): assert isinstance(index, baseinteger), "index for %s is not an int (%s:%r)" % (name, type(index), index) assert isinstance(doc, (basestring, NoneType)) or doc is undefined, "doc is not a str, None, nor undefined (%s:%r)" % (name, type(doc), doc) # create descriptors for fields for name, (index, doc, default) in offsets.items(): clsdict[name] = _TupleAttributeAtIndex(name, index, doc, default) clsdict['__slots__'] = () # create our new NamedTuple type namedtuple_class = super(NamedTupleMeta, metacls).__new__(metacls, cls, bases, clsdict) namedtuple_class._fields_ = fields namedtuple_class._aliases_ = aliases namedtuple_class._defined_len_ = max_len return namedtuple_class @staticmethod def _convert_fields(*namedtuples): "create list of index, doc, default triplets for cls in namedtuples" all_fields = [] for cls in namedtuples: base = len(all_fields) for field in cls._fields_: desc = getattr(cls, field) all_fields.append((field, base+desc.index, desc.__doc__, desc.default)) return all_fields def __add__(cls, other): "A new NamedTuple is created by concatenating the _fields_ and adjusting the descriptors" if not isinstance(other, NamedTupleMeta): return NotImplemented return NamedTupleMeta('%s%s' % (cls.__name__, other.__name__), (cls, other), {}) def __call__(cls, *args, **kwds): """Creates a new NamedTuple class or an instance of a NamedTuple subclass. NamedTuple should have args of (class_name, names, module) `names` can be: * A string containing member names, separated either with spaces or commas. Values are auto-numbered from 1. * An iterable of member names. Values are auto-numbered from 1. * An iterable of (member name, value) pairs. * A mapping of member name -> value. `module`, if set, will be stored in the new class' __module__ attribute; Note: if `module` is not set this routine will attempt to discover the calling module by walking the frame stack; if this is unsuccessful the resulting class will not be pickleable. subclass should have whatever arguments and/or keywords will be used to create an instance of the subclass """ if cls is NamedTuple or cls._defined_len_ == 0: original_args = args original_kwds = kwds.copy() # create a new subclass try: if 'class_name' in kwds: class_name = kwds.pop('class_name') else: class_name, args = args[0], args[1:] if 'names' in kwds: names = kwds.pop('names') else: names, args = args[0], args[1:] if 'module' in kwds: module = kwds.pop('module') elif args: module, args = args[0], args[1:] else: module = None if 'type' in kwds: type = kwds.pop('type') elif args: type, args = args[0], args[1:] else: type = None except IndexError: raise TypeError('too few arguments to NamedTuple: %s, %s' % (original_args, original_kwds)) if args or kwds: raise TypeError('too many arguments to NamedTuple: %s, %s' % (original_args, original_kwds)) if PY2: # if class_name is unicode, attempt a conversion to ASCII if isinstance(class_name, unicode): try: class_name = class_name.encode('ascii') except UnicodeEncodeError: raise TypeError('%r is not representable in ASCII' % (class_name, )) # quick exit if names is a NamedTuple if isinstance(names, NamedTupleMeta): names.__name__ = class_name if type is not None and type not in names.__bases__: names.__bases__ = (type, ) + names.__bases__ return names metacls = cls.__class__ bases = (cls, ) clsdict = metacls.__prepare__(class_name, bases) # special processing needed for names? if isinstance(names, basestring): names = names.replace(',', ' ').split() if isinstance(names, (tuple, list)) and isinstance(names[0], basestring): names = [(e, i) for (i, e) in enumerate(names)] # Here, names is either an iterable of (name, index) or (name, index, doc, default) or a mapping. item = None # in case names is empty for item in names: if isinstance(item, basestring): # mapping field_name, field_index = item, names[item] else: # non-mapping if len(item) == 2: field_name, field_index = item else: field_name, field_index = item[0], item[1:] clsdict[field_name] = field_index if type is not None: if not isinstance(type, tuple): type = (type, ) bases = type + bases namedtuple_class = metacls.__new__(metacls, class_name, bases, clsdict) # TODO: replace the frame hack if a blessed way to know the calling # module is ever developed if module is None: try: module = _sys._getframe(1).f_globals['__name__'] except (AttributeError, ValueError, KeyError): pass if module is None: make_class_unpicklable(namedtuple_class) else: namedtuple_class.__module__ = module return namedtuple_class else: # instantiate a subclass namedtuple_instance = cls.__new__(cls, *args, **kwds) if isinstance(namedtuple_instance, cls): namedtuple_instance.__init__(*args, **kwds) return namedtuple_instance @bltin_property def __fields__(cls): return list(cls._fields_) # collections.namedtuple compatibility _fields = __fields__ @bltin_property def __aliases__(cls): return list(cls._aliases_) def __repr__(cls): return "<NamedTuple %r>" % (cls.__name__, ) namedtuple_dict = _Addendum( dict=NamedTupleMeta.__prepare__('NamedTuple', (object, )), doc="NamedTuple base class.\n\n Derive from this class to define new NamedTuples.\n\n", ns=globals(), ) @namedtuple_dict def __new__(cls, *args, **kwds): if cls._size_ is TupleSize.fixed and len(args) > cls._defined_len_: raise TypeError('%d fields expected, %d received' % (cls._defined_len_, len(args))) unknown = set(kwds) - set(cls._fields_) - set(cls._aliases_) if unknown: raise TypeError('unknown fields: %r' % (unknown, )) final_args = list(args) + [undefined] * (len(cls.__fields__) - len(args)) for field, value in kwds.items(): index = getattr(cls, field).index if final_args[index] != undefined: raise TypeError('field %s specified more than once' % field) final_args[index] = value cls._review_(final_args) missing = [] for index, value in enumerate(final_args): if value is undefined: # look for default values name = cls.__fields__[index] default = getattr(cls, name).default if default is undefined: missing.append(name) else: final_args[index] = default if missing: if cls._size_ in (TupleSize.fixed, TupleSize.minimum): raise TypeError('values not provided for field(s): %s' % ', '.join(missing)) while final_args and final_args[-1] is undefined: final_args.pop() missing.pop() if cls._size_ is not TupleSize.variable or undefined in final_args: raise TypeError('values not provided for field(s): %s' % ', '.join(missing)) return tuple.__new__(cls, tuple(final_args)) @namedtuple_dict def __reduce_ex__(self, proto): return self.__class__, tuple(getattr(self, f) for f in self._fields_) @namedtuple_dict def __repr__(self): if len(self) == len(self._fields_): return "%s(%s)" % ( self.__class__.__name__, ', '.join(['%s=%r' % (f, o) for f, o in zip(self._fields_, self)]) ) else: return '%s(%s)' % (self.__class__.__name__, ', '.join([repr(o) for o in self])) @namedtuple_dict def __str__(self): return "%s(%s)" % ( self.__class__.__name__, ', '.join(['%r' % (getattr(self, f), ) for f in self._fields_]) ) @namedtuple_dict @bltin_property def _fields_(self): return list(self.__class__._fields_) # compatibility methods with stdlib namedtuple @namedtuple_dict @bltin_property def __aliases__(self): return list(self.__class__._aliases_) @namedtuple_dict @bltin_property def _fields(self): return list(self.__class__._fields_) @namedtuple_dict @classmethod def _make(cls, iterable, new=None, len=None): return cls.__new__(cls, *iterable) @namedtuple_dict def _asdict(self): return OrderedDict(zip(self._fields_, self)) @namedtuple_dict def _replace(self, **kwds): current = self._asdict() current.update(kwds) return self.__class__(**current) @namedtuple_dict @classmethod def _review_(cls, final_args): pass NamedTuple = NamedTupleMeta('NamedTuple', (object, ), namedtuple_dict.resolve()) del namedtuple_dict