Source code for asdf.types

# Licensed under a 3-clause BSD style license - see LICENSE.rst
# -*- coding: utf-8 -*-

import re
import inspect
import warnings
import importlib
from collections import defaultdict

from copy import copy

from . import tagged
from . import util
from .versioning import AsdfVersion, AsdfSpec
from .exceptions import AsdfWarning

__all__ = ['format_tag', 'CustomType']

# regex used to parse module name from optional version string
MODULE_RE = re.compile(r'([a-zA-Z]+)(-(\d+\.\d+\.\d+))?')

class AsdfSubclassProperty(property):

[docs]def format_tag(organization, standard, version, tag_name): """ Format a YAML tag. """ tag = 'tag:{0}:{1}/{2}'.format(organization, standard, tag_name) if version is None: return tag if isinstance(version, AsdfSpec): version = str(version.spec) return "{0}-{1}".format(tag, version)
_all_asdftypes = set() def _from_tree_tagged_missing_requirements(cls, tree, ctx): # A special version of AsdfType.from_tree_tagged for when the # required dependencies for an AsdfType are missing. plural, verb = ('s', 'are') if len(cls.requires) else ('', 'is') message = "{0} package{1} {2} required to instantiate '{3}'".format( util.human_list(cls.requires), plural, verb, tree._tag) # This error will be handled by yamlutil.tagged_tree_to_custom_tree, which # will cause a warning to be issued indicating that the tree failed to be # converted. raise TypeError(message) class ExtensionTypeMeta(type): """ Custom class constructor for tag types. """ _import_cache = {} @classmethod def _has_required_modules(cls, requires): for string in requires: has_module = True match = MODULE_RE.match(string) modname, _, version = match.groups() if modname in cls._import_cache: if not cls._import_cache[modname]: return False try: module = importlib.import_module(modname) if version and hasattr(module, '__version__'): if module.__version__ < version: has_module = False except ImportError: has_module = False finally: cls._import_cache[modname] = has_module if not has_module: return False return True @classmethod def _find_in_bases(cls, attrs, bases, name, default=None): if name in attrs: return attrs[name] for base in bases: if hasattr(base, name): return getattr(base, name) return default @property def versioned_siblings(mcls): return getattr(mcls, '__versioned_siblings') or [] def __new__(mcls, name, bases, attrs): requires = mcls._find_in_bases(attrs, bases, 'requires', []) if not mcls._has_required_modules(requires): attrs['from_tree_tagged'] = classmethod( _from_tree_tagged_missing_requirements) attrs['types'] = [] attrs['has_required_modules'] = False else: attrs['has_required_modules'] = True types = mcls._find_in_bases(attrs, bases, 'types', []) new_types = [] for typ in types: if isinstance(typ, str): typ = util.resolve_name(typ) new_types.append(typ) attrs['types'] = new_types cls = super(ExtensionTypeMeta, mcls).__new__(mcls, name, bases, attrs) if hasattr(cls, 'version'): if not isinstance(cls.version, (AsdfVersion, AsdfSpec)): cls.version = AsdfVersion(cls.version) if hasattr(cls, 'name'): if isinstance(, str): if 'yaml_tag' not in attrs: cls.yaml_tag = cls.make_yaml_tag( elif isinstance(, list): pass elif is not None: raise TypeError("name must be string or list") if hasattr(cls, 'supported_versions'): if not isinstance(cls.supported_versions, (list, set)): cls.supported_versions = [cls.supported_versions] supported_versions = set() for version in cls.supported_versions: if not isinstance(version, (AsdfVersion, AsdfSpec)): version = AsdfVersion(version) # This should cause an exception for invalid input supported_versions.add(version) # We need to convert back to a list here so that the 'in' operator # uses actual comparison instead of hash equality cls.supported_versions = list(supported_versions) siblings = list() for version in cls.supported_versions: if version != cls.version: new_attrs = copy(attrs) new_attrs['version'] = version new_attrs['supported_versions'] = set() new_attrs['_latest_version'] = cls.version siblings.append( ExtensionTypeMeta. __new__(mcls, name, bases, new_attrs)) setattr(cls, '__versioned_siblings', siblings) return cls class AsdfTypeMeta(ExtensionTypeMeta): """ Keeps track of `AsdfType` subclasses that are created, and stores them in `AsdfTypeIndex`. """ def __new__(mcls, name, bases, attrs): cls = super(AsdfTypeMeta, mcls).__new__(mcls, name, bases, attrs) # Classes using this metaclass get added to the list of built-in # extensions if name != "AsdfType": _all_asdftypes.add(cls) return cls class ExtensionType: """ The base class of all custom types in the tree. Besides the attributes defined below, most subclasses will also override `to_tree` and `from_tree`. """ name = None organization = '' standard = 'asdf' version = (1, 0, 0) supported_versions = set() types = [] handle_dynamic_subclasses = False validators = {} requires = [] yaml_tag = None _subclass_map = {} _subclass_attr_map = defaultdict(lambda: list()) @classmethod def names(cls): """ Returns the name(s) represented by this tag type as a list. While some tag types represent only a single custom type, others represent multiple types. In the latter case, the `name` attribute of the extension is actually a list, not simply a string. This method normalizes the value of `name` by returning a list in all cases. Returns ------- `list` of names represented by this tag type """ if is None: return None return if isinstance(, list) else [] @classmethod def make_yaml_tag(cls, name, versioned=True): """ Given the name of a type, returns a string representing its YAML tag. Parameters ---------- name : str The name of the type. In most cases this will correspond to the `name` attribute of the tag type. However, it is passed as a parameter since some tag types represent multiple custom types. versioned : bool If `True`, the tag will be versioned. Otherwise, a YAML tag without a version will be returned. Returns ------- `str` representing the YAML tag """ return format_tag( cls.organization, cls.standard, cls.version if versioned else None, name) @classmethod def tag_base(cls): """ Returns the base of the YAML tag for types represented by this class. This method returns the portion of the tag that represents the standard and the organization of any type represented by this class. Returns ------- `str` representing the base of the YAML tag """ return cls.make_yaml_tag('', versioned=False) @classmethod def to_tree(cls, node, ctx): """ Converts instances of custom types into YAML representations. This method should be overridden by custom extension classes in order to define how custom types are serialized into YAML. The method must return a single Python object corresponding to one of the basic YAML types (dict, list, str, or number). However, the types can be nested and combined in order to represent more complex custom types. This method is called as part of the process of writing an `AsdfFile` object. Whenever a custom type (or a subclass of that type) that is listed in the `types` attribute of this class is encountered, this method will be used to serialize that type. The name `to_tree` refers to the act of converting a custom type into part of a YAML object tree. Parameters ---------- node : `object` Instance of a custom type to be serialized. Will be an instance (or an instance of a subclass) of one of the types listed in the `types` attribute of this class. ctx : `AsdfFile` An instance of the `AsdfFile` object that is being written out. Returns ------- A basic YAML type (`dict`, `list`, `str`, `int`, `float`, or `complex`) representing the properties of the custom type to be serialized. These types can be nested in order to represent more complex custom types. """ return node.__class__.__bases__[0](node) @classmethod def to_tree_tagged(cls, node, ctx): """ Converts instances of custom types into tagged objects. It is more common for custom tag types to override `to_tree` instead of this method. This method should be overridden if it is necessary to modify the YAML tag that will be used to tag this object. Parameters ---------- node : `object` Instance of a custom type to be serialized. Will be an instance (or an instance of a subclass) of one of the types listed in the `types` attribute of this class. ctx : `AsdfFile` An instance of the `AsdfFile` object that is being written out. Returns ------- An instance of `asdf.tagged.Tagged`. """ obj = cls.to_tree(node, ctx) node_cls = type(node) cls_name = node_cls.__name__ if node_cls.__name__ in cls._subclass_map and isinstance(obj, dict): from .tags.core import SubclassMetadata from .yamlutil import custom_tree_to_tagged_tree attribute = cls._subclass_map[cls_name][0] subclass = SubclassMetadata(name=cls_name) obj[attribute] = custom_tree_to_tagged_tree(subclass, ctx) if node_cls in cls._subclass_attr_map: if isinstance(obj, dict): for name, member in cls._subclass_attr_map[node_cls]: obj[name] = member.fget(node) else: # TODO: should this be an exception? Should it be a custom warning type? warnings.warn( "Failed to add subclass attribute(s) to node that is " "not an object (is a {}). No subclass attributes are being " "added (tag={}, subclass={})".format( type(obj).__name__, cls, node_cls), AsdfWarning ) return tagged.tag_object(cls.yaml_tag, obj, ctx=ctx) @classmethod def from_tree(cls, tree, ctx): """ Converts basic types representing YAML trees into custom types. This method should be overridden by custom extension classes in order to define how custom types are deserialized from the YAML representation back into their original types. Typically the method will return an instance of the original custom type. It is also permitted to return a generator, which yields a partially constructed result, then completes construction once the generator is drained. This is useful when constructing objects that contain reference cycles. This method is called as part of the process of reading an ASDF file in order to construct an `AsdfFile` object. Whenever a YAML subtree is encountered that has a tag that corresponds to the `yaml_tag` property of this class, this method will be used to deserialize that tree back into an instance of the original custom type. Parameters ---------- tree : `object` representing YAML tree An instance of a basic Python type (possibly nested) that corresponds to a YAML subtree. ctx : `AsdfFile` An instance of the `AsdfFile` object that is being constructed. Returns ------- An instance of the custom type represented by this extension class, or a generator that yields that instance. """ return cls(tree) @classmethod def from_tree_tagged(cls, tree, ctx): """ Converts from tagged tree into custom type. It is more common for extension classes to override `from_tree` instead of this method. This method should only be overridden if it is necessary to access the `_tag` property of the `Tagged` object directly. Parameters ---------- tree : `asdf.tagged.Tagged` object representing YAML tree ctx : `AsdfFile` An instance of the `AsdfFile` object that is being constructed. Returns ------- An instance of the custom type represented by this extension class. """ from .tags.core import SubclassMetadata if isinstance(tree, dict): for k, v in tree.items(): if isinstance(v, SubclassMetadata): tree.pop(k) subclass_name = v['name'] return cls._subclass_map[subclass_name][1](** return cls.from_tree(, ctx) @classmethod def incompatible_version(cls, version): """ Indicates if given version is known to be incompatible with this type. If this tag class explicitly identifies compatible versions then this checks whether a given version is compatible or not (see `supported_versions`). Otherwise, all versions are assumed to be compatible. Child classes can override this method to affect how version compatiblity for this type is determined. Parameters ---------- version : `str` or `~asdf.versioning.AsdfVersion` The version to test for compatibility. """ if cls.supported_versions: if version not in cls.supported_versions: return True return False @classmethod def subclass(cls, *args, attribute='subclass'): """ Decorator to enable serialization of a subclass of an existing type. Use this method to decorate subclasses of custom types that are already handled by an existing ASDF tag class. This enables subclasses of known types to be properly serialized without having to write an entirely separate tag class for the subclass. This feature can only be used for tagged types where the underlying YAML representation of the type is an object (i.e. a Python `dict`). It will not work for nodes that are basic types. The subclass metadata is stored in a new attribute of the YAML node. By default the attribute name is "subclass", but it is customizable by using the optional `attribute` keyword argument of the decorator. The schema of the base custom type is used for validation. This feature will not work if the base schema disallows additional attributes. It is incumbent upon the user to avoid name conflicts with attributes that already exist in the representation of the base custom class. For example, a base class may use the attribute "subclass" for some other purpose, in which case it would be necessary to provide a different custom attribute name here. Parameters ---------- attribute : `str` Custom attribute name used to store subclass metadata in this node. """ def decorator(subclass): cls._subclass_map[subclass.__name__] = (attribute, subclass) for name, member in inspect.getmembers(subclass): if isinstance(member, AsdfSubclassProperty): cls._subclass_attr_map[subclass].append((name, member)) return subclass return decorator(args[0]) if args else decorator @classmethod def subclass_property(cls, attribute): """ Decorator to enable serialization of custom subclass attributes. Use this decorator to serialize attributes that are specific to a subclass of a custom type that is already handled by an existing ASDF tag class. This decorator will only work on subclasses that have been decorated with the `~asdf.AsdfTypes.subclass` decorator. Methods that are decorated in this way are treated as properties (see `property`). The name of the property **must** correspond to a keyword argument of the subclass constructor. The property will be serialized as a YAML object attribute with the same name. Users are responsible for ensuring that any and all additional subclass properties conform to the schema of the base custom type and do not conflict with existing attributes. """ return AsdfSubclassProperty(attribute) class AsdfType(ExtensionType, metaclass=AsdfTypeMeta): """ Base class for all built-in ASDF types. Types that inherit this class will be automatically added to the list of built-ins. This should *not* be used for user-defined extensions. """
[docs]class CustomType(ExtensionType, metaclass=ExtensionTypeMeta): """ Base class for all user-defined types. """ # These attributes are duplicated here with docstrings since a bug in # sphinx prevents the docstrings of class attributes from being inherited # properly (see The # docstrings are not included anywhere else in the class hierarchy since # this class is the only one exposed in the public API. name = None """ `str` or `list`: The name of the type. """ organization = '' """ `str`: The organization responsible for the type. """ standard = 'asdf' """ `str`: The standard the type is defined in. """ version = (1, 0, 0) """ `str`, `tuple`, `AsdfVersion`, or `AsdfSpec`: The version of the type. """ supported_versions = set() """ `set`: Versions that explicitly compatible with this extension class. If provided, indicates explicit compatibility with the given set of versions. Other versions of the same schema that are not included in this set will not be converted to custom types with this class. """ types = [] """ `list`: List of types that this extension class can convert to/from YAML. Custom Python types that, when found in the tree, will be converted into basic types for YAML output. Can be either strings referring to the types or the types themselves.""" handle_dynamic_subclasses = False """ `bool`: Indicates whether dynamically generated subclasses can be serialized Flag indicating whether this type is capable of serializing subclasses of any of the types listed in ``types`` that are generated dynamically. """ validators = {} """ `dict`: Mapping JSON Schema keywords to validation functions for jsonschema. Useful if the type defines extra types of validation that can be performed. """ requires = [] """ `list`: Python packages that are required to instantiate the object. """ yaml_tag = None """ `str`: The YAML tag to use for the type. If not provided, it will be automatically generated from name, organization, standard and version. """ has_required_modules = True """ `bool`: Indicates whether modules specified by `requires` are available. NOTE: This value is automatically generated. Do not set it in subclasses as it will be overwritten. """