Source code for dask.delayed

from __future__ import annotations

import operator
import types
import uuid
import warnings
from collections.abc import Sequence
from dataclasses import fields, is_dataclass, replace
from functools import partial

from tlz import concat, curry, merge, unique

from dask import config
from dask.base import (
    DaskMethodsMixin,
    dont_optimize,
    is_dask_collection,
    named_schedulers,
    replace_name_in_key,
)
from dask.base import tokenize as _tokenize
from dask.context import globalmethod
from dask.core import flatten, quote
from dask.highlevelgraph import HighLevelGraph
from dask.typing import Graph, NestedKeys
from dask.utils import (
    OperatorMethodMixin,
    apply,
    funcname,
    is_namedtuple_instance,
    methodcaller,
)

__all__ = ["Delayed", "delayed"]


DEFAULT_GET = named_schedulers.get("threads", named_schedulers["sync"])


def unzip(ls, nout):
    """Unzip a list of lists into ``nout`` outputs."""
    out = list(zip(*ls))
    if not out:
        out = [()] * nout
    return out


def finalize(collection):
    assert is_dask_collection(collection)

    name = "finalize-" + tokenize(collection)
    keys = collection.__dask_keys__()
    finalize, args = collection.__dask_postcompute__()
    layer = {name: (finalize, keys) + args}
    graph = HighLevelGraph.from_collections(name, layer, dependencies=[collection])
    return Delayed(name, graph)


def unpack_collections(expr):
    """Normalize a python object and merge all sub-graphs.

    - Replace ``Delayed`` with their keys
    - Convert literals to things the schedulers can handle
    - Extract dask graphs from all enclosed values

    Parameters
    ----------
    expr : object
        The object to be normalized. This function knows how to handle
        dask collections, as well as most builtin python types.

    Returns
    -------
    task : normalized task to be run
    collections : a tuple of collections

    Examples
    --------
    >>> import dask
    >>> a = delayed(1, 'a')
    >>> b = delayed(2, 'b')
    >>> task, collections = unpack_collections([a, b, 3])
    >>> task
    ['a', 'b', 3]
    >>> collections
    (Delayed('a'), Delayed('b'))

    >>> task, collections = unpack_collections({a: 1, b: 2})
    >>> task
    (<class 'dict'>, [['a', 1], ['b', 2]])
    >>> collections
    (Delayed('a'), Delayed('b'))
    """
    if isinstance(expr, Delayed):
        return expr._key, (expr,)

    if is_dask_collection(expr):
        if hasattr(expr, "optimize"):
            # Optimize dask-expr collections
            expr = expr.optimize()

        finalized = finalize(expr)
        return finalized._key, (finalized,)

    if type(expr) is type(iter(list())):
        expr = list(expr)
    elif type(expr) is type(iter(tuple())):
        expr = tuple(expr)
    elif type(expr) is type(iter(set())):
        expr = set(expr)

    typ = type(expr)

    if typ in (list, tuple, set):
        args, collections = unzip((unpack_collections(e) for e in expr), 2)
        args = list(args)
        collections = tuple(unique(concat(collections), key=id))
        # Ensure output type matches input type
        if typ is not list:
            args = (typ, args)
        return args, collections

    if typ is dict:
        args, collections = unpack_collections([[k, v] for k, v in expr.items()])
        return (dict, args), collections

    if typ is slice:
        args, collections = unpack_collections([expr.start, expr.stop, expr.step])
        return (slice, *args), collections

    if is_dataclass(expr):
        args, collections = unpack_collections(
            [
                [f.name, getattr(expr, f.name)]
                for f in fields(expr)
                if hasattr(expr, f.name)  # if init=False, field might not exist
            ]
        )
        if not collections:
            return expr, ()
        try:
            _fields = {
                f.name: getattr(expr, f.name)
                for f in fields(expr)
                if hasattr(expr, f.name)
            }
            replace(expr, **_fields)
        except (TypeError, ValueError) as e:
            if isinstance(e, ValueError) or "is declared with init=False" in str(e):
                raise ValueError(
                    f"Failed to unpack {typ} instance. "
                    "Note that using fields with `init=False` are not supported."
                ) from e
            else:
                raise TypeError(
                    f"Failed to unpack {typ} instance. "
                    "Note that using a custom __init__ is not supported."
                ) from e
        return (apply, typ, (), (dict, args)), collections

    if is_namedtuple_instance(expr):
        args, collections = unpack_collections([v for v in expr])
        return (typ, *args), collections

    return expr, ()


def to_task_dask(expr):
    """Normalize a python object and merge all sub-graphs.

    - Replace ``Delayed`` with their keys
    - Convert literals to things the schedulers can handle
    - Extract dask graphs from all enclosed values

    Parameters
    ----------
    expr : object
        The object to be normalized. This function knows how to handle
        ``Delayed``s, as well as most builtin python types.

    Returns
    -------
    task : normalized task to be run
    dask : a merged dask graph that forms the dag for this task

    Examples
    --------
    >>> import dask
    >>> a = delayed(1, 'a')
    >>> b = delayed(2, 'b')
    >>> task, dask = to_task_dask([a, b, 3])  # doctest: +SKIP
    >>> task  # doctest: +SKIP
    ['a', 'b', 3]
    >>> dict(dask)  # doctest: +SKIP
    {'a': 1, 'b': 2}

    >>> task, dasks = to_task_dask({a: 1, b: 2})  # doctest: +SKIP
    >>> task  # doctest: +SKIP
    (dict, [['a', 1], ['b', 2]])
    >>> dict(dask)  # doctest: +SKIP
    {'a': 1, 'b': 2}
    """
    warnings.warn(
        "The dask.delayed.to_dask_dask function has been "
        "Deprecated in favor of unpack_collections",
        stacklevel=2,
    )

    if isinstance(expr, Delayed):
        return expr.key, expr.dask

    if is_dask_collection(expr):
        name = "finalize-" + tokenize(expr, pure=True)
        keys = expr.__dask_keys__()
        opt = getattr(expr, "__dask_optimize__", dont_optimize)
        finalize, args = expr.__dask_postcompute__()
        dsk = {name: (finalize, keys) + args}
        dsk.update(opt(expr.__dask_graph__(), keys))
        return name, dsk

    if type(expr) is type(iter(list())):
        expr = list(expr)
    elif type(expr) is type(iter(tuple())):
        expr = tuple(expr)
    elif type(expr) is type(iter(set())):
        expr = set(expr)
    typ = type(expr)

    if typ in (list, tuple, set):
        args, dasks = unzip((to_task_dask(e) for e in expr), 2)
        args = list(args)
        dsk = merge(dasks)
        # Ensure output type matches input type
        return (args, dsk) if typ is list else ((typ, args), dsk)

    if typ is dict:
        args, dsk = to_task_dask([[k, v] for k, v in expr.items()])
        return (dict, args), dsk

    if is_dataclass(expr):
        args, dsk = to_task_dask(
            [
                [f.name, getattr(expr, f.name)]
                for f in fields(expr)
                if hasattr(expr, f.name)  # if init=False, field might not exist
            ]
        )

        return (apply, typ, (), (dict, args)), dsk

    if is_namedtuple_instance(expr):
        args, dsk = to_task_dask([v for v in expr])
        return (typ, *args), dsk

    if typ is slice:
        args, dsk = to_task_dask([expr.start, expr.stop, expr.step])
        return (slice,) + tuple(args), dsk

    return expr, {}


def tokenize(*args, pure=None, **kwargs):
    """Mapping function from task -> consistent name.

    Parameters
    ----------
    args : object
        Python objects that summarize the task.
    pure : boolean, optional
        If True, a consistent hash function is tried on the input. If this
        fails, then a unique identifier is used. If False (default), then a
        unique identifier is always used.
    """
    if pure is None:
        pure = config.get("delayed_pure", False)

    if pure:
        return _tokenize(*args, **kwargs)
    else:
        return str(uuid.uuid4())


[docs]@curry def delayed(obj, name=None, pure=None, nout=None, traverse=True): """Wraps a function or object to produce a ``Delayed``. ``Delayed`` objects act as proxies for the object they wrap, but all operations on them are done lazily by building up a dask graph internally. Parameters ---------- obj : object The function or object to wrap name : Dask key, optional The key to use in the underlying graph for the wrapped object. Defaults to hashing content. Note that this only affects the name of the object wrapped by this call to delayed, and *not* the output of delayed function calls - for that use ``dask_key_name=`` as described below. .. note:: Because this ``name`` is used as the key in task graphs, you should ensure that it uniquely identifies ``obj``. If you'd like to provide a descriptive name that is still unique, combine the descriptive name with :func:`dask.base.tokenize` of the ``array_like``. See :ref:`graphs` for more. pure : bool, optional Indicates whether calling the resulting ``Delayed`` object is a pure operation. If True, arguments to the call are hashed to produce deterministic keys. If not provided, the default is to check the global ``delayed_pure`` setting, and fallback to ``False`` if unset. nout : int, optional The number of outputs returned from calling the resulting ``Delayed`` object. If provided, the ``Delayed`` output of the call can be iterated into ``nout`` objects, allowing for unpacking of results. By default iteration over ``Delayed`` objects will error. Note, that ``nout=1`` expects ``obj`` to return a tuple of length 1, and consequently for ``nout=0``, ``obj`` should return an empty tuple. traverse : bool, optional By default dask traverses builtin python collections looking for dask objects passed to ``delayed``. For large collections this can be expensive. If ``obj`` doesn't contain any dask objects, set ``traverse=False`` to avoid doing this traversal. Examples -------- Apply to functions to delay execution: >>> from dask import delayed >>> def inc(x): ... return x + 1 >>> inc(10) 11 >>> x = delayed(inc, pure=True)(10) >>> type(x) == Delayed True >>> x.compute() 11 Can be used as a decorator: >>> @delayed(pure=True) ... def add(a, b): ... return a + b >>> add(1, 2).compute() 3 ``delayed`` also accepts an optional keyword ``pure``. If False, then subsequent calls will always produce a different ``Delayed``. This is useful for non-pure functions (such as ``time`` or ``random``). >>> from random import random >>> out1 = delayed(random, pure=False)() >>> out2 = delayed(random, pure=False)() >>> out1.key == out2.key False If you know a function is pure (output only depends on the input, with no global state), then you can set ``pure=True``. This will attempt to apply a consistent name to the output, but will fallback on the same behavior of ``pure=False`` if this fails. >>> @delayed(pure=True) ... def add(a, b): ... return a + b >>> out1 = add(1, 2) >>> out2 = add(1, 2) >>> out1.key == out2.key True Instead of setting ``pure`` as a property of the callable, you can also set it contextually using the ``delayed_pure`` setting. Note that this influences the *call* and not the *creation* of the callable: >>> @delayed ... def mul(a, b): ... return a * b >>> import dask >>> with dask.config.set(delayed_pure=True): ... print(mul(1, 2).key == mul(1, 2).key) True >>> with dask.config.set(delayed_pure=False): ... print(mul(1, 2).key == mul(1, 2).key) False The key name of the result of calling a delayed object is determined by hashing the arguments by default. To explicitly set the name, you can use the ``dask_key_name`` keyword when calling the function: >>> add(1, 2) # doctest: +SKIP Delayed('add-3dce7c56edd1ac2614add714086e950f') >>> add(1, 2, dask_key_name='three') Delayed('three') Note that objects with the same key name are assumed to have the same result. If you set the names explicitly you should make sure your key names are different for different results. >>> add(1, 2, dask_key_name='three') Delayed('three') >>> add(2, 1, dask_key_name='three') Delayed('three') >>> add(2, 2, dask_key_name='four') Delayed('four') ``delayed`` can also be applied to objects to make operations on them lazy: >>> a = delayed([1, 2, 3]) >>> isinstance(a, Delayed) True >>> a.compute() [1, 2, 3] The key name of a delayed object is hashed by default if ``pure=True`` or is generated randomly if ``pure=False`` (default). To explicitly set the name, you can use the ``name`` keyword. To ensure that the key is unique you should include the tokenized value as well, or otherwise ensure that it's unique: >>> from dask.base import tokenize >>> data = [1, 2, 3] >>> a = delayed(data, name='mylist-' + tokenize(data)) >>> a # doctest: +SKIP Delayed('mylist-55af65871cb378a4fa6de1660c3e8fb7') Delayed results act as a proxy to the underlying object. Many operators are supported: >>> (a + [1, 2]).compute() [1, 2, 3, 1, 2] >>> a[1].compute() 2 Method and attribute access also works: >>> a.count(2).compute() 1 Note that if a method doesn't exist, no error will be thrown until runtime: >>> res = a.not_a_real_method() # doctest: +SKIP >>> res.compute() # doctest: +SKIP AttributeError("'list' object has no attribute 'not_a_real_method'") "Magic" methods (e.g. operators and attribute access) are assumed to be pure, meaning that subsequent calls must return the same results. This behavior is not overridable through the ``delayed`` call, but can be modified using other ways as described below. To invoke an impure attribute or operator, you'd need to use it in a delayed function with ``pure=False``: >>> class Incrementer: ... def __init__(self): ... self._n = 0 ... @property ... def n(self): ... self._n += 1 ... return self._n ... >>> x = delayed(Incrementer()) >>> x.n.key == x.n.key True >>> get_n = delayed(lambda x: x.n, pure=False) >>> get_n(x).key == get_n(x).key False In contrast, methods are assumed to be impure by default, meaning that subsequent calls may return different results. To assume purity, set ``pure=True``. This allows sharing of any intermediate values. >>> a.count(2, pure=True).key == a.count(2, pure=True).key True As with function calls, method calls also respect the global ``delayed_pure`` setting and support the ``dask_key_name`` keyword: >>> a.count(2, dask_key_name="count_2") Delayed('count_2') >>> import dask >>> with dask.config.set(delayed_pure=True): ... print(a.count(2).key == a.count(2).key) True """ if isinstance(obj, Delayed): return obj if is_dask_collection(obj) or traverse: task, collections = unpack_collections(obj) else: task = quote(obj) collections = set() if not (nout is None or (type(nout) is int and nout >= 0)): raise ValueError("nout must be None or a non-negative integer, got %s" % nout) if task is obj: if not name: try: prefix = obj.__name__ except AttributeError: prefix = type(obj).__name__ token = tokenize(obj, nout, pure=pure) name = f"{prefix}-{token}" return DelayedLeaf(obj, name, pure=pure, nout=nout) else: if not name: name = f"{type(obj).__name__}-{tokenize(task, pure=pure)}" layer = {name: task} graph = HighLevelGraph.from_collections(name, layer, dependencies=collections) return Delayed(name, graph, nout)
def _swap(method, self, other): return method(other, self) def right(method): """Wrapper to create 'right' version of operator given left version""" return partial(_swap, method) def optimize(dsk, keys, **kwargs): if not isinstance(keys, (list, set)): keys = [keys] if not isinstance(dsk, HighLevelGraph): dsk = HighLevelGraph.from_collections(id(dsk), dsk, dependencies=()) dsk = dsk.cull(set(flatten(keys))) return dsk
[docs]class Delayed(DaskMethodsMixin, OperatorMethodMixin): """Represents a value to be computed by dask. Equivalent to the output from a single key in a dask graph. """ __slots__ = ("_key", "_dask", "_length", "_layer") def __init__(self, key, dsk, length=None, layer=None): self._key = key self._dask = dsk self._length = length # NOTE: Layer is used by `to_delayed` in other collections, but not in normal Delayed use self._layer = layer or key if isinstance(dsk, HighLevelGraph) and self._layer not in dsk.layers: raise ValueError( f"Layer {self._layer} not in the HighLevelGraph's layers: {list(dsk.layers)}" ) @property def key(self): return self._key @property def dask(self): return self._dask def __dask_graph__(self) -> Graph: return self.dask def __dask_keys__(self) -> NestedKeys: return [self.key] def __dask_layers__(self) -> Sequence[str]: return (self._layer,) def __dask_tokenize__(self): return self.key __dask_scheduler__ = staticmethod(DEFAULT_GET) __dask_optimize__ = globalmethod(optimize, key="delayed_optimize") def __dask_postcompute__(self): return single_key, () def __dask_postpersist__(self): return self._rebuild, () def _rebuild(self, dsk, *, rename=None): key = replace_name_in_key(self.key, rename) if rename else self.key if isinstance(dsk, HighLevelGraph) and len(dsk.layers) == 1: # FIXME Delayed is currently the only collection type that supports both high- and low-level graphs. # The HLG output of `optimize` will have a layer name that doesn't match `key`. # Remove this when Delayed is HLG-only (because `optimize` will only be passed HLGs, so it won't have # to generate random layer names). layer = next(iter(dsk.layers)) else: layer = None return Delayed(key, dsk, self._length, layer=layer) def __repr__(self): return f"Delayed({repr(self.key)})" def __hash__(self): return hash(self.key) def __dir__(self): return dir(type(self)) def __getattr__(self, attr): if attr.startswith("_"): raise AttributeError(f"Attribute {attr} not found") if attr == "visualise": # added to warn users in case of spelling error # for more details: https://github.com/dask/dask/issues/5721 warnings.warn( "dask.delayed objects have no `visualise` method. " "Perhaps you meant `visualize`?" ) return DelayedAttr(self, attr) def __setattr__(self, attr, val): try: object.__setattr__(self, attr, val) except AttributeError: # attr is neither in type(self).__slots__ nor in the __slots__ of any of its # parent classes, and all the parent classes define __slots__ too. # This last bit needs to be unit tested: if any of the parent classes omit # the __slots__ declaration, self will gain a __dict__ and this branch will # become unreachable. raise TypeError("Delayed objects are immutable") def __setitem__(self, index, val): raise TypeError("Delayed objects are immutable") def __iter__(self): if self._length is None: raise TypeError("Delayed objects of unspecified length are not iterable") for i in range(self._length): yield self[i] def __len__(self): if self._length is None: raise TypeError("Delayed objects of unspecified length have no len()") return self._length def __call__(self, *args, pure=None, dask_key_name=None, **kwargs): func = delayed(apply, pure=pure) if dask_key_name is not None: return func(self, args, kwargs, dask_key_name=dask_key_name) return func(self, args, kwargs) def __bool__(self): raise TypeError("Truth of Delayed objects is not supported") __nonzero__ = __bool__ def __get__(self, instance, cls): if instance is None: return self return types.MethodType(self, instance) @classmethod def _get_binary_operator(cls, op, inv=False): method = delayed(right(op) if inv else op, pure=True) return lambda *args, **kwargs: method(*args, **kwargs) _get_unary_operator = _get_binary_operator
def call_function(func, func_token, args, kwargs, pure=None, nout=None): dask_key_name = kwargs.pop("dask_key_name", None) pure = kwargs.pop("pure", pure) if dask_key_name is None: name = "{}-{}".format( funcname(func), tokenize(func_token, *args, pure=pure, **kwargs), ) else: name = dask_key_name args2, collections = unzip(map(unpack_collections, args), 2) collections = list(concat(collections)) if kwargs: dask_kwargs, collections2 = unpack_collections(kwargs) collections.extend(collections2) task = (apply, func, list(args2), dask_kwargs) else: task = (func,) + args2 graph = HighLevelGraph.from_collections( name, {name: task}, dependencies=collections ) nout = nout if nout is not None else None return Delayed(name, graph, length=nout) class DelayedLeaf(Delayed): __slots__ = ("_obj", "_pure", "_nout") def __init__(self, obj, key, pure=None, nout=None): super().__init__(key, None) self._obj = obj self._pure = pure self._nout = nout @property def dask(self): return HighLevelGraph.from_collections( self._key, {self._key: self._obj}, dependencies=() ) def __call__(self, *args, **kwargs): return call_function( self._obj, self._key, args, kwargs, pure=self._pure, nout=self._nout ) @property def __name__(self): return self._obj.__name__ @property def __doc__(self): return self._obj.__doc__ @property def __wrapped__(self): return self._obj class DelayedAttr(Delayed): __slots__ = ("_obj", "_attr") def __init__(self, obj, attr): key = "getattr-%s" % tokenize(obj, attr, pure=True) super().__init__(key, None) self._obj = obj self._attr = attr def __getattr__(self, attr): # Calling np.dtype(dask.delayed(...)) used to result in a segfault, as # numpy recursively tries to get `dtype` from the object. This is # likely a bug in numpy. For now, we can do a dumb for if # `x.dtype().dtype()` is called (which shouldn't ever show up in real # code). See https://github.com/dask/dask/pull/4374#issuecomment-454381465 if attr == "dtype" and self._attr == "dtype": raise AttributeError("Attribute dtype not found") return super().__getattr__(attr) @property def dask(self): layer = {self._key: (getattr, self._obj._key, self._attr)} return HighLevelGraph.from_collections( self._key, layer, dependencies=[self._obj] ) def __call__(self, *args, **kwargs): return call_function( methodcaller(self._attr), self._attr, (self._obj,) + args, kwargs ) for op in [ operator.abs, operator.neg, operator.pos, operator.invert, operator.add, operator.sub, operator.mul, operator.floordiv, operator.truediv, operator.mod, operator.pow, operator.and_, operator.or_, operator.xor, operator.lshift, operator.rshift, operator.eq, operator.ge, operator.gt, operator.ne, operator.le, operator.lt, operator.getitem, ]: Delayed._bind_operator(op) try: Delayed._bind_operator(operator.matmul) except AttributeError: pass def single_key(seq): """Pick out the only element of this list, a list of keys""" return seq[0]