brainstate 0.2.1__py2.py3-none-any.whl → 0.2.2__py2.py3-none-any.whl

brainstate/_compatible_import.py

@@ -1,340 +1,340 @@
-# Copyright 2025 BrainX Ecosystem Limited. All Rights Reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-# ==============================================================================
-
-# -*- coding: utf-8 -*-
-
-"""
-Compatibility layer for JAX version differences.
-
-This module provides a compatibility layer to handle differences between various
-versions of JAX, ensuring that BrainState works correctly across different JAX
-versions. It imports the appropriate modules and functions based on the detected
-JAX version and provides fallback implementations when necessary.
-
-Key Features:
-    - Version-aware imports for JAX core functionality
-    - Compatibility wrappers for changed APIs
-    - Fallback implementations for deprecated functions
-    - Type-safe utility functions
-
-Examples:
-    Basic usage:
-
-    >>> from brainstate._compatible_import import safe_map, safe_zip
-    >>> result = safe_map(lambda x: x * 2, [1, 2, 3])
-    >>> pairs = safe_zip([1, 2, 3], ['a', 'b', 'c'])
-
-    Using JAX core types:
-
-    >>> from brainstate._compatible_import import Primitive, ClosedJaxpr
-    >>> # These imports work across different JAX versions
-"""
-
-from contextlib import contextmanager
-from functools import partial
-from typing import Iterable, Hashable, TypeVar, Callable
-
-import jax
-from jax.core import get_aval, Tracer
-from saiunit._compatible_import import wrap_init
-
-__all__ = [
-    'ClosedJaxpr',
-    'Primitive',
-    'extend_axis_env_nd',
-    'jaxpr_as_fun',
-    'get_aval',
-    'Tracer',
-    'to_concrete_aval',
-    'safe_map',
-    'safe_zip',
-    'unzip2',
-    'wraps',
-    'Device',
-    'wrap_init',
-    'Var',
-    'JaxprEqn',
-    'Jaxpr',
-    'Literal',
-
-    'make_iota', 'to_elt', 'BatchTracer', 'BatchTrace',
-]
-
-T = TypeVar("T")
-T1 = TypeVar("T1")
-T2 = TypeVar("T2")
-T3 = TypeVar("T3")
-
-if jax.__version_info__ < (0, 5, 0):
-    from jax.lib.xla_client import Device
-else:
-    from jax import Device
-
-if jax.__version_info__ < (0, 7, 1):
-    from jax.interpreters.batching import make_iota, to_elt, BatchTracer, BatchTrace
-else:
-    from jax._src.interpreters.batching import make_iota, to_elt, BatchTracer, BatchTrace
-
-if jax.__version_info__ < (0, 4, 38):
-    from jax.core import ClosedJaxpr, extend_axis_env_nd, Primitive, jaxpr_as_fun
-    from jax.core import Primitive, Var, JaxprEqn, Jaxpr, ClosedJaxpr, Literal
-else:
-    from jax.extend.core import ClosedJaxpr, Primitive, jaxpr_as_fun
-    from jax.extend.core import Primitive, Var, JaxprEqn, Jaxpr, ClosedJaxpr, Literal
-    from jax.core import trace_ctx
-
-
-    @contextmanager
-    def extend_axis_env_nd(name_size_pairs: Iterable[tuple[Hashable, int]]):
-        """
-        Context manager to temporarily extend the JAX axis environment.
-
-        Extends the current JAX axis environment with new named axes for
-        vectorized computations, then restores the previous environment.
-
-        Args:
-            name_size_pairs: Iterable of (name, size) tuples specifying
-                           the named axes to add to the environment.
-
-        Yields:
-            None: Context with extended axis environment.
-
-        Examples:
-            >>> with extend_axis_env_nd([('batch', 32), ('seq', 128)]):
-            ...     # Code using vectorized operations with named axes
-            ...     pass
-        """
-        prev = trace_ctx.axis_env
-        try:
-            trace_ctx.set_axis_env(prev.extend_pure(name_size_pairs))
-            yield
-        finally:
-            trace_ctx.set_axis_env(prev)
-
-if jax.__version_info__ < (0, 6, 0):
-    from jax.util import safe_map, safe_zip, unzip2, wraps
-
-else:
-    def safe_map(f, *args):
-        """
-        Map a function over multiple sequences with length checking.
-
-        Applies a function to corresponding elements from multiple sequences,
-        ensuring all sequences have the same length.
-
-        Args:
-            f: Function to apply to elements from each sequence.
-            *args: Variable number of sequences to map over.
-
-        Returns:
-            list: Results of applying f to corresponding elements.
-
-        Raises:
-            AssertionError: If input sequences have different lengths.
-
-        Examples:
-            >>> safe_map(lambda x, y: x + y, [1, 2, 3], [4, 5, 6])
-            [5, 7, 9]
-
-            >>> safe_map(str.upper, ['a', 'b', 'c'])
-            ['A', 'B', 'C']
-        """
-        args = list(map(list, args))
-        n = len(args[0])
-        for arg in args[1:]:
-            assert len(arg) == n, f'length mismatch: {list(map(len, args))}'
-        return list(map(f, *args))
-
-
-    def safe_zip(*args):
-        """
-        Zip multiple sequences with length checking.
-
-        Combines corresponding elements from multiple sequences into tuples,
-        ensuring all sequences have the same length.
-
-        Args:
-            *args: Variable number of sequences to zip together.
-
-        Returns:
-            list: List of tuples containing corresponding elements.
-
-        Raises:
-            AssertionError: If input sequences have different lengths.
-
-        Examples:
-            >>> safe_zip([1, 2, 3], ['a', 'b', 'c'])
-            [(1, 'a'), (2, 'b'), (3, 'c')]
-
-            >>> safe_zip([1, 2], [3, 4], [5, 6])
-            [(1, 3, 5), (2, 4, 6)]
-        """
-        args = list(map(list, args))
-        n = len(args[0])
-        for arg in args[1:]:
-            assert len(arg) == n, f'length mismatch: {list(map(len, args))}'
-        return list(zip(*args))
-
-
-    def unzip2(xys: Iterable[tuple[T1, T2]]) -> tuple[tuple[T1, ...], tuple[T2, ...]]:
-        """
-        Unzip sequence of length-2 tuples into two tuples.
-
-        Takes an iterable of 2-tuples and separates them into two tuples
-        containing the first and second elements respectively.
-
-        Args:
-            xys: Iterable of 2-tuples to unzip.
-
-        Returns:
-            tuple: A 2-tuple containing:
-                - Tuple of all first elements
-                - Tuple of all second elements
-
-        Examples:
-            >>> pairs = [(1, 'a'), (2, 'b'), (3, 'c')]
-            >>> nums, letters = unzip2(pairs)
-            >>> nums
-            (1, 2, 3)
-            >>> letters
-            ('a', 'b', 'c')
-
-        Notes:
-            We deliberately don't use zip(*xys) because it is lazily evaluated,
-            is too permissive about inputs, and does not guarantee a length-2 output.
-        """
-        # Note: we deliberately don't use zip(*xys) because it is lazily evaluated,
-        # is too permissive about inputs, and does not guarantee a length-2 output.
-        xs: list[T1] = []
-        ys: list[T2] = []
-        for x, y in xys:
-            xs.append(x)
-            ys.append(y)
-        return tuple(xs), tuple(ys)
-
-
-    def fun_name(fun: Callable):
-        """
-        Extract the name of a function, handling special cases.
-
-        Attempts to get the name of a function, with special handling for
-        partial functions and fallback for unnamed functions.
-
-        Args:
-            fun: The function to get the name from.
-
-        Returns:
-            str: The function name, or "<unnamed function>" if no name available.
-
-        Examples:
-            >>> def my_function():
-            ...     pass
-            >>> fun_name(my_function)
-            'my_function'
-
-            >>> from functools import partial
-            >>> add = lambda x, y: x + y
-            >>> add_one = partial(add, 1)
-            >>> fun_name(add_one)
-            '<lambda>'
-        """
-        name = getattr(fun, "__name__", None)
-        if name is not None:
-            return name
-        if isinstance(fun, partial):
-            return fun_name(fun.func)
-        else:
-            return "<unnamed function>"
-
-
-    def wraps(
-        wrapped: Callable,
-        namestr: str | None = None,
-        docstr: str | None = None,
-        **kwargs,
-    ) -> Callable[[T], T]:
-        """
-        Enhanced function wrapper with fine-grained control.
-
-        Like functools.wraps, but provides more control over the name and docstring
-        of the resulting function. Useful for creating custom decorators.
-
-        Args:
-            wrapped: The function being wrapped.
-            namestr: Optional format string for the wrapper function name.
-                    Can use {fun} placeholder for the original function name.
-            docstr: Optional format string for the wrapper function docstring.
-                   Can use {fun}, {doc}, and other kwargs as placeholders.
-            **kwargs: Additional keyword arguments for format string substitution.
-
-        Returns:
-            Callable: A decorator function that applies the wrapping.
-
-        Examples:
-            >>> def my_decorator(func):
-            ...     @wraps(func, namestr="decorated_{fun}")
-            ...     def wrapper(*args, **kwargs):
-            ...         return func(*args, **kwargs)
-            ...     return wrapper
-
-            >>> @my_decorator
-            ... def example():
-            ...     pass
-            >>> example.__name__
-            'decorated_example'
-        """
-
-        def wrapper(fun: T) -> T:
-            try:
-                name = fun_name(wrapped)
-                doc = getattr(wrapped, "__doc__", "") or ""
-                fun.__dict__.update(getattr(wrapped, "__dict__", {}))
-                fun.__annotations__ = getattr(wrapped, "__annotations__", {})
-                fun.__name__ = name if namestr is None else namestr.format(fun=name)
-                fun.__module__ = getattr(wrapped, "__module__", "<unknown module>")
-                fun.__doc__ = (doc if docstr is None
-                               else docstr.format(fun=name, doc=doc, **kwargs))
-                fun.__qualname__ = getattr(wrapped, "__qualname__", fun.__name__)
-                fun.__wrapped__ = wrapped
-            except Exception:
-                pass
-            return fun
-
-        return wrapper
-
-
-def to_concrete_aval(aval):
-    """
-    Convert an abstract value to its concrete representation.
-
-    Takes an abstract value and attempts to convert it to a concrete value,
-    handling JAX Tracer objects appropriately.
-
-    Args:
-        aval: The abstract value to convert.
-
-    Returns:
-        The concrete value representation, or the original aval if already concrete.
-
-    Examples:
-        >>> import jax.numpy as jnp
-        >>> arr = jnp.array([1, 2, 3])
-        >>> concrete = to_concrete_aval(arr)
-        # Returns the concrete array value
-    """
-    aval = get_aval(aval)
-    if isinstance(aval, Tracer):
-        return aval.to_concrete_value()
-    return aval
+# Copyright 2025 BrainX Ecosystem Limited. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ==============================================================================
+
+# -*- coding: utf-8 -*-
+
+"""
+Compatibility layer for JAX version differences.
+
+This module provides a compatibility layer to handle differences between various
+versions of JAX, ensuring that BrainState works correctly across different JAX
+versions. It imports the appropriate modules and functions based on the detected
+JAX version and provides fallback implementations when necessary.
+
+Key Features:
+    - Version-aware imports for JAX core functionality
+    - Compatibility wrappers for changed APIs
+    - Fallback implementations for deprecated functions
+    - Type-safe utility functions
+
+Examples:
+    Basic usage:
+
+    >>> from brainstate._compatible_import import safe_map, safe_zip
+    >>> result = safe_map(lambda x: x * 2, [1, 2, 3])
+    >>> pairs = safe_zip([1, 2, 3], ['a', 'b', 'c'])
+
+    Using JAX core types:
+
+    >>> from brainstate._compatible_import import Primitive, ClosedJaxpr
+    >>> # These imports work across different JAX versions
+"""
+
+from contextlib import contextmanager
+from functools import partial
+from typing import Iterable, Hashable, TypeVar, Callable
+
+import jax
+from jax.core import get_aval, Tracer
+from saiunit._compatible_import import wrap_init
+
+__all__ = [
+    'ClosedJaxpr',
+    'Primitive',
+    'extend_axis_env_nd',
+    'jaxpr_as_fun',
+    'get_aval',
+    'Tracer',
+    'to_concrete_aval',
+    'safe_map',
+    'safe_zip',
+    'unzip2',
+    'wraps',
+    'Device',
+    'wrap_init',
+    'Var',
+    'JaxprEqn',
+    'Jaxpr',
+    'Literal',
+
+    'make_iota', 'to_elt', 'BatchTracer', 'BatchTrace',
+]
+
+T = TypeVar("T")
+T1 = TypeVar("T1")
+T2 = TypeVar("T2")
+T3 = TypeVar("T3")
+
+if jax.__version_info__ < (0, 5, 0):
+    from jax.lib.xla_client import Device
+else:
+    from jax import Device
+
+if jax.__version_info__ < (0, 7, 1):
+    from jax.interpreters.batching import make_iota, to_elt, BatchTracer, BatchTrace
+else:
+    from jax._src.interpreters.batching import make_iota, to_elt, BatchTracer, BatchTrace
+
+if jax.__version_info__ < (0, 4, 38):
+    from jax.core import ClosedJaxpr, extend_axis_env_nd, Primitive, jaxpr_as_fun
+    from jax.core import Primitive, Var, JaxprEqn, Jaxpr, ClosedJaxpr, Literal
+else:
+    from jax.extend.core import ClosedJaxpr, Primitive, jaxpr_as_fun
+    from jax.extend.core import Primitive, Var, JaxprEqn, Jaxpr, ClosedJaxpr, Literal
+    from jax.core import trace_ctx
+
+
+    @contextmanager
+    def extend_axis_env_nd(name_size_pairs: Iterable[tuple[Hashable, int]]):
+        """
+        Context manager to temporarily extend the JAX axis environment.
+
+        Extends the current JAX axis environment with new named axes for
+        vectorized computations, then restores the previous environment.
+
+        Args:
+            name_size_pairs: Iterable of (name, size) tuples specifying
+                           the named axes to add to the environment.
+
+        Yields:
+            None: Context with extended axis environment.
+
+        Examples:
+            >>> with extend_axis_env_nd([('batch', 32), ('seq', 128)]):
+            ...     # Code using vectorized operations with named axes
+            ...     pass
+        """
+        prev = trace_ctx.axis_env
+        try:
+            trace_ctx.set_axis_env(prev.extend_pure(name_size_pairs))
+            yield
+        finally:
+            trace_ctx.set_axis_env(prev)
+
+if jax.__version_info__ < (0, 6, 0):
+    from jax.util import safe_map, safe_zip, unzip2, wraps
+
+else:
+    def safe_map(f, *args):
+        """
+        Map a function over multiple sequences with length checking.
+
+        Applies a function to corresponding elements from multiple sequences,
+        ensuring all sequences have the same length.
+
+        Args:
+            f: Function to apply to elements from each sequence.
+            *args: Variable number of sequences to map over.
+
+        Returns:
+            list: Results of applying f to corresponding elements.
+
+        Raises:
+            AssertionError: If input sequences have different lengths.
+
+        Examples:
+            >>> safe_map(lambda x, y: x + y, [1, 2, 3], [4, 5, 6])
+            [5, 7, 9]
+
+            >>> safe_map(str.upper, ['a', 'b', 'c'])
+            ['A', 'B', 'C']
+        """
+        args = list(map(list, args))
+        n = len(args[0])
+        for arg in args[1:]:
+            assert len(arg) == n, f'length mismatch: {list(map(len, args))}'
+        return list(map(f, *args))
+
+
+    def safe_zip(*args):
+        """
+        Zip multiple sequences with length checking.
+
+        Combines corresponding elements from multiple sequences into tuples,
+        ensuring all sequences have the same length.
+
+        Args:
+            *args: Variable number of sequences to zip together.
+
+        Returns:
+            list: List of tuples containing corresponding elements.
+
+        Raises:
+            AssertionError: If input sequences have different lengths.
+
+        Examples:
+            >>> safe_zip([1, 2, 3], ['a', 'b', 'c'])
+            [(1, 'a'), (2, 'b'), (3, 'c')]
+
+            >>> safe_zip([1, 2], [3, 4], [5, 6])
+            [(1, 3, 5), (2, 4, 6)]
+        """
+        args = list(map(list, args))
+        n = len(args[0])
+        for arg in args[1:]:
+            assert len(arg) == n, f'length mismatch: {list(map(len, args))}'
+        return list(zip(*args))
+
+
+    def unzip2(xys: Iterable[tuple[T1, T2]]) -> tuple[tuple[T1, ...], tuple[T2, ...]]:
+        """
+        Unzip sequence of length-2 tuples into two tuples.
+
+        Takes an iterable of 2-tuples and separates them into two tuples
+        containing the first and second elements respectively.
+
+        Args:
+            xys: Iterable of 2-tuples to unzip.
+
+        Returns:
+            tuple: A 2-tuple containing:
+                - Tuple of all first elements
+                - Tuple of all second elements
+
+        Examples:
+            >>> pairs = [(1, 'a'), (2, 'b'), (3, 'c')]
+            >>> nums, letters = unzip2(pairs)
+            >>> nums
+            (1, 2, 3)
+            >>> letters
+            ('a', 'b', 'c')
+
+        Notes:
+            We deliberately don't use zip(*xys) because it is lazily evaluated,
+            is too permissive about inputs, and does not guarantee a length-2 output.
+        """
+        # Note: we deliberately don't use zip(*xys) because it is lazily evaluated,
+        # is too permissive about inputs, and does not guarantee a length-2 output.
+        xs: list[T1] = []
+        ys: list[T2] = []
+        for x, y in xys:
+            xs.append(x)
+            ys.append(y)
+        return tuple(xs), tuple(ys)
+
+
+    def fun_name(fun: Callable):
+        """
+        Extract the name of a function, handling special cases.
+
+        Attempts to get the name of a function, with special handling for
+        partial functions and fallback for unnamed functions.
+
+        Args:
+            fun: The function to get the name from.
+
+        Returns:
+            str: The function name, or "<unnamed function>" if no name available.
+
+        Examples:
+            >>> def my_function():
+            ...     pass
+            >>> fun_name(my_function)
+            'my_function'
+
+            >>> from functools import partial
+            >>> add = lambda x, y: x + y
+            >>> add_one = partial(add, 1)
+            >>> fun_name(add_one)
+            '<lambda>'
+        """
+        name = getattr(fun, "__name__", None)
+        if name is not None:
+            return name
+        if isinstance(fun, partial):
+            return fun_name(fun.func)
+        else:
+            return "<unnamed function>"
+
+
+    def wraps(
+        wrapped: Callable,
+        namestr: str | None = None,
+        docstr: str | None = None,
+        **kwargs,
+    ) -> Callable[[T], T]:
+        """
+        Enhanced function wrapper with fine-grained control.
+
+        Like functools.wraps, but provides more control over the name and docstring
+        of the resulting function. Useful for creating custom decorators.
+
+        Args:
+            wrapped: The function being wrapped.
+            namestr: Optional format string for the wrapper function name.
+                    Can use {fun} placeholder for the original function name.
+            docstr: Optional format string for the wrapper function docstring.
+                   Can use {fun}, {doc}, and other kwargs as placeholders.
+            **kwargs: Additional keyword arguments for format string substitution.
+
+        Returns:
+            Callable: A decorator function that applies the wrapping.
+
+        Examples:
+            >>> def my_decorator(func):
+            ...     @wraps(func, namestr="decorated_{fun}")
+            ...     def wrapper(*args, **kwargs):
+            ...         return func(*args, **kwargs)
+            ...     return wrapper
+
+            >>> @my_decorator
+            ... def example():
+            ...     pass
+            >>> example.__name__
+            'decorated_example'
+        """
+
+        def wrapper(fun: T) -> T:
+            try:
+                name = fun_name(wrapped)
+                doc = getattr(wrapped, "__doc__", "") or ""
+                fun.__dict__.update(getattr(wrapped, "__dict__", {}))
+                fun.__annotations__ = getattr(wrapped, "__annotations__", {})
+                fun.__name__ = name if namestr is None else namestr.format(fun=name)
+                fun.__module__ = getattr(wrapped, "__module__", "<unknown module>")
+                fun.__doc__ = (doc if docstr is None
+                               else docstr.format(fun=name, doc=doc, **kwargs))
+                fun.__qualname__ = getattr(wrapped, "__qualname__", fun.__name__)
+                fun.__wrapped__ = wrapped
+            except Exception:
+                pass
+            return fun
+
+        return wrapper
+
+
+def to_concrete_aval(aval):
+    """
+    Convert an abstract value to its concrete representation.
+
+    Takes an abstract value and attempts to convert it to a concrete value,
+    handling JAX Tracer objects appropriately.
+
+    Args:
+        aval: The abstract value to convert.
+
+    Returns:
+        The concrete value representation, or the original aval if already concrete.
+
+    Examples:
+        >>> import jax.numpy as jnp
+        >>> arr = jnp.array([1, 2, 3])
+        >>> concrete = to_concrete_aval(arr)
+        # Returns the concrete array value
+    """
+    aval = get_aval(aval)
+    if isinstance(aval, Tracer):
+        return aval.to_concrete_value()
+    return aval