"""Compressed Block Sparse Row matrix format"""
from __future__ import division, print_function, absolute_import


__docformat__ = "restructuredtext en"

__all__ = ['bsr_matrix', 'isspmatrix_bsr']

from warnings import warn

import numpy as np

from .data import _data_matrix, _minmax_mixin
from .compressed import _cs_matrix
from .base import isspmatrix, _formats, spmatrix
from .sputils import (isshape, getdtype, to_native, upcast, get_index_dtype,
                      check_shape)
from . import _sparsetools
from ._sparsetools import (bsr_matvec, bsr_matvecs, csr_matmat_pass1,
                           bsr_matmat_pass2, bsr_transpose, bsr_sort_indices,
                           bsr_tocsr)


class bsr_matrix(_cs_matrix, _minmax_mixin):
    """Block Sparse Row matrix

    This can be instantiated in several ways:
        bsr_matrix(D, [blocksize=(R,C)])
            where D is a dense matrix or 2-D ndarray.

        bsr_matrix(S, [blocksize=(R,C)])
            with another sparse matrix S (equivalent to S.tobsr())

        bsr_matrix((M, N), [blocksize=(R,C), dtype])
            to construct an empty matrix with shape (M, N)
            dtype is optional, defaulting to dtype='d'.

        bsr_matrix((data, ij), [blocksize=(R,C), shape=(M, N)])
            where ``data`` and ``ij`` satisfy ``a[ij[0, k], ij[1, k]] = data[k]``

        bsr_matrix((data, indices, indptr), [shape=(M, N)])
            is the standard BSR representation where the block column
            indices for row i are stored in ``indices[indptr[i]:indptr[i+1]]``
            and their corresponding block values are stored in
            ``data[ indptr[i]: indptr[i+1] ]``.  If the shape parameter is not
            supplied, the matrix dimensions are inferred from the index arrays.

    Attributes
    ----------
    dtype : dtype
        Data type of the matrix
    shape : 2-tuple
        Shape of the matrix
    ndim : int
        Number of dimensions (this is always 2)
    nnz
        Number of nonzero elements
    data
        Data array of the matrix
    indices
        BSR format index array
    indptr
        BSR format index pointer array
    blocksize
        Block size of the matrix
    has_sorted_indices
        Whether indices are sorted

    Notes
    -----
    Sparse matrices can be used in arithmetic operations: they support
    addition, subtraction, multiplication, division, and matrix power.

    **Summary of BSR format**

    The Block Compressed Row (BSR) format is very similar to the Compressed
    Sparse Row (CSR) format.  BSR is appropriate for sparse matrices with dense
    sub matrices like the last example below.  Block matrices often arise in
    vector-valued finite element discretizations.  In such cases, BSR is
    considerably more efficient than CSR and CSC for many sparse arithmetic
    operations.

    **Blocksize**

    The blocksize (R,C) must evenly divide the shape of the matrix (M,N).
    That is, R and C must satisfy the relationship ``M % R = 0`` and
    ``N % C = 0``.

    If no blocksize is specified, a simple heuristic is applied to determine
    an appropriate blocksize.

    Examples
    --------
    >>> from scipy.sparse import bsr_matrix
    >>> bsr_matrix((3, 4), dtype=np.int8).toarray()
    array([[0, 0, 0, 0],
           [0, 0, 0, 0],
           [0, 0, 0, 0]], dtype=int8)

    >>> row = np.array([0, 0, 1, 2, 2, 2])
    >>> col = np.array([0, 2, 2, 0, 1, 2])
    >>> data = np.array([1, 2, 3 ,4, 5, 6])
    >>> bsr_matrix((data, (row, col)), shape=(3, 3)).toarray()
    array([[1, 0, 2],
           [0, 0, 3],
           [4, 5, 6]])

    >>> indptr = np.array([0, 2, 3, 6])
    >>> indices = np.array([0, 2, 2, 0, 1, 2])
    >>> data = np.array([1, 2, 3, 4, 5, 6]).repeat(4).reshape(6, 2, 2)
    >>> bsr_matrix((data,indices,indptr), shape=(6, 6)).toarray()
    array([[1, 1, 0, 0, 2, 2],
           [1, 1, 0, 0, 2, 2],
           [0, 0, 0, 0, 3, 3],
           [0, 0, 0, 0, 3, 3],
           [4, 4, 5, 5, 6, 6],
           [4, 4, 5, 5, 6, 6]])

    """
    format = 'bsr'

    def __init__(self, arg1, shape=None, dtype=None, copy=False, blocksize=None):
        _data_matrix.__init__(self)

        if isspmatrix(arg1):
            if isspmatrix_bsr(arg1) and copy:
                arg1 = arg1.copy()
            else:
                arg1 = arg1.tobsr(blocksize=blocksize)
            self._set_self(arg1)

        elif isinstance(arg1,tuple):
            if isshape(arg1):
                # it's a tuple of matrix dimensions (M,N)
                self._shape = check_shape(arg1)
                M,N = self.shape
                # process blocksize
                if blocksize is None:
                    blocksize = (1,1)
                else:
                    if not isshape(blocksize):
                        raise ValueError('invalid blocksize=%s' % blocksize)
                    blocksize = tuple(blocksize)
                self.data = np.zeros((0,) + blocksize, getdtype(dtype, default=float))

                R,C = blocksize
                if (M % R) != 0 or (N % C) != 0:
                    raise ValueError('shape must be multiple of blocksize')

                # Select index dtype large enough to pass array and
                # scalar parameters to sparsetools
                idx_dtype = get_index_dtype(maxval=max(M//R, N//C, R, C))
                self.indices = np.zeros(0, dtype=idx_dtype)
                self.indptr = np.zeros(M//R + 1, dtype=idx_dtype)

            elif len(arg1) == 2:
                # (data,(row,col)) format
                from .coo import coo_matrix
                self._set_self(coo_matrix(arg1, dtype=dtype).tobsr(blocksize=blocksize))

            elif len(arg1) == 3:
                # (data,indices,indptr) format
                (data, indices, indptr) = arg1

                # Select index dtype large enough to pass array and
                # scalar parameters to sparsetools
                maxval = 1
                if shape is not None:
                    maxval = max(shape)
                if blocksize is not None:
                    maxval = max(maxval, max(blocksize))
                idx_dtype = get_index_dtype((indices, indptr), maxval=maxval, check_contents=True)

                self.indices = np.array(indices, copy=copy, dtype=idx_dtype)
                self.indptr = np.array(indptr, copy=copy, dtype=idx_dtype)
                self.data = np.array(data, copy=copy, dtype=getdtype(dtype, data))
            else:
                raise ValueError('unrecognized bsr_matrix constructor usage')
        else:
            # must be dense
            try:
                arg1 = np.asarray(arg1)
            except Exception:
                raise ValueError("unrecognized form for"
                        " %s_matrix constructor" % self.format)
            from .coo import coo_matrix
            arg1 = coo_matrix(arg1, dtype=dtype).tobsr(blocksize=blocksize)
            self._set_self(arg1)

        if shape is not None:
            self._shape = check_shape(shape)
        else:
            if self.shape is None:
                # shape not already set, try to infer dimensions
                try:
                    M = len(self.indptr) - 1
                    N = self.indices.max() + 1
                except Exception:
                    raise ValueError('unable to infer matrix dimensions')
                else:
                    R,C = self.blocksize
                    self._shape = check_shape((M*R,N*C))

        if self.shape is None:
            if shape is None:
                # TODO infer shape here
                raise ValueError('need to infer shape')
            else:
                self._shape = check_shape(shape)

        if dtype is not None:
            self.data = self.data.astype(dtype)

        self.check_format(full_check=False)

    def check_format(self, full_check=True):
        """check whether the matrix format is valid

            *Parameters*:
                full_check:
                    True  - rigorous check, O(N) operations : default
                    False - basic check, O(1) operations

        """
        M,N = self.shape
        R,C = self.blocksize

        # index arrays should have integer data types
        if self.indptr.dtype.kind != 'i':
            warn("indptr array has non-integer dtype (%s)"
                    % self.indptr.dtype.name)
        if self.indices.dtype.kind != 'i':
            warn("indices array has non-integer dtype (%s)"
                    % self.indices.dtype.name)

        idx_dtype = get_index_dtype((self.indices, self.indptr))
        self.indptr = np.asarray(self.indptr, dtype=idx_dtype)
        self.indices = np.asarray(self.indices, dtype=idx_dtype)
        self.data = to_native(self.data)

        # check array shapes
        if self.indices.ndim != 1 or self.indptr.ndim != 1:
            raise ValueError("indices, and indptr should be 1-D")
        if self.data.ndim != 3:
            raise ValueError("data should be 3-D")

        # check index pointer
        if (len(self.indptr) != M//R + 1):
            raise ValueError("index pointer size (%d) should be (%d)" %
                                (len(self.indptr), M//R + 1))
        if (self.indptr[0] != 0):
            raise ValueError("index pointer should start with 0")

        # check index and data arrays
        if (len(self.indices) != len(self.data)):
            raise ValueError("indices and data should have the same size")
        if (self.indptr[-1] > len(self.indices)):
            raise ValueError("Last value of index pointer should be less than "
                                "the size of index and data arrays")

        self.prune()

        if full_check:
            # check format validity (more expensive)
            if self.nnz > 0:
                if self.indices.max() >= N//C:
                    raise ValueError("column index values must be < %d (now max %d)" % (N//C, self.indices.max()))
                if self.indices.min() < 0:
                    raise ValueError("column index values must be >= 0")
                if np.diff(self.indptr).min() < 0:
                    raise ValueError("index pointer values must form a "
                                        "non-decreasing sequence")

        # if not self.has_sorted_indices():
        #    warn('Indices were not in sorted order. Sorting indices.')
        #    self.sort_indices(check_first=False)

    def _get_blocksize(self):
        return self.data.shape[1:]
    blocksize = property(fget=_get_blocksize)

    def getnnz(self, axis=None):
        if axis is not None:
            raise NotImplementedError("getnnz over an axis is not implemented "
                                      "for BSR format")
        R,C = self.blocksize
        return int(self.indptr[-1] * R * C)

    getnnz.__doc__ = spmatrix.getnnz.__doc__

    def __repr__(self):
        format = _formats[self.getformat()][1]
        return ("<%dx%d sparse matrix of type '%s'\n"
                "\twith %d stored elements (blocksize = %dx%d) in %s format>" %
                (self.shape + (self.dtype.type, self.nnz) + self.blocksize +
                 (format,)))

    def diagonal(self, k=0):
        rows, cols = self.shape
        if k <= -rows or k >= cols:
            raise ValueError("k exceeds matrix dimensions")
        R, C = self.blocksize
        y = np.zeros(min(rows + min(k, 0), cols - max(k, 0)),
                     dtype=upcast(self.dtype))
        _sparsetools.bsr_diagonal(k, rows // R, cols // C, R, C,
                                  self.indptr, self.indices,
                                  np.ravel(self.data), y)
        return y

    diagonal.__doc__ = spmatrix.diagonal.__doc__

    ##########################
    # NotImplemented methods #
    ##########################

    def __getitem__(self,key):
        raise NotImplementedError

    def __setitem__(self,key,val):
        raise NotImplementedError

    ######################
    # Arithmetic methods #
    ######################

    @np.deprecate(message="BSR matvec is deprecated in scipy 0.19.0. "
                          "Use * operator instead.")
    def matvec(self, other):
        """Multiply matrix by vector."""
        return self * other

    @np.deprecate(message="BSR matmat is deprecated in scipy 0.19.0. "
                          "Use * operator instead.")
    def matmat(self, other):
        """Multiply this sparse matrix by other matrix."""
        return self * other

    def _add_dense(self, other):
        return self.tocoo(copy=False)._add_dense(other)

    def _mul_vector(self, other):
        M,N = self.shape
        R,C = self.blocksize

        result = np.zeros(self.shape[0], dtype=upcast(self.dtype, other.dtype))