passagemath-flint 10.6.1rc10__cp312-cp312-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl

sage/matrix/matrix_rational_dense.pyx

@@ -0,0 +1,2995 @@
+# sage_setup: distribution = sagemath-flint
+"""
+Dense matrices over the rational field
+
+EXAMPLES:
+
+We create a 3x3 matrix with rational entries and do some
+operations with it.
+
+::
+
+    sage: a = matrix(QQ, 3,3, [1,2/3, -4/5, 1,1,1, 8,2, -3/19]); a
+    [    1   2/3  -4/5]
+    [    1     1     1]
+    [    8     2 -3/19]
+    sage: a.det()
+    2303/285
+    sage: a.charpoly()
+    x^3 - 35/19*x^2 + 1259/285*x - 2303/285
+    sage: b = a^(-1); b
+    [ -615/2303  -426/2303   418/2303]
+    [ 2325/2303  1779/2303  -513/2303]
+    [-1710/2303   950/2303    95/2303]
+    sage: b.det()
+    285/2303
+    sage: a == b
+    False
+    sage: a < b
+    False
+    sage: b < a
+    True
+    sage: a > b
+    True
+    sage: a*b
+    [1 0 0]
+    [0 1 0]
+    [0 0 1]
+
+TESTS::
+
+    sage: a = matrix(QQ, 2, range(4), sparse=False)
+    sage: TestSuite(a).run()
+
+Test hashing::
+
+    sage: m = matrix(QQ, 2, [1/2, -1, 2, 3])
+    sage: hash(m)
+    Traceback (most recent call last):
+    ...
+    TypeError: mutable matrices are unhashable
+    sage: m.set_immutable()
+    sage: hash(m)
+    2212268000387745777  # 64-bit
+    1997752305           # 32-bit
+"""
+
+# ****************************************************************************
+#       Copyright (C) 2004,2005,2006 William Stein <wstein@gmail.com>
+#                     2017 Vincent Delecroix <20100.delecroix@gmail.com>
+#
+# This program is free software: you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation, either version 2 of the License, or
+# (at your option) any later version.
+#                  https://www.gnu.org/licenses/
+# ****************************************************************************
+
+from libc.string cimport strcpy, strlen
+
+from sage.categories.rings import Rings
+from sage.cpython.string cimport char_to_str
+
+from sage.modules.vector_rational_dense cimport Vector_rational_dense
+from sage.ext.stdsage cimport PY_NEW
+from sage.misc.randstate cimport randstate, current_randstate
+
+from cysignals.signals cimport sig_on, sig_off
+from cysignals.memory cimport sig_malloc, sig_free
+
+from sage.libs.gmp.types cimport mpz_t, mpq_t
+from sage.libs.gmp.mpz cimport mpz_init, mpz_clear, mpz_cmp_si
+from sage.libs.gmp.mpq cimport mpq_init, mpq_clear, mpq_set_si, mpq_mul, mpq_add, mpq_set
+from sage.libs.gmp.randomize cimport (mpq_randomize_entry,
+                                      mpq_randomize_entry_as_int,
+                                      mpq_randomize_entry_recip_uniform,
+                                      mpq_randomize_entry_nonzero,
+                                      mpq_randomize_entry_as_int_nonzero,
+                                      mpq_randomize_entry_recip_uniform_nonzero)
+
+from sage.libs.flint.fmpz cimport *
+from sage.libs.flint.fmpq cimport *
+from sage.libs.flint.fmpz_mat cimport *
+from sage.libs.flint.fmpq_mat cimport *
+
+cimport sage.structure.element
+
+from sage.structure.richcmp cimport rich_to_bool
+from sage.rings.rational cimport Rational
+from sage.matrix.matrix cimport Matrix
+from sage.matrix.args cimport SparseEntry, MatrixArgs_init
+from sage.matrix.matrix_integer_dense cimport Matrix_integer_dense
+from sage.structure.element cimport Element, Vector
+from sage.rings.integer cimport Integer
+from sage.rings.integer_ring import ZZ, IntegerRing_class
+import sage.rings.abc
+from sage.rings.rational_field import QQ
+
+from sage.matrix.matrix2 import decomp_seq
+from sage.misc.verbose import verbose
+
+
+cdef class Matrix_rational_dense(Matrix_dense):
+    def __cinit__(self):
+        """
+        Create and allocate memory for the matrix.
+
+        EXAMPLES::
+
+            sage: from sage.matrix.matrix_rational_dense import Matrix_rational_dense
+            sage: a = Matrix_rational_dense.__new__(Matrix_rational_dense, Mat(ZZ,3), 0,0,0)
+            sage: type(a)
+            <class 'sage.matrix.matrix_rational_dense.Matrix_rational_dense'>
+
+        .. WARNING::
+
+           This is for internal use only, or if you really know what
+           you're doing.
+        """
+        sig_on()
+        fmpq_mat_init(self._matrix, self._nrows, self._ncols)
+        sig_off()
+
+    cdef inline Matrix_rational_dense _new_matrix(self, Py_ssize_t nrows, Py_ssize_t ncols):
+        if nrows == self._nrows and ncols == self._ncols:
+            parent = self._parent
+        else:
+            parent = self.matrix_space(nrows, ncols)
+
+        return Matrix_rational_dense.__new__(Matrix_rational_dense, parent, None, None, None)
+
+    def __dealloc__(self):
+        fmpq_mat_clear(self._matrix)
+
+    def __init__(self, parent, entries=None, copy=None, bint coerce=True):
+        r"""
+        INPUT:
+
+        - ``parent`` -- a matrix space over ``QQ``
+
+        - ``entries`` -- see :func:`matrix`
+
+        - ``copy`` -- ignored (for backwards compatibility)
+
+        - ``coerce`` -- if ``False``, assume without checking that the
+          entries are of type :class:`Rational`
+
+        TESTS::
+
+            sage: matrix(QQ, 2, 2, 1/4)
+            [1/4   0]
+            [  0 1/4]
+            sage: matrix(QQ, 3, 1, [1/2, -3/4, 0])
+            [ 1/2]
+            [-3/4]
+            [   0]
+            sage: matrix(QQ, 2, 2, 0.5)
+            [1/2   0]
+            [  0 1/2]
+        """
+        ma = MatrixArgs_init(parent, entries)
+        cdef Rational z
+        for t in ma.iter(coerce, True):
+            se = <SparseEntry>t
+            z = <Rational>se.entry
+            fmpq_set_mpq(fmpq_mat_entry(self._matrix, se.i, se.j), z.value)
+
+    def matrix_from_columns(self, columns):
+        """
+        Return the matrix constructed from ``self`` using columns with indices
+        in the columns list.
+
+        EXAMPLES::
+
+            sage: A = matrix(QQ, 3, range(9))
+            sage: A
+            [0 1 2]
+            [3 4 5]
+            [6 7 8]
+            sage: A.matrix_from_columns([2,1])
+            [2 1]
+            [5 4]
+            [8 7]
+            sage: A.matrix_from_columns((2,1,0,2))
+            [2 1 0 2]
+            [5 4 3 5]
+            [8 7 6 8]
+        """
+        cdef Matrix_rational_dense A
+        cdef Py_ssize_t k, r, col
+
+        A = self._new_matrix(self._nrows, len(columns))
+        k = 0
+        for col in columns:
+            if col < 0 or col >= self._ncols:
+                raise IndexError("column out of range")
+            for r in range(self._nrows):
+                fmpq_set(fmpq_mat_entry(A._matrix, r, k), fmpq_mat_entry(self._matrix, r, col))
+            k = k + 1
+        return A
+
+    def add_to_entry(self, Py_ssize_t i, Py_ssize_t j, elt):
+        r"""
+        Add ``elt`` to the entry at position ``(i,j)``.
+
+        EXAMPLES::
+
+            sage: m = matrix(QQ, 2, 2)
+            sage: m.add_to_entry(0, 0, -1/3)
+            sage: m
+            [-1/3    0]
+            [   0    0]
+        """
+        if not isinstance(elt, Rational):
+            elt = Rational(elt)
+        if i < 0:
+            i += self._nrows
+        if i < 0 or i >= self._nrows:
+            raise IndexError("row index out of range")
+        if j < 0:
+            j += self._ncols
+        if j < 0 or j >= self._ncols:
+            raise IndexError("column index out of range")
+        cdef fmpq_t tmp
+        fmpq_init(tmp)
+        fmpq_set_mpq(tmp, (<Rational>elt).value)
+        fmpq_add(fmpq_mat_entry(self._matrix, i, j),
+                 fmpq_mat_entry(self._matrix, i, j),
+                 tmp)
+        fmpq_clear(tmp)
+
+    cdef set_unsafe(self, Py_ssize_t i, Py_ssize_t j, value):
+        fmpq_set_mpq(fmpq_mat_entry(self._matrix, i, j), (<Rational> value).value)
+
+    cdef get_unsafe(self, Py_ssize_t i, Py_ssize_t j):
+        cdef Rational x
+        x = Rational.__new__(Rational)
+        fmpq_get_mpq(x.value, fmpq_mat_entry(self._matrix, i, j))
+        return x
+
+    cdef bint get_is_zero_unsafe(self, Py_ssize_t i, Py_ssize_t j) except -1:
+        """
+        Return 1 if the entry (i, j) is zero, otherwise 0.
+
+        .. WARNING::
+
+           This is very unsafe; it assumes i and j are in the right
+           range.
+        """
+        return fmpq_is_zero(fmpq_mat_entry(self._matrix, i,j))
+
+    cdef _add_ui_unsafe_assuming_int(self, Py_ssize_t i, Py_ssize_t j, unsigned long int n):
+        # doesn't check immutability
+        # doesn't do bounds checks.
+        # assumes that self[i,j] is an integer.
+        cdef fmpz * entry = fmpq_numref(fmpq_mat_entry(self._matrix, i, j))
+        fmpz_add_ui(entry, entry, n)
+
+    cdef _sub_ui_unsafe_assuming_int(self, Py_ssize_t i, Py_ssize_t j, unsigned long int n):
+        # doesn't check immutability
+        # doesn't do bounds checks.
+        # assumes that self[i,j] is an integer.
+        cdef fmpz * entry = fmpq_numref(fmpq_mat_entry(self._matrix, i, j))
+        fmpz_sub_ui(entry, entry, n)
+
+    def _pickle(self):
+        return self._pickle_version0(), 0
+
+    def _unpickle(self, data, int version):
+        if version == 0:
+            self._unpickle_version0(data)
+        else:
+            raise RuntimeError("unknown matrix version (=%s)" % version)
+
+    cdef _pickle_version0(self):
+        return self._export_as_string(32)
+
+    cpdef _export_as_string(self, int base=10):
+        """
+        Return space separated string of the entries in this matrix, in the
+        given base. This is optimized for speed.
+
+        INPUT:
+
+        - ``base`` -- integer (default: `10`)
+
+        EXAMPLES::
+
+            sage: m = matrix(QQ,2,3,[1,2/3,-3/4,1,-2/3,-45/17])
+            sage: m._export_as_string(10)
+            '1 2/3 -3/4 1 -2/3 -45/17'
+            sage: m._export_as_string(16)
+            '1 2/3 -3/4 1 -2/3 -2d/11'
+        """
+        cdef Py_ssize_t i, j, len_so_far, m, n
+        cdef char *s
+        cdef char *t
+        cdef char *tmp
+
+        if self._nrows == 0 or self._ncols == 0:
+            data = ''
+        else:
+            n = self._nrows * self._ncols * 10
+            s = <char*> sig_malloc(n * sizeof(char))
+            t = s
+            len_so_far = 0
+
+            sig_on()
+            for i in range(self._nrows):
+                for j in range(self._ncols):
+                    m = fmpz_sizeinbase (fmpq_mat_entry_num(self._matrix, i, j), base) + \
+                        fmpz_sizeinbase (fmpq_mat_entry_den(self._matrix, i, j), base) + 3
+                    if len_so_far + m + 1 >= n:
+                        # copy to new string with double the size
+                        n = 2*n + m + 1
+                        tmp = <char*> sig_malloc(n * sizeof(char))
+                        strcpy(tmp, s)
+                        sig_free(s)
+                        s = tmp
+                        t = s + len_so_far
+                    fmpq_get_str(t, base, fmpq_mat_entry(self._matrix, i, j))
+                    m = strlen(t)
+                    len_so_far = len_so_far + m + 1
+                    t = t + m
+                    t[0] = <char>32
+                    t[1] = <char>0
+                    t = t + 1
+            sig_off()
+            data = char_to_str(s)[:-1]
+            sig_free(s)
+        return data
+
+    cdef _unpickle_version0(self, data):
+        r"""
+        TESTS::
+
+            sage: a = random_matrix(QQ, 4, 3, num_bound=2**500, den_bound=2**500)
+            sage: loads(dumps(a)) == a  # indirect doctest
+            True
+        """
+        cdef Py_ssize_t i, j, k
+        data = data.split()
+        if len(data) != self._nrows * self._ncols:
+            raise RuntimeError("invalid pickle data")
+        k = 0
+        for i in range(self._nrows):
+            for j in range(self._ncols):
+                s = data[k]
+                k += 1
+                if '/' in s:
+                    num, den = (n.encode() for n in s.split('/'))
+                    if fmpz_set_str(fmpq_mat_entry_num(self._matrix, i, j), num, 32) or \
+                       fmpz_set_str(fmpq_mat_entry_den(self._matrix, i, j), den, 32):
+                        raise RuntimeError("invalid pickle data")
+                else:
+                    num = s.encode()
+                    if fmpz_set_str(fmpq_mat_entry_num(self._matrix, i, j), num, 32):
+                        raise RuntimeError("invalid pickle data")
+                    fmpz_one(fmpq_mat_entry_den(self._matrix, i, j))
+
+    # #######################################################################
+    # LEVEL 2 functionality
+    # x * cdef _add_
+    # x * cdef _mul_
+    # x * cdef _vector_times_matrix_
+    # x * cpdef _richcmp_
+    # x * __neg__
+    #   * __invert__
+    # x * __copy__
+    # x * _multiply_classical
+    #   * _list -- list of underlying elements (need not be a copy)
+    #   * _dict -- sparse dictionary of underlying elements (need not be a copy)
+    # #######################################################################
+
+    cpdef _lmul_(self, Element right):
+        """
+        EXAMPLES::
+
+            sage: a = matrix(QQ, 2, range(6))
+            sage: (3/4) * a
+            [   0  3/4  3/2]
+            [ 9/4    3 15/4]
+        """
+        cdef Matrix_rational_dense M
+        cdef fmpq_t x
+        fmpq_init(x)
+        fmpq_set_mpq(x, (<Rational>right).value)
+        M = Matrix_rational_dense.__new__(Matrix_rational_dense, self._parent, None, None, None)
+        fmpq_mat_scalar_mul_fmpz(M._matrix, self._matrix, fmpq_numref(x))
+        fmpq_mat_scalar_div_fmpz(M._matrix, M._matrix, fmpq_denref(x))
+        fmpq_clear(x)
+        return M
+
+    cpdef _add_(self, right):
+        """
+        Add two dense matrices over QQ.
+
+        EXAMPLES::
+
+            sage: a = MatrixSpace(QQ,3)(range(9))
+            sage: b = MatrixSpace(QQ,3)([1/n for n in range(1,10)])
+            sage: a+b
+            [   1  3/2  7/3]
+            [13/4 21/5 31/6]
+            [43/7 57/8 73/9]
+            sage: b.swap_rows(1,2)
+            sage: #a+b
+        """
+        cdef Matrix_rational_dense ans
+        ans = Matrix_rational_dense.__new__(Matrix_rational_dense, self._parent, None, None, None)
+
+        sig_on()
+        fmpq_mat_add(ans._matrix, self._matrix, (<Matrix_rational_dense> right)._matrix)
+        sig_off()
+        return ans
+
+    cpdef _sub_(self, right):
+        """
+        Subtract two dense matrices over QQ.
+
+        EXAMPLES::
+
+            sage: a = MatrixSpace(QQ,3)(range(9))
+            sage: b = MatrixSpace(QQ,3)([1/n for n in range(1,10)])
+            sage: a-b
+            [  -1  1/2  5/3]
+            [11/4 19/5 29/6]
+            [41/7 55/8 71/9]
+        """
+        cdef Matrix_rational_dense ans
+        ans = Matrix_rational_dense.__new__(Matrix_rational_dense, self._parent, None, None, None)
+
+        sig_on()
+        fmpq_mat_sub(ans._matrix, self._matrix, (<Matrix_rational_dense> right)._matrix)
+        sig_off()
+        return ans
+
+    cpdef _richcmp_(self, right, int op):
+        r"""
+        TESTS::
+
+            sage: M = MatrixSpace(QQ, 1)
+            sage: M(1) < M(2)
+            True
+            sage: M(1/3) >= M(5/2)
+            False
+            sage: M(2) == M(2)
+            True
+            sage: M(3/4) != M(2)
+            True
+
+            sage: matrix(QQ, 2, 3) == matrix(QQ, 2, 3)
+            True
+            sage: matrix(QQ, 2, 2) == matrix(QQ, 2, 3)
+            False
+            sage: matrix(QQ, 2, 2) == matrix(QQ, 3, 2)
+            False
+            sage: matrix(QQ, 2, 3) == matrix(QQ, 3, 2)
+            False
+
+            sage: mats = [matrix(QQ, 2, 2, 1), matrix(QQ, 2, 2, -1), matrix(QQ, 2, 2, 0)]
+            sage: mats.sort()
+            sage: mats == [-1, 0, 1]
+            True
+        """
+        cdef Py_ssize_t i, j
+        cdef int k
+        for i in range(self._nrows):
+            for j in range(self._ncols):
+                k = fmpq_cmp(fmpq_mat_entry(self._matrix, i, j),
+                             fmpq_mat_entry((<Matrix_rational_dense> right)._matrix, i, j))
+                if k:
+                    if k > 0:
+                        return rich_to_bool(op, 1)
+                    else:
+                        return rich_to_bool(op, -1)
+        return rich_to_bool(op, 0)
+
+    cdef _vector_times_matrix_(self, Vector v):
+        r"""
+        Return the vector times matrix product.
+
+        INPUT:
+
+        - ``v`` -- a free module element
+
+        OUTPUT: the vector times matrix product v\*A
+
+        EXAMPLES::
+
+            sage: B = matrix(QQ,2, [1,2,3,4])
+            sage: V = QQ^2
+            sage: w = V([-1,5/2])
+            sage: w * B
+            (13/2, 8)
+        """
+        cdef Vector_rational_dense w, ans
+        cdef Py_ssize_t i, j
+        cdef mpq_t x, y, z
+
+        M = self.row_ambient_module()
+        w = <Vector_rational_dense> v
+        ans = M.zero_vector()
+
+        mpq_init(x)
+        mpq_init(y)
+        mpq_init(z)
+        for i in range(self._ncols):
+            mpq_set_si(x, 0, 1)
+            for j in range(self._nrows):
+                fmpq_get_mpq(z, fmpq_mat_entry(self._matrix, j, i))
+                mpq_mul(y, w._entries[j], z)
+                mpq_add(x, x, y)
+            mpq_set(ans._entries[i], x)
+        mpq_clear(x)
+        mpq_clear(y)
+        mpq_clear(z)
+        return ans
+
+    def __neg__(self):
+        """
+        Negate a matrix over QQ.
+
+        EXAMPLES::
+
+            sage: a = matrix(QQ, 3, [1/n for n in range(1,10)])
+            sage: -a
+            [  -1 -1/2 -1/3]
+            [-1/4 -1/5 -1/6]
+            [-1/7 -1/8 -1/9]
+        """
+        cdef Matrix_rational_dense ans
+        ans = Matrix_rational_dense.__new__(Matrix_rational_dense, self._parent, None, None, None)
+        fmpq_mat_neg(ans._matrix, self._matrix)
+        return ans
+
+    def __copy__(self):
+        """
+        Copy a matrix over QQ.
+
+        TESTS::
+
+            sage: a = matrix(QQ, 3, [1/n for n in range(1,10)])
+            sage: b = a.__copy__()
+            sage: a == b
+            True
+            sage: a is b
+            False
+            sage: b[0,0] = 5
+            sage: a == b
+            False
+
+            sage: a.subdivide(2, 1)
+            sage: b = a.__copy__()
+            sage: b.subdivisions()
+            ([2], [1])
+            sage: a.subdivide(2, 2)
+            sage: b.subdivisions()
+            ([2], [1])
+        """
+        cdef Matrix_rational_dense ans
+        ans = Matrix_rational_dense.__new__(Matrix_rational_dense, self._parent, None, None, None)
+        fmpq_mat_set(ans._matrix, self._matrix)
+        ans._subdivisions = self._subdivisions
+        return ans
+
+    # #######################################################################
+    # LEVEL 3 functionality (Optional)
+    # x * cdef _sub_
+    #   * __deepcopy__
+    #   * __invert__
+    #   * Matrix windows -- only if you need strassen for that base
+    #   * Other functions (list them here):
+    # x * denom(self):
+    # x * mpz_denom(self, mpz_t d):
+    # x * _clear_denom(self):
+    # o * echelon_modular(self, height_guess=None):
+    # #######################################################################
+    def __invert__(self):
+        """
+        EXAMPLES::
+
+            sage: a = matrix(QQ,3,range(9))
+            sage: a.inverse()
+            Traceback (most recent call last):
+            ...
+            ZeroDivisionError: input matrix must be nonsingular
+            sage: a = matrix(QQ, 2, [1, 5, 17, 3])
+            sage: a.inverse()
+            [-3/82  5/82]
+            [17/82 -1/82]
+            sage: ~matrix(QQ, 2, 3)
+            Traceback (most recent call last):
+            ...
+            ArithmeticError: self must be a square matrix
+        """
+        return self.inverse()
+
+    def _invert_flint(self):
+        r"""
+        TESTS::
+
+            sage: matrix(QQ, 2, [1,2,3,4])._invert_flint()
+            [  -2    1]
+            [ 3/2 -1/2]
+            sage: matrix(QQ, 1)._invert_flint()
+            Traceback (most recent call last):
+            ...
+            ZeroDivisionError: input matrix must be nonsingular
+        """
+        cdef int ret
+        cdef Matrix_rational_dense ans
+        ans = Matrix_rational_dense.__new__(Matrix_rational_dense, self._parent, None, None, None)
+        sig_on()
+        ret = fmpq_mat_inv(ans._matrix, self._matrix)
+        sig_off()
+        if ret == 0:
+            raise ZeroDivisionError("input matrix must be nonsingular")
+        return ans
+
+    def inverse(self, algorithm=None, check_invertible=True):
+        """
+        Return the inverse of this matrix.
+
+        INPUT:
+
+        - ``algorithm`` -- an optional specification of an algorithm. It can be one of
+
+          - ``None``: (default) uses flint
+
+          - ``'flint'``: uses flint library
+
+          - ``'pari'``: uses PARI library
+
+          - ``'iml'``: uses IML library
+
+        - ``check_invertible`` -- only used when ``algorithm=iml``; whether to
+          check that matrix is invertible
+
+        EXAMPLES::
+
+            sage: a = matrix(QQ,3,[1,2,5,3,2,1,1,1,1,])
+            sage: a.inverse()
+            [1/2 3/2  -4]
+            [ -1  -2   7]
+            [1/2 1/2  -2]
+
+            sage: a = matrix(QQ, 2, [1, 5, 17, 3])
+            sage: a.inverse(algorithm='flint')
+            [-3/82  5/82]
+            [17/82 -1/82]
+            sage: a.inverse(algorithm='flint')  * a
+            [1 0]
+            [0 1]
+
+            sage: # needs sage.libs.iml
+            sage: a = matrix(QQ, 2, [-1, 5, 12, -3])
+            sage: a.inverse(algorithm='iml')
+            [1/19 5/57]
+            [4/19 1/57]
+            sage: a.inverse(algorithm='iml') * a
+            [1 0]
+            [0 1]
+
+            sage: a = matrix(QQ, 4, primes_first_n(16))
+            sage: a.inverse(algorithm='pari')                                           # needs sage.libs.pari
+            [   3/11  -12/55    -1/5    2/11]
+            [  -5/11   -2/55    3/10   -3/22]
+            [ -13/22 307/440   -1/10   -9/88]
+            [  15/22  -37/88       0    7/88]
+
+        On singular matrices this method raises a :exc:`ZeroDivisionError`::
+
+            sage: a = matrix(QQ, 2)
+            sage: a.inverse(algorithm='flint')
+            Traceback (most recent call last):
+            ...
+            ZeroDivisionError: input matrix must be nonsingular
+            sage: a.inverse(algorithm='iml')                                            # needs sage.libs.iml
+            Traceback (most recent call last):
+            ...
+            ZeroDivisionError: input matrix must be nonsingular
+            sage: a.inverse(algorithm='pari')                                           # needs sage.libs.pari
+            Traceback (most recent call last):
+            ...
+            ZeroDivisionError: input matrix must be nonsingular
+
+        TESTS::
+
+            sage: a = matrix(QQ, 2)
+            sage: a.inverse(algorithm="IAmNotAnAlgorithm")
+            Traceback (most recent call last):
+            ...
+            ValueError: unknown algorithm 'IAmNotAnAlgorithm'
+
+            sage: for _ in range(30):
+            ....:     dim = randint(1, 20)
+            ....:     a = random_matrix(QQ, dim, num_bound=10, den_bound=10)
+            ....:     while a.rank() != dim: a = random_matrix(QQ, dim)
+            ....:     inv_flint = a.inverse(algorithm='flint')
+            ....:     inv_pari = a.inverse(algorithm='pari')
+            ....:     inv_iml = a.inverse(algorithm='iml')
+            ....:     assert inv_flint == inv_pari == inv_iml
+        """
+        if self._nrows != self._ncols:
+            raise ArithmeticError("self must be a square matrix")
+
+        if self._nrows == 0:
+            return self
+
+        if algorithm is None:
+            algorithm = "flint"
+
+        if algorithm == "flint":
+            return self._invert_flint()
+        elif algorithm == "pari":
+            from cypari2.handle_error import PariError
+            from .matrix_rational_pari import _invert_pari
+            try:
+                return _invert_pari(self)
+            except PariError:
+                raise ZeroDivisionError("input matrix must be nonsingular")
+        elif algorithm == "iml":
+            from .matrix_integer_iml import _invert_iml
+            AZ, denom = self._clear_denom()
+            B, d = _invert_iml(AZ, check_invertible=check_invertible)
+            return (denom/d)*B
+
+        else:
+            raise ValueError("unknown algorithm '%s'" % algorithm)
+
+    def determinant(self, algorithm=None, proof=None):
+        """
+        Return the determinant of this matrix.
+
+        INPUT:
+
+        - ``algorithm`` -- an optional specification of an algorithm. It can be one of
+
+          - ``None``: (default) uses flint
+
+          - ``'flint'``: uses flint library
+
+          - ``'pari'``: uses PARI library
+
+          - ``'integer'``: removes denominator and call determinant on the corresponding
+             integer matrix
+
+          - ``'generic'``: calls the generic Sage implementation
+
+        - ``proof`` -- boolean or ``None``; if ``None`` use
+          proof.linear_algebra(); only relevant for the padic algorithm
+
+        .. NOTE::
+
+           It would be *VERY VERY* hard for det to fail even with
+           proof=False.
+
+        EXAMPLES::
+
+            sage: m = matrix(QQ,3,[1,2/3,4/5, 2,2,2, 5,3,2/5])
+            sage: m.determinant()
+            -34/15
+            sage: m.charpoly()
+            x^3 - 17/5*x^2 - 122/15*x + 34/15
+
+            sage: m = matrix(QQ, 3, [(1/i)**j for i in range(2,5) for j in range(3)])
+            sage: m.determinant(algorithm='flint')
+            -1/288
+
+            sage: m = matrix(QQ, 4, [(-1)**n/n for n in range(1,17)])
+            sage: m.determinant(algorithm='pari')                                       # needs sage.libs.pari
+            2/70945875
+
+            sage: m = matrix(QQ, 5, [1/(i+j+1) for i in range(5) for j in range(5)])
+            sage: m.determinant(algorithm='integer')
+            1/266716800000
+
+        On non-square matrices, the method raises a :exc:`ValueError`::
+
+            sage: matrix(QQ, 2, 3).determinant(algorithm='flint')
+            Traceback (most recent call last):
+            ...
+            ValueError: non square matrix
+            sage: matrix(QQ, 2, 3).determinant(algorithm='pari')                        # needs sage.libs.pari
+            Traceback (most recent call last):
+            ...
+            ValueError: non square matrix
+            sage: matrix(QQ, 2, 3).determinant(algorithm='integer')
+            Traceback (most recent call last):
+            ...
+            ValueError: non square matrix
+            sage: matrix(QQ, 2, 3).determinant(algorithm='generic')
+            Traceback (most recent call last):
+            ...
+            ValueError: non square matrix
+
+        TESTS:
+
+        Check that the four algorithms agree::
+
+            sage: # needs sage.libs.pari
+            sage: for _ in range(20):
+            ....:     dim = randint(0, 30)
+            ....:     m = random_matrix(QQ, dim, num_bound=10, den_bound=10)
+            ....:     det_flint = m.determinant("flint"); m._clear_cache()
+            ....:     det_pari = m.determinant("pari"); m._clear_cache()
+            ....:     det_int = m.determinant("integer"); m._clear_cache()
+            ....:     det_gen = m.determinant("generic")
+            ....:     assert det_flint == det_pari == det_int == det_gen
+        """
+        if self._nrows != self._ncols:
+            raise ValueError("non square matrix")
+
+        det = self.fetch('det')
+        if det is not None:
+            return det
+
+        if algorithm is None or algorithm == "flint":
+            det = self._det_flint()
+        elif algorithm == "pari":
+            from .matrix_rational_pari import _det_pari
+            det = _det_pari(self)
+        elif algorithm == "integer":
+            A, denom = self._clear_denom()
+            det = Rational(A.determinant(proof=proof))
+            if not denom.is_one():
+                det = det / (denom ** self.nrows())
+        elif algorithm == "generic":
+            det = Matrix_dense.determinant(self)
+        else:
+            raise ValueError("unknown algorithm '%s'" % algorithm)
+
+        self.cache('det', det)
+        return det
+
+    def _det_flint(self):
+        r"""
+        Return the determinant (computed using flint).
+
+        EXAMPLES::
+
+            sage: matrix(QQ, 2, [1/3, 2/5, 3/4, 7/8])._det_flint()
+            -1/120
+            sage: matrix(QQ, 0)._det_flint()
+            1
+            sage: matrix(QQ, 1, [0])._det_flint()
+            0
+        """
+        cdef Rational d = Rational.__new__(Rational)
+        cdef fmpq_t e
+        fmpq_init(e)
+        sig_on()
+        fmpq_mat_det(e, self._matrix)
+        fmpq_get_mpq(d.value, e)
+        sig_off()
+        return d
+
+    def denominator(self):
+        """
+        Return the denominator of this matrix.
+
+        OUTPUT: a Sage Integer
+
+        EXAMPLES::
+
+            sage: b = matrix(QQ,2,range(6)); b[0,0]=-5007/293; b
+            [-5007/293         1         2]
+            [        3         4         5]
+            sage: b.denominator()
+            293
+
+            sage: matrix(QQ, 2, [1/2, 1/3, 1/4, 1/5]).denominator()
+            60
+        """
+        cdef Integer z = Integer.__new__(Integer)
+        cdef fmpz_t tmp
+        fmpz_init(tmp)
+        self.fmpz_denom(tmp)
+        fmpz_get_mpz(z.value, tmp)
+        fmpz_clear(tmp)
+        return z
+
+    cdef int fmpz_denom(self, fmpz_t d) except -1:
+        cdef Py_ssize_t i, j
+        sig_on()
+        fmpz_one(d)
+        for i in range(self._nrows):
+            for j in range(self._ncols):
+                fmpz_lcm(d, d, fmpq_mat_entry_den(self._matrix, i, j))
+        sig_off()
+        return 0
+
+    def _clear_denom(self):
+        r"""
+        INPUT:
+
+        - ``self`` -- a matrix
+
+        OUTPUT: ``D*self, D``
+
+        The product is a matrix over `\ZZ`.
+
+        EXAMPLES::
+
+            sage: a = matrix(QQ,2,[-1/6,-7,3,5/4]); a
+            [-1/6   -7]
+            [   3  5/4]
+            sage: b, d = a._clear_denom()
+            sage: b
+            [ -2 -84]
+            [ 36  15]
+            sage: d
+            12
+            sage: b == d * a
+            True
+        """
+        X = self.fetch('clear_denom')
+        if X is not None:
+            return X
+
+        cdef Py_ssize_t i, j
+        cdef Matrix_integer_dense A
+        cdef fmpz * entry
+        cdef fmpz_t denom
+        fmpz_init(denom)
+        self.fmpz_denom(denom)
+
+        from sage.matrix.matrix_space import MatrixSpace
+        MZ = MatrixSpace(ZZ, self._nrows, self._ncols, sparse=False)
+        A = Matrix_integer_dense.__new__(Matrix_integer_dense, MZ, None, None, None)
+
+        sig_on()
+        for i in range(self._nrows):
+            for j in range(self._ncols):
+                entry = fmpz_mat_entry(A._matrix, i, j)
+                fmpz_divexact(entry, denom, fmpq_mat_entry_den(self._matrix, i, j))
+                fmpz_mul(entry, entry, fmpq_mat_entry_num(self._matrix, i, j))
+        sig_off()
+
+        cdef Integer D = PY_NEW(Integer)
+        fmpz_get_mpz(D.value, denom)
+        fmpz_clear(denom)
+        X = (A, D)
+        self.cache('clear_denom', X)
+        return X
+
+    def charpoly(self, var='x', algorithm=None):
+        r"""
+        Return the characteristic polynomial of this matrix.
+
+        .. NOTE::
+
+            The characteristic polynomial is defined as `\det(xI-A)`.
+
+        INPUT:
+
+        - ``var`` -- (optional) name of the variable as a string
+
+        - ``algorithm`` -- an optional specification of an algorithm. It can be
+          one of:
+
+           - ``None``: (default) will use flint for small dimensions and linbox
+             otherwise
+
+           - ``'flint'``: uses flint library
+
+           - ``'linbox'``: uses linbox library
+
+           - ``'generic'``: uses Sage generic implementation
+
+        OUTPUT: a polynomial over the rational numbers
+
+        EXAMPLES::
+
+            sage: a = matrix(QQ, 3, [4/3, 2/5, 1/5, 4, -3/2, 0, 0, -2/3, 3/4])
+            sage: f = a.charpoly(); f
+            x^3 - 7/12*x^2 - 149/40*x + 97/30
+            sage: f(a)
+            [0 0 0]
+            [0 0 0]
+            [0 0 0]
+
+        TESTS:
+
+        The cached polynomial should be independent of the ``var``
+        argument (:issue:`12292`). We check (indirectly) that the
+        second call uses the cached value by noting that its result is
+        not cached::
+
+            sage: M = MatrixSpace(QQ, 2)
+            sage: A = M(range(0, 2^2))
+            sage: type(A)
+            <class 'sage.matrix.matrix_rational_dense.Matrix_rational_dense'>
+            sage: A.charpoly('x')
+            x^2 - 3*x - 2
+            sage: A.charpoly('y')
+            y^2 - 3*y - 2
+            sage: A._cache['charpoly']
+            x^2 - 3*x - 2
+
+        Check consistency::
+
+            sage: for _ in range(100):                                                  # needs sage.libs.linbox
+            ....:     dim = randint(0, 10)
+            ....:     m = random_matrix(QQ, dim, num_bound=8, den_bound=8)
+            ....:     p_flint = m.charpoly(algorithm='flint'); m._clear_cache()
+            ....:     p_linbox = m.charpoly(algorithm='linbox'); m._clear_cache()
+            ....:     p_generic = m.charpoly(algorithm='generic')
+            ....:     assert p_flint == p_linbox == p_generic
+        """
+        poly = self.fetch('charpoly')
+        if poly is not None:
+            return poly.change_variable_name(var)
+
+        if algorithm is None:
+            algorithm = 'flint' if self._nrows <= 40 else 'linbox'
+
+        if algorithm == 'flint' or algorithm == 'linbox':
+            A, denom = self._clear_denom()
+            f = A.charpoly(var, algorithm=algorithm)
+            x = f.parent().gen()
+            g = f(x * denom) / denom ** f.degree()
+        elif algorithm == 'generic':
+            g = Matrix_dense.charpoly(self, var)
+        else:
+            raise ValueError("no algorithm '%s'" % algorithm)
+
+        self.cache('charpoly', g)
+        return g
+
+    def minpoly(self, var='x', algorithm=None):
+        """
+        Return the minimal polynomial of this matrix.
+
+        INPUT:
+
+        - ``var`` -- (optional) the variable name as a string (default: ``'x'``)
+
+        - ``algorithm`` -- an optional specification of an algorithm. It can
+          be one of
+
+           - ``None``: (default) will use linbox
+
+           - ``'linbox'``: uses the linbox library
+
+           - ``'generic'``: uses the generic Sage implementation
+
+        OUTPUT: a polynomial over the rationals
+
+        EXAMPLES::
+
+            sage: a = matrix(QQ, 3, [4/3, 2/5, 1/5, 4, -3/2, 0, 0, -2/3, 3/4])
+            sage: f = a.minpoly(); f
+            x^3 - 7/12*x^2 - 149/40*x + 97/30
+            sage: a = Mat(ZZ,4)(range(16))
+            sage: f = a.minpoly(); f.factor()                                           # needs sage.libs.pari
+            x * (x^2 - 30*x - 80)
+            sage: f(a) == 0                                                             # needs sage.libs.pari
+            True
+
+        ::
+
+            sage: # needs sage.libs.pari
+            sage: a = matrix(QQ, 4, [1..4^2])
+            sage: factor(a.minpoly())
+            x * (x^2 - 34*x - 80)
+            sage: factor(a.minpoly('y'))
+            y * (y^2 - 34*y - 80)
+            sage: factor(a.charpoly())
+            x^2 * (x^2 - 34*x - 80)
+            sage: b = matrix(QQ, 4, [-1, 2, 2, 0, 0, 4, 2, 2, 0, 0, -1, -2, 0, -4, 0, 4])
+            sage: a = matrix(QQ, 4, [1, 1, 0,0, 0,1,0,0, 0,0,5,0, 0,0,0,5])
+            sage: c = b^(-1)*a*b
+            sage: factor(c.minpoly())
+            (x - 5) * (x - 1)^2
+            sage: factor(c.charpoly())
+            (x - 5)^2 * (x - 1)^2
+
+        Check consistency::
+
+            sage: for _ in range(100):                                                  # needs sage.libs.linbox sage.libs.pari
+            ....:     dim = randint(0, 10)
+            ....:     m = random_matrix(QQ, dim, num_bound=8, den_bound=8)
+            ....:     p_linbox = m.charpoly(algorithm='linbox'); m._clear_cache()
+            ....:     p_generic = m.charpoly(algorithm='generic')
+            ....:     assert p_linbox == p_generic
+        """
+        poly = self.fetch('minpoly')
+        if poly is not None:
+            return poly.change_variable_name(var)
+
+        if algorithm is None:
+            try:
+                from .matrix_integer_linbox import _minpoly_linbox
+            except ImportError:
+                algorithm = 'generic'
+            else:
+                algorithm = 'linbox'
+
+        if algorithm == 'linbox':
+            A, denom = self._clear_denom()
+            f = A.minpoly(var, algorithm='linbox')
+            x = f.parent().gen()
+            g = f(x * denom) / denom**f.degree()
+        elif algorithm == 'generic':
+            g = Matrix_dense.minpoly(self, var)
+        else:
+            raise ValueError("no algorithm '%s'" % algorithm)
+
+        self.cache('minpoly', g)
+        return g
+
+    cdef sage.structure.element.Matrix _matrix_times_matrix_(self, sage.structure.element.Matrix right):
+        """
+        EXAMPLES::
+
+            sage: a = matrix(QQ, 3, range(9))/3
+            sage: b = matrix(QQ, 3, range(1, 10))/5
+            sage: a * b   # indirect doctest
+            [ 6/5  7/5  8/5]
+            [18/5 22/5 26/5]
+            [   6 37/5 44/5]
+
+            sage: matrix(QQ, 2, 3) * matrix(QQ, 4, 5)
+            Traceback (most recent call last):
+            ...
+            TypeError: unsupported operand parent(s) for *: 'Full MatrixSpace of 2 by 3 dense matrices over Rational Field' and 'Full MatrixSpace of 4 by 5 dense matrices over Rational Field'
+        """
+        return self._multiply_flint(right)
+
+    def _multiply_flint(self, Matrix_rational_dense right):
+        r"""
+        Multiply this matrix by ``right`` using the flint library.
+
+        EXAMPLES::
+
+            sage: n = 3
+            sage: a = matrix(QQ,n,range(n^2))/3
+            sage: b = matrix(QQ,n,range(1, n^2 + 1))/5
+            sage: a._multiply_flint(b)
+            [ 6/5  7/5  8/5]
+            [18/5 22/5 26/5]
+            [   6 37/5 44/5]
+        """
+        if self._nrows == right._nrows:
+            # self acts on the space of right
+            parent = right.parent()
+        if self._ncols == right._ncols:
+            # right acts on the space of self
+            parent = self.parent()
+        else:
+            parent = self.matrix_space(self._nrows, right._ncols)
+
+        cdef Matrix_rational_dense ans
+        ans = Matrix_rational_dense.__new__(Matrix_rational_dense, parent, None, None, None)
+
+        sig_on()
+        fmpq_mat_mul(ans._matrix, self._matrix, (<Matrix_rational_dense> right)._matrix)
+        sig_off()
+        return ans
+
+    def _multiply_over_integers(self, Matrix_rational_dense right, algorithm='default'):
+        r"""
+        Multiply this matrix by right using a multimodular algorithm and
+        return the result.
+
+        INPUT:
+
+        - ``self`` -- matrix over QQ
+
+        - ``right`` -- matrix over QQ
+
+        - ``algorithm``
+
+           - 'default': use whatever is the default for A\*B when A, B
+             are over ZZ.
+
+           - 'multimodular': use a multimodular algorithm
+
+        EXAMPLES::
+
+            sage: a = MatrixSpace(QQ,10,5)(range(50))
+            sage: b = MatrixSpace(QQ,5,12)([1/n for n in range(1,61)])
+            sage: a._multiply_over_integers(b) == a._multiply_over_integers(b, algorithm='multimodular')
+            True
+
+        ::
+
+            sage: a = MatrixSpace(QQ,3)(range(9))
+            sage: b = MatrixSpace(QQ,3)([1/n for n in range(1,10)])
+            sage: c = a._multiply_over_integers(b, algorithm = 'multimodular')
+            sage: c
+            [ 15/28   9/20   7/18]
+            [  33/7 117/40   20/9]
+            [249/28   27/5  73/18]
+            sage: c == a._multiply_flint(b)
+            True
+        """
+        cdef Matrix_integer_dense A, B, AB
+        cdef Matrix_rational_dense res
+        cdef Integer D
+        sig_on()
+        A, A_denom = self._clear_denom()
+        B, B_denom = right._clear_denom()
+        if algorithm == 'default' or algorithm == 'multimodular':
+            AB = A*B
+        else:
+            sig_off()
+            raise ValueError("unknown algorithm '%s'" % algorithm)
+        D = A_denom * B_denom
+        if self._nrows == right._nrows:
+            # self acts on the space of right
+            res = Matrix_rational_dense.__new__(Matrix_rational_dense, right.parent(), 0, 0, 0)
+        if self._ncols == right._ncols:
+            # right acts on the space of self
+            res = Matrix_rational_dense.__new__(Matrix_rational_dense, self.parent(), 0, 0, 0)
+        else:
+            res = Matrix_rational_dense.__new__(Matrix_rational_dense, self.matrix_space(AB._nrows, AB._ncols), 0, 0, 0)
+        for i in range(res._nrows):
+            for j in range(res._ncols):
+                fmpz_set(fmpq_mat_entry_num(res._matrix, i, j), fmpz_mat_entry(AB._matrix,i,j))
+                fmpz_set_mpz(fmpq_mat_entry_den(res._matrix, i, j), D.value)
+                fmpq_canonicalise(fmpq_mat_entry(res._matrix, i, j))
+        sig_off()
+        return res
+
+    def height(self):
+        """
+        Return the height of this matrix, which is the maximum of the
+        absolute values of all numerators and denominators of entries in
+        this matrix.
+
+        OUTPUT: integer
+
+        EXAMPLES::
+
+            sage: b = matrix(QQ,2,range(6)); b[0,0]=-5007/293; b
+            [-5007/293         1         2]
+            [        3         4         5]
+            sage: b.height()
+            5007
+        """
+        cdef Integer z
+        cdef fmpz_t tmp
+        fmpz_init(tmp)
+        self.fmpz_height(tmp)
+        z = PY_NEW(Integer)
+        fmpz_get_mpz(z.value, tmp)
+        fmpz_clear(tmp)
+        return z
+
+    cdef int fmpz_height(self, fmpz_t h) except -1:
+        cdef fmpz_t x
+        cdef Py_ssize_t i, j
+        sig_on()
+        fmpz_init(x)
+        fmpz_zero(h)
+        for i in range(self._nrows):
+            for j in range(self._ncols):
+                fmpz_abs(x, fmpq_mat_entry_num(self._matrix, i, j))
+                if fmpz_cmp(h, x) < 0:
+                    fmpz_set(h, x)
+                fmpz_abs(x, fmpq_mat_entry_den(self._matrix, i, j))
+                if fmpz_cmp(h, x) < 0:
+                    fmpz_set(h, x)
+        fmpz_clear(x)
+        sig_off()
+        return 0
+
+    def _adjugate(self):
+        """
+        Return the adjugate of this matrix.
+
+        Assumes ``self`` is a square matrix (checked in adjugate).
+
+        EXAMPLES::
+
+            sage: m = matrix(QQ,3,[1..9])/9; m
+            [1/9 2/9 1/3]
+            [4/9 5/9 2/3]
+            [7/9 8/9   1]
+            sage: m.adjugate()                                                          # needs sage.libs.pari
+            [-1/27  2/27 -1/27]
+            [ 2/27 -4/27  2/27]
+            [-1/27  2/27 -1/27]
+        """
+        return self.parent()(self.__pari__().matadjoint().sage())
+
+    def _magma_init_(self, magma):
+        """
+        EXAMPLES::
+
+            sage: m = matrix(QQ,2,3,[1,2/3,-3/4,1,-2/3,-45/17])
+            sage: m._magma_init_(magma)
+            'Matrix(RationalField(),2,3,StringToIntegerSequence("204 136 -153 204 -136 -540"))/204'
+            sage: magma(m)                                                # optional - magma
+            [     1    2/3   -3/4]
+            [     1   -2/3 -45/17]
+        """
+        X, d = self._clear_denom()
+        s = X._magma_init_(magma).replace('IntegerRing','RationalField')
+        if d != 1:
+            s += '/%s' % d._magma_init_(magma)
+        return s
+
+    def prod_of_row_sums(self, cols):
+        cdef Py_ssize_t i, c
+        cdef fmpq_t s, pr
+        fmpq_init(s)
+        fmpq_init(pr)
+
+        fmpq_one(pr)
+        for i in range(self._nrows):
+            fmpq_zero(s)
+            for c in cols:
+                if c < 0 or c >= self._ncols:
+                    raise IndexError("matrix column index out of range")
+                fmpq_add(s, s, fmpq_mat_entry(self._matrix, i, c))
+            fmpq_mul(pr, pr, s)
+        cdef Rational ans
+        ans = Rational.__new__(Rational)
+        fmpq_get_mpq(ans.value, pr)
+        fmpq_clear(s)
+        fmpq_clear(pr)
+        return ans
+
+    def _right_kernel_matrix(self, **kwds):
+        r"""
+        Return a pair that includes a matrix of basis vectors
+        for the right kernel of ``self``.
+
+        INPUT:
+
+        - ``kwds`` -- these are provided for consistency with other versions
+          of this method.  Here they are ignored as there is no optional
+          behavior available.
+
+        OUTPUT:
+
+        Returns a pair.  First item is the string 'computed-iml-rational'
+        that identifies the nature of the basis vectors.
+
+        Second item is a matrix whose rows are a basis for the right kernel,
+        over the rationals, as computed by the IML library.  Notice that the
+        IML library returns a matrix that is in the 'pivot' format, once the
+        whole matrix is multiplied by -1.  So the 'computed' format is very
+        close to the 'pivot' format.
+
+        EXAMPLES::
+
+            sage: A = matrix(QQ, [
+            ....:                 [1, 0, 1, -3, 1],
+            ....:                 [-5, 1, 0, 7, -3],
+            ....:                 [0, -1, -4, 6, -2],
+            ....:                 [4, -1, 0, -6, 2]])
+            sage: result = A._right_kernel_matrix()
+            sage: result[0]                                               # needs sage.libs.iml
+            'computed-iml-rational'
+            sage: result[1]
+            [-1  2 -2 -1  0]
+            [ 1  2  0  0 -1]
+            sage: X = result[1].transpose()
+            sage: A*X == zero_matrix(QQ, 4, 2)
+            True
+
+        Computed result is the negative of the pivot basis, which
+        is just slightly more efficient to compute. ::
+
+            sage: A.right_kernel_matrix(basis='pivot') == -A.right_kernel_matrix(basis='computed')
+            True
+
+        TESTS:
+
+        We test three trivial cases. ::
+
+            sage: A = matrix(QQ, 0, 2)
+            sage: A._right_kernel_matrix()[1]
+            [1 0]
+            [0 1]
+            sage: A = matrix(QQ, 2, 0)
+            sage: A._right_kernel_matrix()[1].parent()
+            Full MatrixSpace of 0 by 0 dense matrices over Rational Field
+            sage: A = zero_matrix(QQ, 4, 3)
+            sage: A._right_kernel_matrix()[1]
+            [1 0 0]
+            [0 1 0]
+            [0 0 1]
+       """
+        tm = verbose("computing right kernel matrix over the rationals for %sx%s matrix" % (self.nrows(), self.ncols()),level=1)
+        # _rational_kernel_flint() gets the zero-row case wrong, fix it there
+        if self.nrows()==0:
+            from sage.matrix.constructor import identity_matrix
+            K = identity_matrix(QQ, self.ncols())
+        else:
+            try:
+                from .matrix_integer_iml import _rational_kernel_iml
+            except ImportError:
+                return self._right_kernel_matrix_over_field()
+            A, _ = self._clear_denom()
+            K = _rational_kernel_iml(A).transpose().change_ring(QQ)
+        verbose("done computing right kernel matrix over the rationals for %sx%s matrix" % (self.nrows(), self.ncols()),level=1, t=tm)
+        return 'computed-iml-rational', K
+
+    # ###############################################
+    # Change ring
+    # ###############################################
+    def change_ring(self, R):
+        """
+        Create the matrix over R with entries the entries of ``self`` coerced
+        into R.
+
+        EXAMPLES::
+
+            sage: a = matrix(QQ,2,[1/2,-1,2,3])
+            sage: a.change_ring(GF(3))
+            [2 2]
+            [2 0]
+            sage: a.change_ring(ZZ)
+            Traceback (most recent call last):
+            ...
+            TypeError: matrix has denominators so can...t change to ZZ
+            sage: b = a.change_ring(QQ['x']); b
+            [1/2  -1]
+            [  2   3]
+            sage: b.parent()
+            Full MatrixSpace of 2 by 2 dense matrices over Univariate Polynomial Ring in x over Rational Field
+
+        TESTS:
+
+        Make sure that subdivisions are preserved when changing rings::
+
+            sage: a = matrix(QQ, 3, range(9))
+            sage: a.subdivide(2,1); a
+            [0|1 2]
+            [3|4 5]
+            [-+---]
+            [6|7 8]
+            sage: a.change_ring(ZZ).change_ring(QQ)
+            [0|1 2]
+            [3|4 5]
+            [-+---]
+            [6|7 8]
+            sage: a.change_ring(GF(3))
+            [0|1 2]
+            [0|1 2]
+            [-+---]
+            [0|1 2]
+        """
+        if R not in Rings():
+            raise TypeError("R must be a ring")
+        if R == self._base_ring:
+            if self._is_immutable:
+                return self
+            return self.__copy__()
+        if isinstance(R, IntegerRing_class):
+            A, d = self._clear_denom()
+            if not d.is_one():
+                raise TypeError("matrix has denominators so can't change to ZZ")
+            if self._subdivisions is not None:
+                A.subdivide(self.subdivisions())
+            return A
+
+        if isinstance(R, sage.rings.abc.IntegerModRing):
+            try:
+                from sage.matrix.matrix_modn_dense_double import MAX_MODULUS
+            except ImportError:
+                pass
+            else:
+                if R.order() < MAX_MODULUS:
+                    b = R.order()
+                    A, d = self._clear_denom()
+                    if not b.gcd(d).is_one():
+                        raise TypeError("matrix denominator not coprime to modulus")
+                    B = A._mod_int(b)
+                    C = (1/(B.base_ring()(d))) * B
+                    if self._subdivisions is not None:
+                        C.subdivide(self.subdivisions())
+                    return C
+
+        # fallback to the generic version
+        return Matrix_dense.change_ring(self, R)
+
+    # ###############################################
+    # Echelon form
+    # ###############################################
+    def echelonize(self, algorithm=None,
+                   height_guess=None, proof=None, **kwds):
+        """
+        Transform the matrix ``self`` into reduced row echelon form
+        in place.
+
+        INPUT:
+
+        - ``algorithm`` -- an optional specification of an algorithm. One of
+
+          - ``None``: (default) try to pick the best choice,
+
+          - ``'flint'``: use flint library
+            `function <https://flintlib.org/doc/fmpq_mat.html#c.fmpq_mat_rref>`_,
+            which automatically chooses between
+            `classical algorithm <https://flintlib.org/doc/fmpq_mat.html#c.fmpq_mat_rref_classical>`_
+            (Gaussian elimination),
+            `fraction-free multimodular <https://flintlib.org/doc/fmpz_mat.html#c.fmpz_mat_rref_mul>`_,
+            and `fraction-free LU decomposition <https://flintlib.org/doc/fmpz_mat.html#c.fmpz_mat_rref_fflu>`_,
+
+          - ``'flint:classical'``, ``'flint:multimodular'``, ``'flint:fflu'``: use the
+            flint library as above, but select an algorithm explicitly,
+
+          - ``'padic'``: an algorithm based on the IML `p`-adic solver,
+
+          - ``'multimodular'``: uses a multimodular algorithm implemented in Cython
+            that uses linbox modulo many primes,
+            see :func:`~sage.matrix.misc.matrix_rational_echelon_form_multimodular`,
+
+          - ``'classical'``: just clear each column using Gauss elimination.
+
+        - ``height_guess``, ``**kwds`` -- all passed to the
+          ``'multimodular'`` algorithm; ignored by other algorithms
+
+        - ``proof`` -- boolean or ``None`` (default: None, see
+          proof.linear_algebra or sage.structure.proof). Passed to the
+          ``'multimodular'`` algorithm. Note that the Sage global default is
+          ``proof=True``.
+
+        EXAMPLES::
+
+            sage: a = matrix(QQ, 4, range(16)); a[0,0] = 1/19; a[0,1] = 1/5; a
+            [1/19  1/5    2    3]
+            [   4    5    6    7]
+            [   8    9   10   11]
+            [  12   13   14   15]
+            sage: a.echelonize()
+            sage: a
+            [      1       0       0 -76/157]
+            [      0       1       0  -5/157]
+            [      0       0       1 238/157]
+            [      0       0       0       0]
+
+        ::
+
+            sage: # needs sage.libs.linbox
+            sage: a = matrix(QQ, 4, range(16)); a[0,0] = 1/19; a[0,1] = 1/5
+            sage: a.echelonize(algorithm='multimodular')
+            sage: a
+            [      1       0       0 -76/157]
+            [      0       1       0  -5/157]
+            [      0       0       1 238/157]
+            [      0       0       0       0]
+
+        TESTS:
+
+        Echelonizing a matrix in place throws away the cache of
+        the old matrix (:issue:`14506`)::
+
+            sage: for algo in ["flint", "padic", "multimodular", "classical",           # needs sage.libs.linbox
+            ....:              "flint:classical", "flint:multimodular", "flint:fflu"]:
+            ....:      a = Matrix(QQ, [[1,2],[3,4]])
+            ....:      _ = a.det()          # fills the cache
+            ....:      _ = a._clear_denom() # fills the cache
+            ....:      a.echelonize(algorithm=algo)
+            ....:      assert sorted(a._cache.keys()) == ['echelon_form', 'in_echelon_form', 'pivots', 'rank'], (algo, a._cache.keys())
+        """
+        if self.fetch('in_echelon_form'):
+            return  # already known to be in echelon form
+        self.check_mutability()
+
+        if algorithm is None:
+            algorithm = 'flint:multimodular'
+
+        if algorithm in ('flint', 'flint:classical', 'flint:multimodular', 'flint:fflu'):
+            pivots = self._echelonize_flint(algorithm)
+        elif algorithm == 'multimodular':
+            pivots = self._echelonize_multimodular(height_guess, proof, **kwds)
+        elif algorithm == 'classical':
+            pivots = self._echelon_in_place_classical()
+        elif algorithm == 'padic':
+            pivots = self._echelonize_padic()
+        else:
+            raise ValueError("no algorithm '%s'" % algorithm)
+
+        if type(pivots) is not tuple:
+            raise RuntimeError("BUG: pivots must get set as a tuple. Got {} for algo {} with {}x{} matrix.".format(
+                type(pivots), algorithm, self._nrows, self._ncols))
+
+        self.cache('in_echelon_form', True)
+        self.cache('echelon_form', self)
+        self.cache('pivots', pivots)
+        self.cache('rank', len(pivots))
+
+    def echelon_form(self, algorithm=None,
+                     height_guess=None, proof=None, **kwds):
+        r"""
+        Return the echelon form of this matrix.
+
+        The (row) echelon form of a matrix, see :wikipedia:`Row_echelon_form`,
+        is the matrix obtained by performing Gauss elimination on the rows
+        of the matrix.
+
+        INPUT: See :meth:`echelonize` for the options.
+
+        EXAMPLES::
+
+            sage: a = matrix(QQ, 4, range(16)); a[0,0] = 1/19; a[0,1] = 1/5; a
+            [1/19  1/5    2    3]
+            [   4    5    6    7]
+            [   8    9   10   11]
+            [  12   13   14   15]
+            sage: a.echelon_form()
+            [      1       0       0 -76/157]
+            [      0       1       0  -5/157]
+            [      0       0       1 238/157]
+            [      0       0       0       0]
+            sage: a.echelon_form(algorithm='multimodular')
+            [      1       0       0 -76/157]
+            [      0       1       0  -5/157]
+            [      0       0       1 238/157]
+            [      0       0       0       0]
+
+        The result is an immutable matrix, so if you want to modify the result
+        then you need to make a copy.  This checks that :issue:`10543` is
+        fixed.::
+
+            sage: A = matrix(QQ, 2, range(6))
+            sage: E = A.echelon_form()
+            sage: E.is_mutable()
+            False
+            sage: F = copy(E)
+            sage: F[0,0] = 50
+            sage: F
+            [50  0 -1]
+            [ 0  1  2]
+
+        TESTS:
+
+        Check consistency::
+
+            sage: for _ in range(100):                                                  # needs sage.libs.linbox
+            ....:     nrows = randint(0, 30)
+            ....:     ncols = randint(0, 30)
+            ....:     m = random_matrix(QQ, nrows, ncols, num_bound=10, den_bound=10)
+            ....:     ech_flint = m.echelon_form('flint'); m._clear_cache()
+            ....:     ech_padic = m.echelon_form('padic'); m._clear_cache()
+            ....:     ech_multi = m.echelon_form('multimodular'); m._clear_cache()
+            ....:     ech_class = m.echelon_form('classical')
+            ....:     assert ech_flint == ech_padic == ech_multi == ech_class
+        """
+        x = self.fetch('echelon_form')
+        if x is not None:
+            return x
+        if self.fetch('in_echelon_form'):
+            raise RuntimeError('in_echelon_form set but not echelon_form')
+
+        E = self.__copy__()
+        E.echelonize(algorithm)
+        E.set_immutable()
+        self.cache('echelon_form', E)
+        self.cache('pivots', E.pivots())
+        self.cache('rank', len(E.pivots()))
+        return E
+
+    def _echelonize_flint(self, algorithm: str):
+        r"""
+        INPUT: See :meth:`echelonize` for the options.
+        Only options that use flint are allowed, passing other algorithms may
+        trigger undefined behavior.
+
+        EXAMPLES::
+
+            sage: m = matrix(QQ, 4, range(16))
+            sage: m._echelonize_flint("flint")
+            (0, 1)
+            sage: m
+            [ 1  0 -1 -2]
+            [ 0  1  2  3]
+            [ 0  0  0  0]
+            [ 0  0  0  0]
+            sage: m = matrix(QQ, 4, 6, [-1,0,0,-2,-1,-2,-1,0,0,-2,-1,0,3,3,-2,0,0,3,-2,-3,1,1,-2,3])
+            sage: m._echelonize_flint("flint")
+            (0, 1, 2, 5)
+            sage: m
+            [   1    0    0    2    1    0]
+            [   0    1    0 -4/3    1    0]
+            [   0    0    1    1    3    0]
+            [   0    0    0    0    0    1]
+        """
+        self.clear_cache()
+
+        if fmpq_mat_is_empty(self._matrix):
+            return ()
+
+        cdef long r
+        cdef fmpz_mat_t Aclear
+        cdef fmpz_t den
+
+        sig_on()
+
+        if algorithm == 'flint':
+            r = fmpq_mat_rref(self._matrix, self._matrix)
+        elif algorithm == 'flint:classical':
+            r = fmpq_mat_rref_classical(self._matrix, self._matrix)
+        else:
+            # copied from fmpq_mat_rref_fraction_free
+            fmpz_mat_init(Aclear, self._nrows, self._ncols)
+            fmpq_mat_get_fmpz_mat_rowwise(Aclear, NULL, self._matrix)
+            fmpz_init(den)
+
+            if algorithm == 'flint:fflu':
+                r = fmpz_mat_rref_fflu(Aclear, den, Aclear)
+            else:
+                assert algorithm == 'flint:multimodular'
+                r = fmpz_mat_rref_mul(Aclear, den, Aclear)
+
+            if r == 0:
+                fmpq_mat_zero(self._matrix)
+            else:
+                fmpq_mat_set_fmpz_mat_div_fmpz(self._matrix, Aclear, den)
+
+            fmpz_mat_clear(Aclear)
+            fmpz_clear(den)
+
+        sig_off()
+
+        # compute pivots
+        cdef long i, j, k
+        cdef list p = []
+        k = 0
+        for i in range(r):
+            for j in range(k, self._ncols):
+                if not fmpq_is_zero(fmpq_mat_entry(self._matrix, i, j)):
+                    p.append(j)
+                    k = j+1  # so start at next position next time
+                    break
+            else:
+                break
+        return tuple(p)
+
+    def _echelonize_padic(self):
+        """
+        Echelonize ``self`` using a `p`-adic nullspace algorithm.
+
+        EXAMPLES::
+
+            sage: # needs sage.libs.linbox
+            sage: m = matrix(QQ, 4, range(16))
+            sage: m._echelonize_padic()
+            (0, 1)
+            sage: m
+            [ 1  0 -1 -2]
+            [ 0  1  2  3]
+            [ 0  0  0  0]
+            [ 0  0  0  0]
+            sage: m = matrix(QQ, 4, 6, [-1,0,0,-2,-1,-2,-1,0,0,-2,-1,0,3,3,-2,0,0,3,-2,-3,1,1,-2,3])
+            sage: m._echelonize_padic()
+            (0, 1, 2, 5)
+            sage: m
+            [   1    0    0    2    1    0]
+            [   0    1    0 -4/3    1    0]
+            [   0    0    1    1    3    0]
+            [   0    0    0    0    0    1]
+        """
+        cdef Matrix_integer_dense X
+        cdef Integer d
+        cdef fmpq * entry
+
+        A, _ = self._clear_denom()
+        pivots, nonpivots, X, d = A._rational_echelon_via_solve()
+        self.clear_cache()
+
+        # FIXME: we should always have X.nrows() == len(pivots)
+        if X.nrows() != len(pivots):
+            assert X.ncols() == len(pivots) == 0
+            assert type(pivots) is list
+            fmpq_mat_zero(self._matrix)
+            return tuple(pivots)
+
+        cdef Py_ssize_t i,j
+        for i in range(X.nrows()):
+            # 1 at pivot
+            fmpq_one(fmpq_mat_entry(self._matrix, i, pivots[i]))
+
+            # nonzero part
+            for j in range(X.ncols()):
+                entry = fmpq_mat_entry(self._matrix, i, nonpivots[j])
+                fmpz_set(fmpq_numref(entry), fmpz_mat_entry(X._matrix, i, j))
+                fmpz_set_mpz(fmpq_denref(entry), d.value)
+                fmpq_canonicalise(entry)
+
+            # zeros on the left of the pivot
+            for j in range(pivots[i]):
+                fmpq_zero(fmpq_mat_entry(self._matrix, i, j))
+
+            # zeros on top of the other pivots
+            for j in range(i):
+                fmpq_zero(fmpq_mat_entry(self._matrix, j, pivots[i]))
+
+        # Fill in the 0-rows at the bottom.
+        for i in range(len(pivots), self._nrows):
+            for j in range(self._ncols):
+                fmpq_zero(fmpq_mat_entry(self._matrix, i, j))
+
+        # FIXME: pivots should already be a tuple in all cases
+        return tuple(pivots)
+
+    def _echelonize_multimodular(self, height_guess=None, proof=None):
+        """
+        Echelonize ``self`` using multimodular recomposition.
+
+        REFERENCE:
+
+        - Chapter 7 of Stein's "Explicitly Computing Modular Forms".
+
+        INPUT:
+
+        - ``height_guess`` -- integer or ``None``
+
+        - ``proof`` -- boolean (default: ``None``, see
+          proof.linear_algebra or sage.structure.proof); Note that the Sage
+          global default is ``proof=True``
+
+        EXAMPLES::
+
+            sage: # needs sage.libs.linbox
+            sage: m = matrix(QQ, 4, range(16))
+            sage: m._echelonize_multimodular()
+            (0, 1)
+            sage: m
+            [ 1  0 -1 -2]
+            [ 0  1  2  3]
+            [ 0  0  0  0]
+            [ 0  0  0  0]
+            sage: m = matrix(QQ, 4, 6, [-1,0,0,-2,-1,-2,-1,0,0,-2,-1,0,3,3,-2,0,0,3,-2,-3,1,1,-2,3])
+            sage: m._echelonize_multimodular()
+            (0, 1, 2, 5)
+            sage: m
+            [   1    0    0    2    1    0]
+            [   0    1    0 -4/3    1    0]
+            [   0    0    1    1    3    0]
+            [   0    0    0    0    0    1]
+        """
+        from sage.matrix.misc import matrix_rational_echelon_form_multimodular
+        E, pivots = matrix_rational_echelon_form_multimodular(self, height_guess, proof=proof)
+        self.clear_cache()
+        fmpq_mat_swap(self._matrix, (<Matrix_rational_dense>E)._matrix)
+        return pivots
+
+    cdef swap_rows_c(self, Py_ssize_t r1, Py_ssize_t r2):
+        """
+        EXAMPLES::
+
+            sage: a = matrix(QQ,2,[1..6])
+            sage: a.swap_rows(0,1)             # indirect doctest
+            sage: a
+            [4 5 6]
+            [1 2 3]
+        """
+        # no bounds checking!
+        cdef Py_ssize_t c
+        for c in range(self._ncols):
+            fmpq_swap(fmpq_mat_entry(self._matrix, r1, c),
+                      fmpq_mat_entry(self._matrix, r2, c))
+
+    cdef swap_columns_c(self, Py_ssize_t c1, Py_ssize_t c2):
+        """
+        EXAMPLES::
+
+            sage: a = matrix(QQ,2,[1..6])
+            sage: a.swap_columns(0,1)          # indirect doctest
+            sage: a
+            [2 1 3]
+            [5 4 6]
+        """
+        # no bounds checking!
+        for r in range(self._nrows):
+            fmpq_swap(fmpq_mat_entry(self._matrix, r, c1),
+                      fmpq_mat_entry(self._matrix, r, c2))
+
+    def decomposition(self, is_diagonalizable=False, dual=False,
+                      algorithm=None, height_guess=None, proof=None):
+        """
+        Return the decomposition of the free module on which this matrix A
+        acts from the right (i.e., the action is x goes to x A), along with
+        whether this matrix acts irreducibly on each factor. The factors
+        are guaranteed to be sorted in the same way as the corresponding
+        factors of the characteristic polynomial.
+
+        Let A be the matrix acting from the on the vector space V of column
+        vectors. Assume that A is square. This function computes maximal
+        subspaces W_1, ..., W_n corresponding to Galois conjugacy classes
+        of eigenvalues of A. More precisely, let f(X) be the characteristic
+        polynomial of A. This function computes the subspace
+        `W_i = ker(g_(A)^n)`, where g_i(X) is an irreducible
+        factor of f(X) and g_i(X) exactly divides f(X). If the optional
+        parameter is_diagonalizable is True, then we let W_i = ker(g(A)),
+        since then we know that ker(g(A)) = `ker(g(A)^n)`.
+
+        If dual is True, also returns the corresponding decomposition of V
+        under the action of the transpose of A. The factors are guaranteed
+        to correspond.
+
+        INPUT:
+
+        - ``is_diagonalizable`` -- ignored
+
+        - ``dual`` -- whether to also return decompositions for
+          the dual
+
+        - ``algorithm`` -- an optional specification of an algorithm
+
+           - ``None`` -- (default) use default algorithm for computing Echelon
+             forms
+
+           - 'multimodular': much better if the answers
+             factors have small height
+
+        - ``height_guess`` -- positive integer; only used by
+          the multimodular algorithm
+
+        - ``proof`` -- boolean or ``None`` (default: ``None``, see
+          proof.linear_algebra or sage.structure.proof); only used by the
+          multimodular algorithm. Note that the Sage global default is
+          proof=True.
+
+
+        .. NOTE::
+
+           IMPORTANT: If you expect that the subspaces in the answer
+           are spanned by vectors with small height coordinates, use
+           algorithm='multimodular' and height_guess=1; this is
+           potentially much faster than the default. If you know for a
+           fact the answer will be very small, use
+           algorithm='multimodular', height_guess=bound on height,
+           proof=False.
+
+        You can get very very fast decomposition with proof=False.
+
+        EXAMPLES::
+
+            sage: a = matrix(QQ,3,[1..9])
+            sage: a.decomposition()
+            [(Vector space of degree 3 and dimension 1 over Rational Field
+              Basis matrix:
+              [ 1 -2  1],
+              True),
+             (Vector space of degree 3 and dimension 2 over Rational Field
+              Basis matrix:
+              [ 1  0 -1]
+              [ 0  1  2],
+              True)]
+        """
+        X = self._decomposition_rational(is_diagonalizable=is_diagonalizable,
+                                         echelon_algorithm=algorithm,
+                                         height_guess=height_guess, proof=proof)
+        if not dual:
+            return X
+
+        Y = self.transpose()._decomposition_rational(is_diagonalizable=is_diagonalizable,
+            echelon_algorithm=algorithm, height_guess=height_guess, proof=proof)
+        return X, Y
+
+    def _decomposition_rational(self, is_diagonalizable = False,
+                                echelon_algorithm=None,
+                                kernel_algorithm='default',
+                                **kwds):
+        """
+        Return the decomposition of the free module on which this matrix A
+        acts from the right (i.e., the action is x goes to x A), along with
+        whether this matrix acts irreducibly on each factor. The factors
+        are guaranteed to be sorted in the same way as the corresponding
+        factors of the characteristic polynomial.
+
+        INPUT:
+
+        - ``self`` -- a square matrix over the rational
+          numbers
+
+        - ``echelon_algorithm`` -- an optional algorithm to be passed to the
+          method ``echelon_form``
+
+        - ``'multimodular'`` -- use this if the answers have
+          small height
+
+        - ``**kwds`` -- passed on to echelon function
+
+        .. NOTE::
+
+           IMPORTANT: If you expect that the subspaces in the answer are
+           spanned by vectors with small height coordinates, use
+           algorithm='multimodular' and height_guess=1; this is potentially
+           much faster than the default. If you know for a fact the answer
+           will be very small, use algorithm='multimodular',
+           height_guess=bound on height, proof=False
+
+        OUTPUT:
+
+        - ``Sequence`` -- list of tuples (V,t), where V is a
+          vector spaces and t is ``True`` if and only if the charpoly of ``self`` on
+          V is irreducible. The tuples are in order corresponding to the
+          elements of the sorted list self.charpoly().factor().
+        """
+        cdef Py_ssize_t k
+
+        if not self.is_square():
+            raise ArithmeticError("self must be a square matrix")
+
+        if self.nrows() == 0:
+            return decomp_seq([])
+
+        A, _ = self._clear_denom()
+
+        f = A.charpoly('x')
+        E = decomp_seq([])
+
+        t = verbose('factoring the characteristic polynomial', level=2, caller_name='rational decomp')
+        F = f.factor()
+        verbose('done factoring', t=t, level=2, caller_name='rational decomp')
+
+        if len(F) == 1:
+            V = QQ**self.nrows()
+            m = F[0][1]
+            return decomp_seq([(V, m==1)])
+
+        V = ZZ**self.nrows()
+        v = V.random_element()
+        num_iterates = max([0] + [f.degree() - g.degree() for g, _ in F if g.degree() > 1]) + 1
+
+        S = []
+
+        F.sort()
+        for i in range(len(F)):
+            g, m = F[i]
+
+            if g.degree() == 1:
+                # Just use kernel -- much easier.
+                B = A.__copy__()
+                for k from 0 <= k < A.nrows():
+                    B[k,k] += g[0]
+                if m > 1 and not is_diagonalizable:
+                    B = B**m
+                B = B.change_ring(QQ)
+                W = B.kernel(algorithm = kernel_algorithm, **kwds)
+                E.append((W, m==1))
+                continue
+
+            # General case, i.e., deg(g) > 1:
+            W = None
+            tries = m
+            while True:
+
+                # Compute the complementary factor of the charpoly.
+                h = f // (g**m)
+                v = h.list()
+
+                while len(S) < tries:
+                    t = verbose('%s-spinning %s-th random vector' % (num_iterates, len(S)),
+                                level=2, caller_name='rational decomp')
+                    S.append(A.iterates(V.random_element(x=-10,y=10), num_iterates))
+                    verbose('done spinning', level=2, t=t, caller_name='rational decomp')
+
+                for j in range(0 if W is None else W.nrows() // g.degree(), len(S)):
+                    # Compute one element of the kernel of g(A)**m.
+                    t = verbose('compute element of kernel of g(A), for g of degree %s' % g.degree(), level=2,
+                                caller_name='rational decomp')
+                    w = S[j].linear_combination_of_rows(h.list())
+                    t = verbose('done computing element of kernel of g(A)', t=t,level=2, caller_name='rational decomp')
+
+                    # Get the rest of the kernel.
+                    t = verbose('fill out rest of kernel',level=2, caller_name='rational decomp')
+                    if W is None:
+                        W = A.iterates(w, g.degree())
+                    else:
+                        W = W.stack(A.iterates(w, g.degree()))
+                    t = verbose('finished filling out more of kernel',level=2, t=t, caller_name='rational decomp')
+
+                if W.rank() == m * g.degree():
+                    W = W.change_ring(QQ)
+                    t = verbose('now computing row space', level=2, caller_name='rational decomp')
+                    W.echelonize(algorithm = echelon_algorithm, **kwds)
+                    E.append((W.row_space(), m==1))
+                    verbose('computed row space', level=2,t=t, caller_name='rational decomp')
+                    break
+                else:
+                    verbose('we have not yet generated all the kernel (rank so far=%s, target rank=%s)' % (W.rank(), m*g.degree()),
+                            level=2, caller_name='rational decomp')
+                    tries += 1
+                    if tries > 5*m:
+                        raise RuntimeError("likely bug in decomposition")
+                # end if
+            # end while
+        # end for
+        return decomp_seq(E)
+
+#      def simple_decomposition(self, echelon_algorithm='default', **kwds):
+#          """
+#          Returns the decomposition of the free module on which this
+#          matrix A acts from the right (i.e., the action is x goes to x
+#          A), as a direct sum of simple modules.
+
+#          NOTE: self *must* be diagonalizable.
+
+#          INPUT:
+#              self -- a square matrix that is assumed to be diagonalizable
+#              echelon_algorithm -- 'default'
+#                                   'multimodular' -- use this if the answers
+#                                   have small height
+#              **kwds -- passed on to echelon function.
+
+#          IMPORTANT NOTE:
+#          If you expect that the subspaces in the answer are spanned by vectors
+#          with small height coordinates, use algorithm='multimodular' and
+#          height_guess=1; this is potentially much faster than the default.
+#          If you know for a fact the answer will be very small, use
+#             algorithm='multimodular', height_guess=bound on height, proof=False
+
+#          OUTPUT:
+#              Sequence -- list of tuples (V,g), where V is a subspace
+#                          and an irreducible polynomial g, which is the
+#                          charpoly (=minpoly) of self acting on V.
+#          """
+#          cdef Py_ssize_t k
+
+#          if not self.is_square():
+#              raise ArithmeticError("self must be a square matrix")
+
+#          if self.nrows() == 0:
+#              return decomp_seq([])
+
+#          A, _ = self._clear_denom()
+
+#          f = A.charpoly('x')
+#          E = decomp_seq([])
+
+#          t = verbose('factoring the characteristic polynomial', level=2, caller_name='simple decomp')
+#          F = f.factor()
+#          G = [g for g, _ in F]
+#          minpoly = prod(G)
+#          squarefree_degree = sum([g.degree() for g in G])
+#          verbose('done factoring', t=t, level=2, caller_name='simple decomp')
+
+#          V = ZZ**self.nrows()
+#          v = V.random_element()
+#          num_iterates = max([squarefree_degree - g.degree() for g in G]) + 1
+
+#          S = [ ]
+
+#          F.sort()
+#          for i in range(len(F)):
+#              g, m = F[i]
+
+#              if g.degree() == 1:
+#                  # Just use kernel -- much easier.
+#                  B = A.__copy__()
+#                  for k from 0 <= k < A.nrows():
+#                      B[k,k] += g[0]
+#                  if m > 1 and not is_diagonalizable:
+#                      B = B**m
+#                  W = B.change_ring(QQ).kernel()
+#                  for b in W.basis():
+#                      E.append((W.span(b), g))
+#                  continue
+
+#              # General case, i.e., deg(g) > 1:
+#              W = None
+#              while True:
+
+#                  # Compute the complementary factor of the charpoly.
+#                  h = minpoly // g
+#                  v = h.list()
+
+#                  while len(S) < m:
+#                      t = verbose('%s-spinning %s-th random vector'%(num_iterates, len(S)),
+#                                  level=2, caller_name='simple decomp')
+#                      S.append(A.iterates(V.random_element(x=-10,y=10), num_iterates))
+#                      verbose('done spinning', level=2, t=t, caller_name='simple decomp')
+
+#                  for j in range(len(S)):
+#                      # Compute one element of the kernel of g(A).
+#                      t = verbose('compute element of kernel of g(A), for g of degree %s'%g.degree(),level=2,
+#                              caller_name='simple decomp')
+#                      w = S[j].linear_combination_of_rows(h.list())
+#                      t = verbose('done computing element of kernel of g(A)', t=t,level=2, caller_name='simple decomp')
+
+#                      # Get the rest of the kernel.
+#                      t = verbose('fill out rest of kernel',level=2, caller_name='simple decomp')
+#                      if W is None:
+#                          W = A.iterates(w, g.degree())
+#                      else:
+#                          W = W.stack(A.iterates(w, g.degree()))
+#                      t = verbose('finished filling out more of kernel',level=2, t=t, caller_name='simple decomp')
+
+#                  if W.rank() == m * g.degree():
+#                      W = W.change_ring(QQ)
+#                      t = verbose('now computing row space', level=2, caller_name='simple decomp')
+#                      W.echelonize(algorithm = echelon_algorithm, **kwds)
+#                      E.append((W.row_space(), m==1))
+#                      verbose('computed row space', level=2,t=t, caller_name='simple decomp')
+#                      break
+#                  else:
+#                      verbose('we have not yet generated all the kernel (rank so far=%s, target rank=%s)'%(
+#                          W.rank(), m*g.degree()), level=2, caller_name='simple decomp')
+#                      j += 1
+#                      if j > 3*m:
+#                          raise RuntimeError("likely bug in decomposition")
+#                  # end if
+#              #end while
+#          #end for
+#          return E
+
+    def _lift_crt_rr(self, res, mm):
+        from .matrix_rational_linbox import _lift_crt_rr
+        return _lift_crt_rr(self, res, mm)
+
+    def randomize(self, density=1, num_bound=2, den_bound=2,
+                  distribution=None, nonzero=False):
+        """
+        Randomize ``density`` proportion of the entries of this matrix, leaving
+        the rest unchanged.
+
+        If ``x`` and ``y`` are given, randomized entries of this matrix have
+        numerators and denominators bounded by ``x`` and ``y`` and have
+        density 1.
+
+        INPUT:
+
+        - ``density`` -- number between 0 and 1 (default: 1)
+
+        - ``num_bound`` -- numerator bound (default: 2)
+
+        - ``den_bound`` -- denominator bound (default: 2)
+
+        - ``distribution`` -- ``None`` or '1/n' (default: ``None``); if '1/n'
+          then ``num_bound``, ``den_bound`` are ignored and numbers are chosen
+          using the GMP function ``mpq_randomize_entry_recip_uniform``
+
+        OUTPUT: none; the matrix is modified in-space
+
+        EXAMPLES:
+
+        The default distribution::
+
+            sage: from collections import defaultdict
+            sage: total_count = 0
+            sage: dic = defaultdict(Integer)
+            sage: def add_samples(distribution=None):
+            ....:     global dic, total_count
+            ....:     for _ in range(100):
+            ....:         A = Matrix(QQ, 2, 4, 0)
+            ....:         A.randomize(distribution=distribution)
+            ....:         for a in A.list():
+            ....:             dic[a] += 1
+            ....:             total_count += 1.0
+
+            sage: expected = {-2: 1/9, -1: 3/18, -1/2: 1/18, 0: 3/9,
+            ....:             1/2: 1/18, 1: 3/18, 2: 1/9}
+            sage: add_samples()
+            sage: while not all(abs(dic[a]/total_count - expected[a]) < 0.001 for a in dic):
+            ....:     add_samples()
+
+        The distribution ``'1/n'``::
+
+            sage: def mpq_randomize_entry_recip_uniform():
+            ....:     r = 2*random() - 1
+            ....:     if r == 0: r = 1
+            ....:     num = int(4/(5*r))
+            ....:     r = random()
+            ....:     if r == 0: r = 1
+            ....:     den = int(1/random())
+            ....:     return Integer(num)/Integer(den)
+
+            sage: total_count = 0
+            sage: dic = defaultdict(Integer)
+            sage: dic2 = defaultdict(Integer)
+            sage: add_samples('1/n')
+            sage: for _ in range(8):
+            ....:     dic2[mpq_randomize_entry_recip_uniform()] += 1
+            sage: while not all(abs(dic[a] - dic2[a])/total_count < 0.005 for a in dic):
+            ....:     add_samples('1/n')
+            ....:     for _ in range(800):
+            ....:         dic2[mpq_randomize_entry_recip_uniform()] += 1
+
+        The default can be used to obtain matrices of different rank::
+
+            sage: ranks = [False]*11
+            sage: while not all(ranks):
+            ....:     for dens in (0.05, 0.1, 0.2, 0.5):
+            ....:         A = Matrix(QQ, 10, 10, 0)
+            ....:         A.randomize(dens)
+            ....:         ranks[A.rank()] = True
+
+        The default density is `6/9`::
+
+            sage: def add_sample(density, num_rows, num_cols):
+            ....:     global density_sum, total_count
+            ....:     total_count += 1.0
+            ....:     A = Matrix(QQ, num_rows, num_cols, 0)
+            ....:     A.randomize(density)
+            ....:     density_sum += float(A.density())
+
+            sage: density_sum = 0.0
+            sage: total_count = 0.0
+            sage: expected_density = 6/9
+            sage: add_sample(1.0, 100, 100)
+            sage: while abs(density_sum/total_count - expected_density) > 0.001:
+            ....:     add_sample(1.0, 100, 100)
+
+        The modified density depends on the number of columns::
+
+            sage: density_sum = 0.0
+            sage: total_count = 0.0
+            sage: expected_density = 6/9*0.5
+            sage: add_sample(0.5, 100, 2)
+            sage: while abs(density_sum/total_count - expected_density) > 0.001:
+            ....:     add_sample(0.5, 100, 2)
+
+            sage: density_sum = 0.0
+            sage: total_count = 0.0
+            sage: expected_density = 6/9*(1.0 - (99/100)^50)
+            sage: expected_density
+            0.263...
+
+            sage: add_sample(0.5, 100, 100)
+            sage: while abs(density_sum/total_count - expected_density) > 0.001:
+            ....:     add_sample(0.5, 100, 100)
+
+        Modifying the bounds for numerator and denominator::
+
+            sage: num_dic = defaultdict(Integer)
+            sage: den_dic = defaultdict(Integer)
+            sage: while not (all(num_dic[i] for i in range(-200, 201))
+            ....:            and all(den_dic[i] for i in range(1, 101))):
+            ....:     a = matrix(QQ, 2, 4)
+            ....:     a.randomize(num_bound=200, den_bound=100)
+            ....:     for q in a.list():
+            ....:         num_dic[q.numerator()] += 1
+            ....:         den_dic[q.denominator()] += 1
+            sage: len(num_dic)
+            401
+            sage: len(den_dic)
+            100
+
+        TESTS:
+
+        Check that the option ``nonzero`` is meaningful (:issue:`22970`)::
+
+            sage: a = matrix(QQ, 10, 10, 1)
+            sage: b = a.__copy__()
+            sage: b.randomize(nonzero=True)
+            sage: a == b
+            False
+            sage: any(b[i,j].is_zero() for i in range(10) for j in range(10))
+            False
+
+        Check that :issue:`34103` is fixed::
+
+            sage: a = matrix(QQ, 10, 10, 1)
+            sage: a.randomize(nonzero=True, distribution='1/n')
+            sage: bool(a)
+            True
+        """
+        density = float(density)
+        if density <= 0.0:
+            return
+
+        self.check_mutability()
+        self.clear_cache()
+
+        cdef Integer B, C
+        cdef Py_ssize_t i, j, k, num_per_row
+        cdef randstate rstate
+        cdef mpq_t tmp
+
+        B = Integer(num_bound + 1)
+        C = Integer(den_bound + 1)
+
+        mpq_init(tmp)
+
+        if not nonzero:
+            if density >= 1.0:
+                if distribution == "1/n":
+                    sig_on()
+                    for i in range(self._nrows):
+                        for j in range(self._ncols):
+                            mpq_randomize_entry_recip_uniform(tmp)
+                            fmpq_set_mpq(fmpq_mat_entry(self._matrix, i, j), tmp)
+                    sig_off()
+                elif mpz_cmp_si(C.value, 2):   # denom is > 1
+                    sig_on()
+                    for i in range(self._nrows):
+                        for j in range(self._ncols):
+                            mpq_randomize_entry(tmp, B.value, C.value)
+                            fmpq_set_mpq(fmpq_mat_entry(self._matrix, i, j), tmp)
+                    sig_off()
+                else:
+                    sig_on()
+                    for i in range(self._nrows):
+                        for j in range(self._ncols):
+                            mpq_randomize_entry_as_int(tmp, B.value)
+                            fmpq_set_mpq(fmpq_mat_entry(self._matrix, i, j), tmp)
+                    sig_off()
+            else:
+                rstate = current_randstate()
+                num_per_row = int(density * self._ncols)
+                if distribution == "1/n":
+                    sig_on()
+                    for i in range(self._nrows):
+                        for j in range(num_per_row):
+                            k = rstate.c_random() % self._ncols
+                            mpq_randomize_entry_recip_uniform(tmp)
+                            fmpq_set_mpq(fmpq_mat_entry(self._matrix, i, k), tmp)
+                    sig_off()
+                elif mpz_cmp_si(C.value, 2):   # denom is > 1
+                    sig_on()
+                    for i in range(self._nrows):
+                        for j in range(num_per_row):
+                            k = rstate.c_random() % self._ncols
+                            mpq_randomize_entry(tmp, B.value, C.value)
+                            fmpq_set_mpq(fmpq_mat_entry(self._matrix, i, k), tmp)
+                    sig_off()
+                else:
+                    sig_on()
+                    for i in range(self._nrows):
+                        for j in range(num_per_row):
+                            k = rstate.c_random() % self._ncols
+                            mpq_randomize_entry_as_int(tmp, B.value)
+                            fmpq_set_mpq(fmpq_mat_entry(self._matrix, i, k), tmp)
+                    sig_off()
+        else:
+            if density >= 1.0:
+                if distribution == "1/n":
+                    sig_on()
+                    for i in range(self._nrows):
+                        for j in range(self._ncols):
+                            mpq_randomize_entry_recip_uniform_nonzero(tmp)
+                            fmpq_set_mpq(fmpq_mat_entry(self._matrix, i, j), tmp)
+                    sig_off()
+                elif mpz_cmp_si(C.value, 2):   # denom is > 1
+                    sig_on()
+                    for i in range(self._nrows):
+                        for j in range(self._ncols):
+                            mpq_randomize_entry_nonzero(tmp, B.value, C.value)
+                            fmpq_set_mpq(fmpq_mat_entry(self._matrix, i, j), tmp)
+                    sig_off()
+                else:
+                    sig_on()
+                    for i in range(self._nrows):
+                        for j in range(self._ncols):
+                            mpq_randomize_entry_as_int_nonzero(tmp, B.value)
+                            fmpq_set_mpq(fmpq_mat_entry(self._matrix, i, j), tmp)
+                    sig_off()
+            else:
+                rstate = current_randstate()
+                num_per_row = int(density * self._ncols)
+                if distribution == "1/n":
+                    sig_on()
+                    for i in range(self._nrows):
+                        for j in range(num_per_row):
+                            k = rstate.c_random() % self._ncols
+                            mpq_randomize_entry_recip_uniform_nonzero(tmp)
+                            fmpq_set_mpq(fmpq_mat_entry(self._matrix, i, k), tmp)
+                    sig_off()
+                elif mpz_cmp_si(C.value, 2):   # denom is > 1
+                    sig_on()
+                    for i in range(self._nrows):
+                        for j in range(num_per_row):
+                            k = rstate.c_random() % self._ncols
+                            mpq_randomize_entry_nonzero(tmp, B.value, C.value)
+                            fmpq_set_mpq(fmpq_mat_entry(self._matrix, i, k), tmp)
+                    sig_off()
+                else:
+                    sig_on()
+                    for i in range(self._nrows):
+                        for j in range(num_per_row):
+                            k = rstate.c_random() % self._ncols
+                            mpq_randomize_entry_as_int_nonzero(tmp, B.value)
+                            fmpq_set_mpq(fmpq_mat_entry(self._matrix, i, k), tmp)
+                    sig_off()
+
+        mpq_clear(tmp)
+
+    def rank(self, algorithm=None):
+        """
+        Return the rank of this matrix.
+
+        INPUT:
+
+        - ``algorithm`` -- an optional specification of an algorithm. One of
+
+          - ``None`` -- (default) will use flint
+
+          - ``'flint'`` -- uses the flint library
+
+          - ``'pari'`` -- uses the PARI library
+
+          - ``'integer'`` -- eliminate denominators and calls the rank function
+            on the corresponding integer matrix
+
+        EXAMPLES::
+
+            sage: matrix(QQ,3,[1..9]).rank()
+            2
+            sage: matrix(QQ,100,[1..100^2]).rank()
+            2
+
+        TESTS::
+
+            sage: for _ in range(100):                                                  # needs sage.libs.pari
+            ....:     dim = randint(0, 30)
+            ....:     m = random_matrix(QQ, dim, num_bound=2, density=0.5)
+            ....:     r_pari = m.rank('pari'); m._clear_cache()
+            ....:     r_flint = m.rank('flint'); m._clear_cache()
+            ....:     r_int = m.rank('integer'); m._clear_cache()
+            ....:     assert r_pari == r_flint == r_int
+        """
+        r = self.fetch('rank')
+        if r is not None:
+            return r
+
+        if algorithm is None:
+            algorithm = "flint"
+
+        if algorithm == "flint":
+            self.echelon_form(algorithm='flint')
+            return self.fetch('rank')
+        elif algorithm == "pari":
+            from .matrix_rational_pari import _rank_pari
+            r = _rank_pari(self)
+        elif algorithm == "integer":
+            A, _ = self._clear_denom()
+            r = A.rank()
+        else:
+            raise ValueError("unknown algorithm %s" % algorithm)
+
+        self.cache('rank', r)
+        return r
+
+    def transpose(self):
+        """
+        Return the transpose of ``self``, without changing ``self``.
+
+        EXAMPLES:
+
+        We create a matrix, compute its transpose, and note that the
+        original matrix is not changed.
+
+        ::
+
+            sage: A = matrix(QQ, 2, 3, range(6))
+            sage: type(A)
+            <class 'sage.matrix.matrix_rational_dense.Matrix_rational_dense'>
+            sage: B = A.transpose()
+            sage: print(B)
+            [0 3]
+            [1 4]
+            [2 5]
+            sage: print(A)
+            [0 1 2]
+            [3 4 5]
+
+        ``.T`` is a convenient shortcut for the transpose::
+
+            sage: print(A.T)
+            [0 3]
+            [1 4]
+            [2 5]
+
+        ::
+
+            sage: A.subdivide(None, 1); A
+            [0|1 2]
+            [3|4 5]
+            sage: A.transpose()
+            [0 3]
+            [---]
+            [1 4]
+            [2 5]
+        """
+        cdef Matrix_rational_dense ans
+        if self._nrows == self._ncols:
+            parent = self._parent
+        else:
+            parent = self._parent.matrix_space(self._ncols, self._nrows)
+        ans = Matrix_rational_dense.__new__(Matrix_rational_dense, parent, None, None, None)
+        sig_on()
+        fmpq_mat_transpose(ans._matrix, self._matrix)
+        sig_off()
+
+        if self._subdivisions is not None:
+            row_divs, col_divs = self.subdivisions()
+            ans.subdivide(col_divs, row_divs)
+        return ans
+
+    def antitranspose(self):
+        """
+        Return the antitranspose of ``self``, without changing ``self``.
+
+        EXAMPLES::
+
+            sage: A = matrix(QQ,2,3,range(6))
+            sage: type(A)
+            <class 'sage.matrix.matrix_rational_dense.Matrix_rational_dense'>
+            sage: A.antitranspose()
+            [5 2]
+            [4 1]
+            [3 0]
+            sage: A
+            [0 1 2]
+            [3 4 5]
+
+            sage: A.subdivide(1,2); A
+            [0 1|2]
+            [---+-]
+            [3 4|5]
+            sage: A.antitranspose()
+            [5|2]
+            [-+-]
+            [4|1]
+            [3|0]
+        """
+        if self._nrows == self._ncols:
+            parent = self._parent
+        else:
+            parent = self._parent.matrix_space(self._ncols, self._nrows)
+
+        cdef Matrix_rational_dense ans
+        ans = Matrix_rational_dense.__new__(Matrix_rational_dense, parent, None, None, None)
+
+        cdef Py_ssize_t i,j
+        cdef Py_ssize_t ri,rj  # reversed i and j
+        sig_on()
+        ri = self._nrows
+        for i in range(self._nrows):
+            rj = self._ncols
+            ri = ri - 1
+            for j in range(self._ncols):
+                rj = rj - 1
+                fmpq_set(fmpq_mat_entry(ans._matrix, rj, ri),
+                         fmpq_mat_entry(self._matrix, i, j))
+        sig_off()
+
+        if self._subdivisions is not None:
+            row_divs, col_divs = self.subdivisions()
+            ans.subdivide([self._ncols - t for t in reversed(col_divs)],
+                          [self._nrows - t for t in reversed(row_divs)])
+        return ans
+
+    def set_row_to_multiple_of_row(self, Py_ssize_t i, Py_ssize_t j, s):
+        """
+        Set row i equal to s times row j.
+
+        EXAMPLES::
+
+            sage: a = matrix(QQ,2,3,range(6)); a
+            [0 1 2]
+            [3 4 5]
+            sage: a.set_row_to_multiple_of_row(1,0,-3)
+            sage: a
+            [ 0  1  2]
+            [ 0 -3 -6]
+        """
+        self.check_row_bounds_and_mutability(i, j)
+        cdef Py_ssize_t k
+        cdef fmpq_t ss
+        fmpq_init(ss)
+        fmpq_set_mpq(ss, (<Rational> Rational(s)).value)
+        for k in range(self._ncols):
+            fmpq_mul(fmpq_mat_entry(self._matrix, i, k),
+                     fmpq_mat_entry(self._matrix, j, k),
+                     ss)
+        fmpq_clear(ss)
+
+    def _set_row_to_negative_of_row_of_A_using_subset_of_columns(self, Py_ssize_t i, Matrix A,
+                                                                 Py_ssize_t r, cols,
+                                                                 cols_index=None):
+        """
+        Set row i of ``self`` to -(row r of A), but where we only take the
+        given column positions in that row of A. We do not zero out the
+        other entries of ``self``'s row i either.
+
+        .. NOTE::
+
+            This function exists just because it is useful for modular symbols presentations.
+
+        INPUT:
+
+        - ``i`` -- integer, index into the rows of self
+
+        - ``A`` -- a matrix
+
+        - ``r`` -- integer, index into rows of A
+
+        - ``cols`` -- a *sorted* list of integers
+
+        EXAMPLES::
+
+            sage: a = matrix(QQ,2,3,range(6)); a
+            [0 1 2]
+            [3 4 5]
+            sage: a._set_row_to_negative_of_row_of_A_using_subset_of_columns(0,a,1,[1,2])
+            sage: a
+            [-4 -5  2]
+            [ 3  4  5]
+        """
+        self.check_row_bounds_and_mutability(i,i)
+        cdef Matrix_rational_dense _A
+        if r < 0 or r >= A.nrows():
+            raise IndexError("invalid row")
+        cdef Py_ssize_t l = 0
+
+        if not A.base_ring() == QQ:
+            A = A.change_ring(QQ)
+        if not A.is_dense():
+            A = A.dense_matrix()
+
+        _A = A
+        for k in cols:
+            entry = fmpq_mat_entry(self._matrix, i, l)
+            fmpq_set(entry, fmpq_mat_entry(_A._matrix, r, k))
+            fmpq_neg(entry, entry)
+            l += 1
+
+    def _add_col_j_of_A_to_col_i_of_self(self, Py_ssize_t i,
+                                         Matrix_rational_dense A, Py_ssize_t j):
+        """
+        Unsafe technical function that very quickly adds the `j`-th column of
+        A to the `i`-th column of ``self``.
+
+        Does not check mutability.
+        """
+        if A._nrows != self._nrows:
+            raise TypeError("nrows of self and A must be the same")
+        cdef Py_ssize_t r
+        for r in range(self._nrows):
+            fmpq_add(fmpq_mat_entry(self._matrix, r, i),
+                     fmpq_mat_entry(self._matrix, r, i),
+                     fmpq_mat_entry(A._matrix, r, j))
+
+    # ###############################################
+    # Methods using PARI library
+    # ###############################################
+
+    def __pari__(self):
+        """
+        Return pari version of this matrix.
+
+        EXAMPLES::
+
+            sage: matrix(QQ,2,[1/5,-2/3,3/4,4/9]).__pari__()                            # needs sage.libs.pari
+            [1/5, -2/3; 3/4, 4/9]
+        """
+        from .matrix_rational_pari import _pari
+        return _pari(self)
+
+    def _det_pari(self, int flag=0):
+        """
+        Return the determinant of this matrix computed using pari.
+
+        EXAMPLES::
+
+            sage: matrix(QQ,3,[1..9])._det_pari()                                       # needs sage.libs.pari
+            0
+            sage: matrix(QQ,3,[1..9])._det_pari(1)                                      # needs sage.libs.pari
+            0
+            sage: matrix(QQ,3,[0]+[2..9])._det_pari()                                   # needs sage.libs.pari
+            3
+        """
+        from .matrix_rational_pari import _det_pari
+        return _det_pari(self, flag)
+
+    def _rank_pari(self):
+        """
+        Return the rank of this matrix computed using pari.
+
+        EXAMPLES::
+
+            sage: matrix(QQ,3,[1..9])._rank_pari()                                      # needs sage.libs.pari
+            2
+            sage: matrix(QQ, 0, 0)._rank_pari()                                         # needs sage.libs.pari
+            0
+        """
+        from .matrix_rational_pari import _rank_pari
+        return _rank_pari(self)
+
+    def _multiply_pari(self, Matrix_rational_dense right):
+        """
+        Return the product of ``self`` and ``right``, computed using PARI.
+
+        EXAMPLES::
+
+            sage: matrix(QQ,2,[1/5,-2/3,3/4,4/9])._multiply_pari(matrix(QQ,2,[1,2,3,4]))    # needs sage.libs.pari
+            [  -9/5 -34/15]
+            [ 25/12  59/18]
+
+        We verify that 0 rows or columns works::
+
+            sage: x = matrix(QQ,2,0); y = matrix(QQ,0,2); x*y
+            [0 0]
+            [0 0]
+            sage: matrix(ZZ, 0, 0) * matrix(QQ, 0, 5)
+            []
+        """
+        from .matrix_rational_pari import _multiply_pari
+        return _multiply_pari(self, right)
+
+    def _invert_pari(self):
+        """
+        Return the inverse of this matrix computed using PARI.
+
+        EXAMPLES::
+
+            sage: matrix(QQ,2,[1,2,3,4])._invert_pari()                                 # needs sage.libs.pari
+            [  -2    1]
+            [ 3/2 -1/2]
+            sage: matrix(QQ,2,[1,2,2,4])._invert_pari()                                 # needs sage.libs.pari
+            Traceback (most recent call last):
+            ...
+            PariError: impossible inverse in ginv: [1, 2; 2, 4]
+        """
+        from .matrix_rational_pari import _invert_pari
+        return _invert_pari(self)
+
+    def row(self, Py_ssize_t i, from_list=False):
+        """
+        Return the `i`-th row of this matrix as a dense vector.
+
+        INPUT:
+
+        - ``i`` -- integer
+
+        - ``from_list`` -- ignored
+
+        EXAMPLES::
+
+            sage: m = matrix(QQ, 2, [1/5, -2/3, 3/4, 4/9])
+            sage: m.row(0)
+            (1/5, -2/3)
+            sage: m.row(1)
+            (3/4, 4/9)
+            sage: m.row(1, from_list=True)
+            (3/4, 4/9)
+            sage: m.row(-2)
+            (1/5, -2/3)
+
+            sage: m.row(2)
+            Traceback (most recent call last):
+            ...
+            IndexError: row index out of range
+            sage: m.row(-3)
+            Traceback (most recent call last):
+            ...
+            IndexError: row index out of range
+        """
+        if i < 0:
+            i = i + self._nrows
+        if i < 0 or i >= self._nrows:
+            raise IndexError("row index out of range")
+
+        cdef Py_ssize_t j
+        parent = self.row_ambient_module()
+        cdef Vector_rational_dense v = Vector_rational_dense.__new__(Vector_rational_dense)
+        v._init(self._ncols, parent)
+        for j in range(self._ncols):
+            fmpq_get_mpq(v._entries[j], fmpq_mat_entry(self._matrix, i, j))
+        return v
+
+    def column(self, Py_ssize_t i, from_list=False):
+        """
+        Return the `i`-th column of this matrix as a dense vector.
+
+        INPUT:
+
+        - ``i`` -- integer
+
+        - ``from_list`` -- ignored
+
+        EXAMPLES::
+
+            sage: m = matrix(QQ, 3, 2, [1/5,-2/3,3/4,4/9,-1,0])
+            sage: m.column(1)
+            (-2/3, 4/9, 0)
+            sage: m.column(1,from_list=True)
+            (-2/3, 4/9, 0)
+            sage: m.column(-1)
+            (-2/3, 4/9, 0)
+            sage: m.column(-2)
+            (1/5, 3/4, -1)
+
+            sage: m.column(2)
+            Traceback (most recent call last):
+            ...
+            IndexError: column index out of range
+            sage: m.column(-3)
+            Traceback (most recent call last):
+            ...
+            IndexError: column index out of range
+        """
+        if i < 0:
+            i += self._ncols
+        if i < 0 or i >= self._ncols:
+            raise IndexError("column index out of range")
+
+        cdef Py_ssize_t j
+        parent = self.column_ambient_module()
+        cdef Vector_rational_dense v = Vector_rational_dense.__new__(Vector_rational_dense)
+        v._init(self._nrows, parent)
+        for j in range(self._nrows):
+            fmpq_get_mpq(v._entries[j], fmpq_mat_entry(self._matrix, j, i))
+        return v
+
+    # ###############################################
+    # LLL
+    # ###############################################
+
+    def BKZ(self, *args, **kwargs):
+        """
+        Return the result of running Block Korkin-Zolotarev reduction on
+        ``self`` interpreted as a lattice.
+
+        The arguments ``*args`` and ``**kwargs`` are passed onto
+        :meth:`sage.matrix.matrix_integer_dense.Matrix_integer_dense.BKZ`,
+        see there for more details.
+
+        EXAMPLES::
+
+            sage: A = Matrix(QQ, 3, 3, [1/n for n in range(1, 10)])
+            sage: A.BKZ()                                                 # needs fpylll
+            [ 1/28 -1/40 -1/18]
+            [ 1/28 -1/40  1/18]
+            [-1/14 -1/40     0]
+
+            sage: A = random_matrix(QQ, 10, 10)
+            sage: d = lcm(a.denom() for a in A.list())
+            sage: A.BKZ() == (A * d).change_ring(ZZ).BKZ() / d            # needs fpylll
+            True
+        """
+        A, d = self._clear_denom()
+        return A.BKZ(*args, **kwargs) / d
+
+    def LLL(self, *args, **kwargs):
+        """
+        Return an LLL reduced or approximated LLL reduced lattice for
+        ``self`` interpreted as a lattice.
+
+        The arguments ``*args`` and ``**kwargs`` are passed onto
+        :meth:`sage.matrix.matrix_integer_dense.Matrix_integer_dense.LLL`,
+        see there for more details.
+
+        EXAMPLES::
+
+            sage: # needs fpylll
+            sage: A = Matrix(QQ, 3, 3, [1/n for n in range(1, 10)])
+            sage: A.LLL()
+            [ 1/28 -1/40 -1/18]
+            [ 1/28 -1/40  1/18]
+            [    0 -3/40     0]
+            sage: L, U = A.LLL(transformation=True)
+            sage: U * A == L
+            True
+            sage: A = random_matrix(QQ, 10, 10)
+            sage: d = lcm(a.denom() for a in A.list())
+            sage: A.LLL() == (A * d).change_ring(ZZ).LLL() / d
+            True
+        """
+        A, d = self._clear_denom()
+        if kwargs.get('transformation', False):
+            L, U = A.LLL(*args, **kwargs)
+            return L / d, U
+        return A.LLL(*args, **kwargs) / d
+
+    def is_LLL_reduced(self, delta=None, eta=None):
+        r"""
+        Return ``True`` if this lattice is `(\delta, \eta)`-LLL reduced.
+        For a definition of LLL reduction, see
+        :meth:`sage.matrix.matrix_integer_dense.Matrix_integer_dense.LLL`.
+
+        EXAMPLES::
+
+            sage: # needs fpylll
+            sage: A = random_matrix(QQ, 10, 10)
+            sage: L = A.LLL()
+            sage: A.is_LLL_reduced()
+            False
+            sage: L.is_LLL_reduced()
+            True
+        """
+        A, _ = self._clear_denom()
+        return A.is_LLL_reduced(delta, eta)