passagemath-flint 10.6.1rc10__cp310-cp310-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl

sage/modular/pollack_stevens/dist.pyx

@@ -0,0 +1,1439 @@
+# sage_setup: distribution = sagemath-flint
+# distutils: libraries = gmp
+# distutils: extra_compile_args = -D_XPG6
+# sage.doctest: needs sage.schemes
+"""
+`p`-adic distributions spaces
+
+This module implements `p`-adic distributions, a `p`-adic Banach
+space dual to locally analytic functions on a disc.
+
+EXAMPLES::
+
+    sage: D = OverconvergentDistributions(5, 7, 15)
+    sage: v = D([7,14,21,28,35]); v
+    (7 + O(7^5), 2*7 + O(7^4), 3*7 + O(7^3), 4*7 + O(7^2), O(7))
+
+REFERENCES:
+
+- [PS2011]_
+"""
+
+# ****************************************************************************
+#       Copyright (C) 2012 Robert Pollack <rpollack@math.bu.edu>
+#
+#  Distributed under the terms of the GNU General Public License (GPL)
+#  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/
+# ****************************************************************************
+import operator
+
+from sage.arith.misc import bernoulli
+from sage.categories.fields import Fields
+from sage.matrix.constructor import matrix
+from sage.matrix.matrix cimport Matrix
+from sage.misc.verbose import verbose
+from sage.modular.pollack_stevens.sigma0 import Sigma0
+from sage.rings.finite_rings.integer_mod_ring import Zmod
+from sage.rings.infinity import Infinity
+from sage.rings.integer cimport Integer
+from sage.rings.integer_ring import ZZ
+from sage.rings.padics.padic_generic import pAdicGeneric
+from sage.rings.power_series_ring import PowerSeriesRing
+from sage.rings.rational_field import QQ
+from sage.structure.element cimport Element
+from sage.structure.richcmp cimport richcmp_not_equal, rich_to_bool
+
+# from sage.libs.flint.ulong_extras cimport *
+
+cdef long overflow = 1 << (4 * sizeof(long) - 1)
+cdef long underflow = -overflow
+cdef long maxordp = (1L << (sizeof(long) * 8 - 2)) - 1
+
+
+def get_dist_classes(p, prec_cap, base, symk, implementation):
+    r"""
+    Determine the element and action classes to be used for given inputs.
+
+    INPUT:
+
+    - ``p`` -- prime
+
+    - ``prec_cap`` -- the `p`-adic precision cap
+
+    - ``base`` -- the base ring
+
+    - ``symk`` -- an element of Symk
+
+    - ``implementation`` -- string; if not ``None``, override the
+      automatic choice of implementation. May be 'long' or 'vector',
+      otherwise raise a :exc:`NotImplementedError`.
+
+    OUTPUT:
+
+    - Either a Dist_vector and WeightKAction_vector, or a Dist_vector_long
+      and WeightKAction_vector_long
+
+    EXAMPLES::
+
+        sage: D = OverconvergentDistributions(2, 3, 5); D # indirect doctest
+        Space of 3-adic distributions with k=2 action and precision cap 5
+    """
+    if implementation is not None:
+        if implementation == 'long':
+            raise NotImplementedError('The optimized implementation -using longs- has been disabled and may return wrong results.')
+        elif implementation == 'vector':
+            return Dist_vector, WeightKAction_vector
+        else:
+            raise NotImplementedError('The implementation "%s" does not exist yet' % (implementation))
+
+    return Dist_vector, WeightKAction_vector
+
+
+cdef class Dist(ModuleElement):
+    r"""
+        The main `p`-adic distribution class, implemented as per the paper [PS2011]__.
+    """
+    def moment(self, n):
+        r"""
+        Return the `n`-th moment.
+
+        INPUT:
+
+        - ``n`` -- integer or slice, to be passed on to moments
+
+        OUTPUT:
+
+        - the `n`-th moment, or a list of moments in the case that `n`
+          is a slice.
+
+        EXAMPLES::
+
+            sage: D = OverconvergentDistributions(4, 7, 10)
+            sage: v = D([7,14,21,28,35])
+            sage: v.moment(3)
+            4*7 + O(7^2)
+            sage: v.moment(0)
+            7 + O(7^5)
+        """
+        return self.parent().prime() ** (self.ordp) * self._unscaled_moment(n)
+
+    def moments(self):
+        r"""
+        Return the vector of moments.
+
+        OUTPUT: the vector of moments
+
+        EXAMPLES::
+
+            sage: D = OverconvergentDistributions(4, 5, 10, base = Qp(5))
+            sage: v = D([1,7,4,2,-1])
+            sage: v = 1/5^3 * v
+            sage: v
+            5^-3 * (1 + O(5^5), 2 + 5 + O(5^4), 4 + O(5^3), 2 + O(5^2), 4 + O(5))
+            sage: v.moments()
+            (5^-3 + O(5^2), 2*5^-3 + 5^-2 + O(5), 4*5^-3 + O(5^0), 2*5^-3 + O(5^-1), 4*5^-3 + O(5^-2))
+        """
+        return self.parent().prime() ** (self.ordp) * self._moments
+
+    cpdef normalize(self, include_zeroth_moment=True):
+        r"""
+        Normalize so that the precision of the `i`-th moment is `n-i`,
+        where `n` is the number of moments.
+
+        OUTPUT: normalized entries of the distribution
+
+        EXAMPLES::
+
+            sage: D = OverconvergentDistributions(5, 7, 15); D
+            Space of 7-adic distributions with k=5 action and precision cap 15
+            sage: v = D([1,2,3,4,5]); v
+            (1 + O(7^5), 2 + O(7^4), 3 + O(7^3), 4 + O(7^2), 5 + O(7))
+            sage: v.normalize()
+            (1 + O(7^5), 2 + O(7^4), 3 + O(7^3), 4 + O(7^2), 5 + O(7))
+        """
+        raise NotImplementedError
+
+    cdef long _relprec(self) noexcept:
+        raise NotImplementedError
+
+    cdef _unscaled_moment(self, long i):
+        raise NotImplementedError
+
+    cpdef long _ord_p(self) noexcept:
+        r"""
+        Return power of `p` by which the moments are shifted.
+
+        .. NOTE::
+
+            This is not necessarily the same as the valuation,
+            since the moments could all be divisible by `p`.
+
+        EXAMPLES::
+
+            sage: D = OverconvergentDistributions(5, 7, 15)
+            sage: v = D([7,14,21,28,35]); v
+            (7 + O(7^5), 2*7 + O(7^4), 3*7 + O(7^3), 4*7 + O(7^2), O(7))
+            sage: v._ord_p()
+            0
+        """
+        return self.ordp
+
+    def scale(self, left):
+        r"""
+        Scale the moments of the distribution by ``left``.
+
+        INPUT:
+
+        - ``left`` -- scalar
+
+        OUTPUT: scales the moments by ``left``
+
+        EXAMPLES::
+
+            sage: D = OverconvergentDistributions(5, 7, 15)
+            sage: v = D([1,2,3,4,5]); v
+            (1 + O(7^5), 2 + O(7^4), 3 + O(7^3), 4 + O(7^2), 5 + O(7))
+            sage: v.scale(2)
+            (2 + O(7^5), 4 + O(7^4), 6 + O(7^3), 1 + 7 + O(7^2), 3 + O(7))
+        """
+        # if isinstance(self, Dist_long) and isinstance(left, (Integer, pAdicCappedRelativeElement, pAdicCappedAbsoluteElement, pAdicFixedModElement)):
+        #     return self._lmul_(left)
+        R = left.parent()
+        base = self.parent().base_ring()
+        if base is R:
+            return self._lmul_(left)
+        elif base.has_coerce_map_from(R):
+            return self._lmul_(base(left))
+        else:
+            from sage.categories.pushout import pushout
+            new_base = pushout(base, R)
+            V = self.parent().change_ring(new_base)
+            scalar = new_base(left)
+            return V([scalar * new_base(self.moment(i)) for i in range(self.precision_absolute())])
+
+    def is_zero(self, p=None, M=None) -> bool:
+        r"""
+        Return ``True`` if the `i`-th moment is zero for all `i` (case ``M`` is
+        ``None``) or zero modulo `p^{M-i}` for all `i` (when ``M`` is not
+        ``None``).
+
+        Note that some moments are not known to precision ``M``, in which
+        case they are only checked to be equal to zero modulo the
+        precision to which they are defined.
+
+        INPUT:
+
+        - ``p`` -- prime
+
+        - ``M`` -- precision
+
+        OUTPUT: boolean
+
+        EXAMPLES::
+
+            sage: D = OverconvergentDistributions(5, 7, 15)
+            sage: v = D([1,2,3,4,5]); v
+            (1 + O(7^5), 2 + O(7^4), 3 + O(7^3), 4 + O(7^2), 5 + O(7))
+            sage: v.is_zero()
+            False
+            sage: v = D(5*[0])
+            sage: v.is_zero()
+            True
+
+        ::
+
+            sage: D = Symk(0)
+            sage: v = D([0])
+            sage: v.is_zero(5,3)
+            True
+        """
+        n = self.precision_relative()
+        aprec = self.precision_absolute()
+        if M is None:
+            M = n
+        # elif M > aprec: # DEBUG
+        #     return False
+        elif M < aprec:
+            n -= (aprec - M)
+            M -= self.ordp
+        if p is None:
+            p = self.parent().prime()
+        cdef bint usearg = True
+        if n == 0:
+            return True
+        else:
+            try:
+                z = self._unscaled_moment(0).is_zero(M)
+            except TypeError:
+                z = self._unscaled_moment(0).is_zero()
+                use_arg = False
+            if not z:
+                return False
+            for a in range(1, n):
+                if usearg:
+                    try:
+                        z = self._unscaled_moment(a).is_zero(M - a)
+                    except TypeError:
+                        z = self._unscaled_moment(a).is_zero()
+                        use_arg = False
+                else:
+                    z = self._unscaled_moment(a).is_zero()
+                if not z:
+                    return False
+            return True
+
+    def find_scalar(self, _other, p, M=None, check=True):
+        r"""
+        Return an ``alpha`` with ``other = self * alpha``, or raises
+        a :exc:`ValueError`.
+
+        It will also raise a :exc:`ValueError` if this distribution is zero.
+
+        INPUT:
+
+        - ``other`` -- another distribution
+
+        - ``p`` -- an integral prime (only used if the parent is not a Symk)
+
+        - ``M`` -- (default: ``None``) an integer, the relative precision
+          to which the scalar must be determined
+
+        - ``check`` -- boolean (default: ``True``); whether to validate
+          that ``other`` is actually a multiple of this element
+
+        OUTPUT: a scalar ``alpha`` with ``other = self * alpha``
+
+        EXAMPLES::
+
+            sage: D = OverconvergentDistributions(5, 7, 15)
+            sage: v = D([1,2,3,4,5])
+            sage: w = D([3,6,9,12,15])
+            sage: v.find_scalar(w,p=7)
+            3 + O(7^5)
+            sage: v.find_scalar(w,p=7,M=4)
+            3 + O(7^4)
+
+            sage: u = D([1,4,9,16,25])
+            sage: v.find_scalar(u,p=7)
+            Traceback (most recent call last):
+            ...
+            ValueError: not a scalar multiple
+        """
+        cdef Dist other = _other
+        i = 0
+        n = self.precision_relative()
+        other_pr = other.precision_relative()
+        if n == 0:
+            raise ValueError("self is zero")
+        verbose("n = %s" % n, level=2)
+        verbose("moment 0", level=2)
+        a = self._unscaled_moment(i)
+        verbose("a = %s" % a, level = 2)
+        padic = isinstance(a.parent(), pAdicGeneric)
+        if self.parent().is_symk():
+            while a == 0:
+                if other._unscaled_moment(i) != 0:
+                    raise ValueError("not a scalar multiple")
+                i += 1
+                verbose("moment %s" % i, level = 2)
+                try:
+                    a = self._unscaled_moment(i)
+                except IndexError:
+                    raise ValueError("self is zero")
+            alpha = other._unscaled_moment(i) / a
+            if check:
+                i += 1
+                while i < n:
+                    verbose("comparing moment %s" % i, level = 2)
+                    if alpha * self._unscaled_moment(i) != other._unscaled_moment(i):
+                        raise ValueError("not a scalar multiple")
+                    i += 1
+        else:
+            p = self.parent().prime()
+            v = a.valuation(p)
+            while v >= n - i:
+                i += 1
+                verbose("p moment %s" % i, level = 2)
+                try:
+                    a = self._unscaled_moment(i)
+                except IndexError:
+                    raise ValueError("self is zero")
+                v = a.valuation(p)
+            relprec = n - i - v
+            #            verbose("p=%s, n-i=%s\nself.moment=%s, other.moment=%s" % (p, n-i, a, other._unscaled_moment(i)),level=2)
+            # RP: This code was crashing because other may have too few moments -- so I added this bound with other's relative precision
+            if padic:
+                if i < other_pr:
+                    alpha = (other._unscaled_moment(i) / a).add_bigoh(n - i)
+                else:
+                    alpha = (0 * a).add_bigoh(other_pr - i)
+            else:
+                if i < other_pr:
+                    alpha = (other._unscaled_moment(i) / a) % p ** (n - i)
+                else:
+                    alpha = 0
+            verbose("alpha = %s" % alpha, level = 2)
+            # RP: This code was crashing because other may have too few moments -- so I added this bound with other's relative precision
+            while i < other_pr - 1:
+                i += 1
+                verbose("comparing p moment %s" % i, level = 2)
+                a = self._unscaled_moment(i)
+                if check:
+                    # verbose("self.moment=%s, other.moment=%s" % (a, other._unscaled_moment(i)))
+                    if (padic and other._unscaled_moment(i) != alpha * a) or \
+                       (not padic and other._unscaled_moment(i) % p ** (n - i) != alpha * a % p ** (n - i)):
+                        raise ValueError("not a scalar multiple")
+                v = a.valuation(p)
+                if n - i - v > relprec:
+                    verbose("Resetting alpha: relprec=%s, n-i=%s, v=%s" % (relprec, n - i, v), level = 2)
+                    relprec = n - i - v
+                    if padic:
+                        alpha = (other._unscaled_moment(i) / a).add_bigoh(n - i)
+                    else:
+                        alpha = (other._unscaled_moment(i) / a) % p ** (n - i)
+                    verbose("alpha=%s" % alpha, level = 2)
+            if relprec < M:
+                raise ValueError("result not determined to high enough precision")
+        alpha = alpha * self.parent().prime() ** (other.ordp - self.ordp)
+        verbose("alpha=%s" % alpha, level = 2)
+        try:
+            alpha = self.parent().base_ring()(alpha)
+            if M is not None:
+                alpha = alpha.add_bigoh(M)
+        except (ValueError, AttributeError):
+            pass
+        return alpha
+
+    def find_scalar_from_zeroth_moment(self, _other, p, M=None, check=True):
+        r"""
+        Return an ``alpha`` with ``other = self * alpha`` using only
+        the zeroth moment, or raises a :exc:`ValueError`.
+
+        It will also raise a :exc:`ValueError` if the zeroth moment of the
+        distribution is zero.
+
+        INPUT:
+
+        - ``other`` -- another distribution
+
+        - ``p`` -- an integral prime (only used if the parent is not a Symk)
+
+        - ``M`` -- (default: ``None``) an integer, the relative precision
+          to which the scalar must be determined
+
+        - ``check`` -- boolean (default: ``True``); whether to validate
+          that ``other`` is actually a multiple of this element
+
+        OUTPUT: a scalar ``alpha`` with ``other = self * alpha``
+
+        EXAMPLES::
+
+            sage: D = OverconvergentDistributions(5, 7, 15)
+            sage: v = D([1,2,3,4,5])
+            sage: w = D([3,6,9,12,15])
+            sage: v.find_scalar_from_zeroth_moment(w,p=7)
+            3 + O(7^5)
+            sage: v.find_scalar_from_zeroth_moment(w,p=7,M=4)
+            3 + O(7^4)
+
+            sage: u = D([1,4,9,16,25])
+            sage: v.find_scalar_from_zeroth_moment(u,p=7)
+            Traceback (most recent call last):
+            ...
+            ValueError: not a scalar multiple
+        """
+        cdef Dist other = _other
+        n = self.precision_relative()
+        if n == 0:
+            raise ValueError("zeroth moment is zero")
+        verbose("n = %s" % n, level = 2)
+        a = self.moment(0)
+        if a.is_zero():
+            raise ValueError("zeroth moment is zero")
+        alpha = other.moment(0) / a
+        if check:
+            for i in range(1, n):
+                verbose("comparing moment %s" % i, level = 2)
+                if alpha * self.moment(i) != other.moment(i):
+                    raise ValueError("not a scalar multiple")
+        alpha = self.parent().base_ring()(alpha)
+        if M is not None:
+            try:
+                absprec = alpha.precision_absolute()
+                if absprec < M:
+                    raise ValueError("result not determined to high "
+                                     "enough precision")
+                verbose("alpha=%s" % (alpha), level = 2)
+                alpha = alpha.add_bigoh(M)
+            except AttributeError:
+                pass
+        return alpha
+
+    cpdef _richcmp_(_left, _right, int op):
+        r"""
+        Comparison.
+
+        EXAMPLES:
+
+        Equality of two distributions::
+
+            sage: D = OverconvergentDistributions(0, 5, 10)
+            sage: D([1, 2]) == D([1])
+            True
+            sage: D([1]) == D([1, 2])
+            True
+            sage: v = D([1+O(5^3),2+O(5^2),3+O(5)])
+            sage: w = D([1+O(5^2),2+O(5)])
+            sage: v == w
+            True
+            sage: D = Symk(0,Qp(5,5))
+            sage: v = 5 * D([4*5^-1+3+O(5^2)])
+            sage: w = D([4+3*5+O(5^2)])
+            sage: v == w
+            True
+        """
+        cdef Dist left = _left
+        cdef Dist right = _right
+        left.normalize()
+        right.normalize()
+        cdef long rprec = min(left._relprec(), right._relprec())
+        cdef long i
+        p = left.parent().prime()
+        if left.ordp > right.ordp:
+            shift = p ** (left.ordp - right.ordp)
+            for i in range(rprec):
+                lx = shift * left._unscaled_moment(i)
+                rx = right._unscaled_moment(i)
+                if lx != rx:
+                    return richcmp_not_equal(lx, rx, op)
+        elif left.ordp < right.ordp:
+            shift = p ** (right.ordp - left.ordp)
+            for i in range(rprec):
+                lx = left._unscaled_moment(i)
+                rx = shift * right._unscaled_moment(i)
+                if lx != rx:
+                    return richcmp_not_equal(lx, rx, op)
+        else:
+            for i in range(rprec):
+                lx = left.moment(i)
+                rx = right.moment(i)
+                if lx != rx:
+                    return richcmp_not_equal(lx, rx, op)
+        return rich_to_bool(op, 0)
+
+    def diagonal_valuation(self, p=None):
+        """
+        Return the largest `m` so that this distribution lies in `Fil^m`.
+
+        INPUT:
+
+        - ``p`` -- (default: ``None``) a positive integral prime
+
+        OUTPUT:
+
+        The largest integer `m` so that `p^m` divides the `0`-th
+        moment, `p^{m-1}` divides the first moment, etc.
+
+        EXAMPLES::
+
+            sage: D = OverconvergentDistributions(8, 7, 15)
+            sage: v = D([7^(5-i) for i in range(1,5)])
+            sage: v
+            (O(7^4), O(7^3), O(7^2), O(7))
+            sage: v.diagonal_valuation(7)
+            4
+        """
+        if p is None:
+            p = self.parent()._p
+        n = self.precision_relative()
+        return self.ordp + min([n] + [a + self._unscaled_moment(a).valuation(p) for a in range(n)])
+
+    def valuation(self, p=None):
+        """
+        Return the minimum valuation of any moment.
+
+        INPUT:
+
+        - ``p`` -- (default: ``None``) a positive integral prime
+
+        OUTPUT: integer
+
+        .. WARNING::
+
+            Since only finitely many moments are computed, this valuation may
+            be larger than the actual valuation of this distribution.
+            Moreover, this valuation may be smaller than the actual
+            valuation if all entries are zero to the known precision.
+
+        EXAMPLES::
+
+            sage: D = OverconvergentDistributions(8, 7, 15)
+            sage: v = D([7^(5-i) for i in range(1,5)])
+            sage: v
+            (O(7^4), O(7^3), O(7^2), O(7))
+            sage: v.valuation(7)
+            4
+        """
+        if p is None:
+            p = self.parent()._p
+        n = self.precision_relative()
+        if self.parent().is_symk():
+            return self.ordp + min([self._unscaled_moment(a).valuation(p) for a in range(n)])
+        else:
+            return self.ordp + min([n] + [self._unscaled_moment(a).valuation(p) for a in range(n) if not self._unscaled_moment(a).is_zero()])
+
+    def specialize(self, new_base_ring=None):
+        """
+        Return the image of this overconvergent distribution under
+        the canonical projection from distributions of weight `k` to
+        `Sym^k`.
+
+        INPUT:
+
+        - ``new_base_ring`` -- (default: ``None``) a ring giving the
+          desired base ring of the result
+
+        OUTPUT:
+
+        An element of `Sym^k(K)`, where `K` is the specified base ring.
+
+        EXAMPLES::
+
+            sage: D = OverconvergentDistributions(4, 13)
+            sage: d = D([0,2,4,6,8,10,12])
+            sage: d.specialize()
+            (O(13^7), 2 + O(13^6), 4 + O(13^5), 6 + O(13^4), 8 + O(13^3))
+        """
+        # self.normalize() # This method should not change self
+        k = self.parent()._k
+        if k < 0:
+            raise ValueError("negative weight")
+        if self.precision_absolute() < k + 1:
+            raise ValueError("not enough moments")
+        V = self.parent().specialize(new_base_ring)
+        new_base_ring = V.base_ring()
+        if self.precision_relative() == 0:
+            return V.zero()
+        return V([new_base_ring.coerce(self.moment(j)) for j in range(k + 1)])
+
+    def lift(self, p=None, M=None, new_base_ring=None):
+        r"""
+        Lift a distribution or element of `Sym^k` to an overconvergent distribution.
+
+        INPUT:
+
+        - ``p`` -- (default: ``None``) a positive integral prime.  If ``None``
+          then ``p`` must be available in the parent
+
+        - ``M`` -- (default: ``None``) a positive integer giving the
+          desired number of moments. If ``None``, returns a distribution having one
+          more moment than this one.
+
+        - ``new_base_ring`` -- (default: ``None``) a ring giving the desired base
+          ring of the result. If ``None``, a base ring is chosen automatically.
+
+        OUTPUT:
+
+        - An overconvergent distribution with `M` moments whose image
+          under the specialization map is this element.
+
+        EXAMPLES::
+
+            sage: V = Symk(0)
+            sage: x = V(1/4)
+            sage: y = x.lift(17, 5)
+            sage: y
+            (13 + 12*17 + 12*17^2 + 12*17^3 + 12*17^4 + O(17^5), O(17^4), O(17^3), O(17^2), O(17))
+            sage: y.specialize()._moments == x._moments
+            True
+        """
+        V = self.parent().lift(p, M, new_base_ring)
+        k = V._k
+        p = V.prime()
+        M = V.precision_cap()
+        R = V.base_ring()
+        moments = [R(self.moment(j)) for j in range(k + 1)]
+        zero = R(0)
+        moments.extend([zero] * (M - k - 1))
+        mu = V(moments)
+        # val = mu.valuation()
+        # if val < 0:
+        #     # This seems unnatural
+        #     print("scaling by ", p, "^", -val, " to keep things integral")
+        #     mu *= p**(-val)
+        return mu
+
+    def _is_malformed(self):
+        r"""
+        Check that the precision of ``self`` is sensible.
+
+        EXAMPLES::
+
+            sage: D = sage.modular.pollack_stevens.distributions.Symk(2, base=Qp(5))
+            sage: v = D([1, 2, 3])
+            sage: v._is_malformed()
+            False
+            sage: v = D([1 + O(5), 2, 3])
+            sage: v._is_malformed()
+            True
+        """
+        n = self.precision_absolute()
+        for i in range(n):
+            if self.moment(i).precision_absolute() < n - i:
+                return True
+        return False
+
+    def act_right(self, gamma):
+        r"""
+        The image of this element under the right action by a
+        `2 \times 2` matrix.
+
+        INPUT:
+
+        - ``gamma`` -- the matrix by which to act
+
+        OUTPUT:
+
+        - ``self | gamma``
+
+        .. NOTE::
+
+            You may also just use multiplication ``self * gamma``.
+
+        EXAMPLES::
+
+            sage: D = OverconvergentDistributions(4, 7, 10)
+            sage: v = D([98,49,21,28,35])
+            sage: M = matrix([[1,0], [7,1]])
+            sage: v.act_right(M)
+            (2*7^2 + 7^3 + 5*7^4 + O(7^5), 3*7^2 + 6*7^3 + O(7^4), 3*7 + 7^2 + O(7^3), 4*7 + O(7^2), O(7))
+        """
+        return self.parent()._act(self, gamma)
+
+cdef class Dist_vector(Dist):
+    r"""
+    A distribution is stored as a vector whose `j`-th entry is the `j`-th moment of the distribution.
+
+    The `j`-th entry is stored modulo `p^{N-j}` where `N` is the total number of moments.
+    (This is the accuracy that is maintained after acting by `\Gamma_0(p)`.)
+
+    INPUT:
+
+    - ``moments`` -- the list of moments.  If ``check == False`` it
+      must be a vector in the appropriate approximation module.
+
+    - ``parent`` -- a :class:`distributions.OverconvergentDistributions_class` or
+      :class:`distributions.Symk_class` instance
+
+    - ``ordp`` -- integer;  this *must* be zero in the case of Symk
+      of an exact ring
+
+    - ``check`` -- boolean (default: ``True``); whether to validate input
+
+    EXAMPLES::
+
+        sage: D = OverconvergentDistributions(3,5,6) # indirect doctest
+        sage: v = D([1,1,1])
+    """
+    def __init__(self, moments, parent, ordp=0, check=True, normalize=True):
+        """
+        Initialization.
+
+        TESTS::
+
+            sage: Symk(4)(0)
+            (0, 0, 0, 0, 0)
+        """
+        # if not hasattr(parent,'Element'):
+        #     parent, moments = moments, parent
+
+        Dist.__init__(self, parent)
+        if check:
+            # case 1: input is a distribution already
+            if isinstance(moments, Dist):
+                ordp = moments._ord_p()
+                moments = moments._moments.change_ring(parent.base_ring())
+            # case 2: input is a vector, or something with a len
+            elif hasattr(moments, '__len__'):
+                M = len(moments)
+                moments = parent.approx_module(M)(moments)
+            # case 3: input is zero
+            elif moments == 0:
+                moments = parent.approx_module(parent.precision_cap())(moments)
+            # case 4: everything else
+            else:
+                moments = parent.approx_module(1)([moments])
+            # TODO: This is not quite right if the input is an inexact zero.
+            if ordp != 0 and parent.prime() == 0:
+                raise ValueError("cannot specify a valuation shift for an exact ring")
+
+        self._moments = moments
+        self.ordp = ordp
+        if normalize:
+            self.normalize()
+
+    def __reduce__(self):
+        r"""
+        Used for pickling.
+
+        EXAMPLES::
+
+            sage: D = sage.modular.pollack_stevens.distributions.Symk(2)
+            sage: x = D([2,3,4])
+            sage: x.__reduce__()
+            (<class 'sage.modular.pollack_stevens.dist.Dist_vector'>, ((2, 3, 4), Sym^2 Q^2, 0, False))
+        """
+        return (self.__class__, (self._moments, self.parent(), self.ordp, False))
+
+    cdef Dist_vector _new_c(self):
+        r"""
+        Create an empty distribution.
+
+        Note that you MUST fill in the ordp attribute on the resulting distribution.
+
+        OUTPUT:
+
+        - A distribution with no moments.  The moments are then filled
+          in by the calling function.
+
+        EXAMPLES::
+
+            sage: D = OverconvergentDistributions(3,5,4) # indirect doctest
+            sage: v = D([1,1,1])
+        """
+        cdef Dist_vector ans = Dist_vector.__new__(Dist_vector)
+        ans._parent = self._parent
+        return ans
+
+    def _repr_(self):
+        r"""
+        String representation.
+
+        EXAMPLES::
+
+            sage: D = OverconvergentDistributions(5, 7, 15)
+            sage: v = D([1,2,3,4,5]); v
+            (1 + O(7^5), 2 + O(7^4), 3 + O(7^3), 4 + O(7^2), 5 + O(7))
+            sage: repr(v)
+            '(1 + O(7^5), 2 + O(7^4), 3 + O(7^3), 4 + O(7^2), 5 + O(7))'
+        """
+        valstr = ""
+        if self.ordp == 1:
+            valstr = "%s * " % (self.parent().prime())
+        elif self.ordp != 0:
+            valstr = "%s^%s * " % (self.parent().prime(), self.ordp)
+        if len(self._moments) == 1:
+            return valstr + repr(self._moments[0])
+        else:
+            return valstr + repr(self._moments)
+
+    def _rational_(self):
+        """
+        Convert to a rational number.
+
+        EXAMPLES::
+
+            sage: D = Symk(0); d = D(4/3); d
+            4/3
+            sage: QQ(d)
+            4/3
+
+        We get a :exc:`TypeError` if there is more than 1 moment::
+
+            sage: D = Symk(1); d = D([1,2]); d
+            (1, 2)
+            sage: QQ(d)
+            Traceback (most recent call last):
+            ...
+            TypeError: k must be 0
+        """
+        if len(self._moments) == 1:
+            return QQ(self.moment(0))
+        raise TypeError("k must be 0")
+
+    cdef long _relprec(self) noexcept:
+        """
+        Return the number of moments.
+
+        EXAMPLES::
+
+            sage: D = Symk(4)
+            sage: d = D([1,2,3,4,5]); e = D([2,3,4,5,6])
+            sage: d == e # indirect doctest
+            False
+        """
+        return len(self._moments)
+
+    cdef _unscaled_moment(self, long n):
+        r"""
+        Return the `n`-th moment, unscaled by the overall power of `p`
+        stored in ``self.ordp``.
+
+        EXAMPLES::
+
+            sage: D = OverconvergentDistributions(4,3,5)
+            sage: d = D([3,3,3,3,3])
+            sage: d.moment(2) # indirect doctest
+            3 + O(3^3)
+        """
+        return self._moments[n]
+
+    cdef Dist_vector _addsub(self, Dist_vector right, bint negate):
+        r"""
+        Common code for the sum and the difference of two distributions.
+
+        EXAMPLES::
+
+            sage: D = Symk(2)
+            sage: u = D([1,2,3]); v = D([4,5,6])
+            sage: u + v # indirect doctest
+            (5, 7, 9)
+            sage: u - v # indirect doctest
+            (-3, -3, -3)
+        """
+        cdef Dist_vector ans = self._new_c()
+        cdef long aprec = min(self.ordp + len(self._moments), right.ordp + len(right._moments))
+        ans.ordp = min(self.ordp, right.ordp)
+        cdef long rprec = aprec - ans.ordp
+        # In the case of symk, rprec will always be k
+        V = ans.parent().approx_module(rprec)
+        R = V.base_ring()
+        smoments = self._moments
+        rmoments = right._moments
+        # We truncate if the moments are too long; extend by zero if too short
+        if smoments.parent() is not V:
+            vec = smoments.list(copy=False)[:rprec] + ([R(0)] * (rprec - len(smoments)) if rprec > len(smoments) else [])
+            smoments = V(vec)
+        if rmoments.parent() is not V:
+            vec = rmoments.list(copy=False)[:rprec] + ([R(0)] * (rprec - len(rmoments)) if rprec > len(rmoments) else [])
+            rmoments = V(vec)
+        # We multiply by the relative power of p
+        if self.ordp > right.ordp:
+            smoments *= self.parent().prime() ** (self.ordp - right.ordp)
+        elif self.ordp < right.ordp:
+            rmoments *= self.parent().prime() ** (right.ordp - self.ordp)
+        if negate:
+            rmoments = -rmoments
+        ans._moments = smoments + rmoments
+        return ans
+
+    cpdef _add_(self, _right):
+        r"""
+        Sum of two distributions.
+
+        EXAMPLES::
+
+            sage: D = OverconvergentDistributions(5, 7, 15)
+            sage: v = D([1,2,3,4,5]); w = D([3,6,9,12,15])
+            sage: v+w
+            (4 + O(7^5), 1 + 7 + O(7^4), 5 + 7 + O(7^3), 2 + 2*7 + O(7^2), 6 + O(7))
+        """
+        return self._addsub(<Dist_vector>_right, False)
+
+    cpdef _sub_(self, _right):
+        r"""
+        Difference of two distributions.
+
+        EXAMPLES::
+
+            sage: D = OverconvergentDistributions(5, 7, 15)
+            sage: v = D([1,2,3,4,5]); w = D([1,1,1,8,8])
+            sage: v-w
+            (O(7^5), 1 + O(7^4), 2 + O(7^3), 3 + 6*7 + O(7^2), 4 + O(7))
+        """
+        return self._addsub(<Dist_vector>_right, True)
+
+    cpdef _lmul_(self, Element right):
+        r"""
+        Scalar product of a distribution with a ring element that coerces into the base ring.
+
+        EXAMPLES::
+
+            sage: D = OverconvergentDistributions(5, 7, 15)
+            sage: v = D([1,2,3,4,5]); v
+            (1 + O(7^5), 2 + O(7^4), 3 + O(7^3), 4 + O(7^2), 5 + O(7))
+            sage: 3*v; 7*v
+            (3 + O(7^5), 6 + O(7^4), 2 + 7 + O(7^3), 5 + 7 + O(7^2), 1 + O(7))
+            7 * (1 + O(7^5), 2 + O(7^4), 3 + O(7^3), 4 + O(7^2), 5 + O(7))
+            sage: v*3; v*7
+            (3 + O(7^5), 6 + O(7^4), 2 + 7 + O(7^3), 5 + 7 + O(7^2), 1 + O(7))
+            7 * (1 + O(7^5), 2 + O(7^4), 3 + O(7^3), 4 + O(7^2), 5 + O(7))
+        """
+        cdef Dist_vector ans = self._new_c()
+        p = self.parent().prime()
+        if p == 0:
+            ans._moments = self._moments * right
+            ans.ordp = self.ordp
+        elif right.valuation(p) == Infinity:
+            ans._moments = self.parent().approx_module(0)([])
+            ans.ordp += self.precision_relative()
+        else:
+            try:
+                v, u = right.val_unit(p)
+            except TypeError:  # bug in p-adics: they should accept p here
+                v, u = right.val_unit()
+            ans._moments = self._moments * u
+            ans.ordp = self.ordp + v
+            # if the relative precision of u is less than that of self, ans may not be normalized.
+        return ans
+
+    def precision_relative(self):
+        r"""
+        Return the relative precision of this distribution.
+
+        The precision is just the number of moments stored, which is
+        also `k+1` in the case of `Sym^k(R)`.  For overconvergent
+        distributions, the precision is the integer `m` so that the
+        sequence of moments is known modulo `Fil^m`.
+
+        OUTPUT: integer giving the number of moments
+
+        EXAMPLES::
+
+            sage: D = OverconvergentDistributions(2, 11, 15)
+            sage: v = D([1,1,10,9,6,15])
+            sage: v.precision_relative()
+            6
+            sage: v = v.reduce_precision(4); v.precision_relative()
+            4
+            sage: D = Symk(10)
+            sage: v = D.random_element()
+            sage: v.precision_relative()
+            11
+        """
+        return Integer(len(self._moments))
+
+    def precision_absolute(self):
+        r"""
+        Return the absolute precision of this distribution.
+
+        The absolute precision is the sum of the relative precision
+        (number of moments) and the valuation.
+
+        EXAMPLES::
+
+            sage: D = OverconvergentDistributions(3, 7, base = Qp(7))
+            sage: v = D([3,1,10,0])
+            sage: v.precision_absolute()
+            4
+            sage: v *= 7
+            sage: v.precision_absolute()
+            5
+            sage: v = 1/7^10 * v
+            sage: v.precision_absolute()
+            -5
+        """
+        return Integer(len(self._moments) + self.ordp)
+
+    cpdef normalize(self, include_zeroth_moment=True):
+        r"""
+        Normalize by reducing modulo `Fil^N`, where `N` is the number of moments.
+
+        If the parent is Symk, then normalize has no effect.  If the
+        parent is a space of distributions, then normalize reduces the
+        `i`-th moment modulo `p^{N-i}`.
+
+        OUTPUT: this distribution, after normalizing
+
+        .. WARNING::
+
+            This function modifies the distribution in place as well as returning it.
+
+        EXAMPLES::
+
+            sage: D = OverconvergentDistributions(3,7,10)
+            sage: v = D([1,2,3,4,5]) ; v
+            (1 + O(7^5), 2 + O(7^4), 3 + O(7^3), 4 + O(7^2), 5 + O(7))
+            sage: w = v.reduce_precision(3) ; w
+            (1 + O(7^5), 2 + O(7^4), 3 + O(7^3))
+            sage: w.normalize()
+            (1 + O(7^3), 2 + O(7^2), 3 + O(7))
+            sage: w
+            (1 + O(7^3), 2 + O(7^2), 3 + O(7))
+            sage: v.reduce_precision(3).normalize(include_zeroth_moment=False)
+            (1 + O(7^5), 2 + O(7^2), 3 + O(7))
+        """
+        if not self.parent().is_symk() and self._moments != 0:  # non-classical
+            if not self._moments:
+                return self
+            V = self._moments.parent()
+            R = V.base_ring()
+            n = self.precision_relative()
+            p = self.parent()._p
+            shift = self.ordp
+            if include_zeroth_moment:
+                if isinstance(R, pAdicGeneric):
+                    self._moments = V([self._moments[i].add_bigoh(n -shift - i) for i in range(n)])
+                else:
+                    self._moments = V([self._moments[i] % (p ** (n -shift - i)) for i in range(n)])
+            else:
+                if isinstance(R, pAdicGeneric):
+                    self._moments = V([self._moments[0]] + [self._moments[i].add_bigoh(n -shift - i) for i in range(1, n)])  # Don't normalize the zeroth moment
+                else:
+                    self._moments = V([self._moments[0]] + [self._moments[i] % (p ** (n -shift- i)) for i in range(1, n)])  # Don't normalize the zeroth moment
+        return self
+
+    def reduce_precision(self, M):
+        r"""
+        Only hold on to `M` moments.
+
+        INPUT:
+
+        - ``M`` -- positive integer less than the precision of this
+          distribution
+
+        OUTPUT:
+
+        - a new distribution with `M` moments equal to the first `M`
+          moments of this distribution.
+
+        EXAMPLES::
+
+            sage: D = OverconvergentDistributions(3,7,10)
+            sage: v = D([3,4,5])
+            sage: v
+            (3 + O(7^3), 4 + O(7^2), 5 + O(7))
+            sage: v.reduce_precision(2)
+            (3 + O(7^3), 4 + O(7^2))
+        """
+        assert M <= self.precision_relative(), "not enough moments"
+
+        cdef Dist_vector ans = self._new_c()
+        ans._moments = self._moments[:M]
+        ans.ordp = self.ordp
+        return ans
+
+    def solve_difference_equation(self):
+        r"""
+        Solve the difference equation. `self = v | \Delta`, where `\Delta = [1, 1; 0, 1] - 1`.
+
+        See Theorem 4.5 and Lemma 4.4 of [PS2011]_.
+
+        OUTPUT:
+
+        - a distribution `v` so that `self = v | Delta` , assuming ``self.moment(0) == 0``.
+          Otherwise solves the difference equation for ``self - (self.moment(0),0,...,0)``.
+
+        EXAMPLES::
+
+            sage: D = OverconvergentDistributions(5,7,15)
+            sage: v = D(([0,2,3,4,5]))
+            sage: g = D._act.actor()(Matrix(ZZ,2,2,[1,1,0,1]))
+            sage: w = v.solve_difference_equation()
+            sage: v - (w*g - w)
+            (O(7^4), O(7^3), O(7^2), O(7))
+            sage: v = D(([7,2,3,4,5]))
+            sage: w = v.solve_difference_equation()
+            sage: v - (w*g - w)
+            (7 + O(7^4), O(7^3), O(7^2), O(7))
+        """
+        # assert self._moments[0][0]==0, "not total measure zero"
+        # print("result accurate modulo p^",self.moment(0).valuation(self.p) )
+        # v=[0 for j in range(i)]+[binomial(j,i)*bernoulli(j-i) for j in range(i,M)]
+        M = self.precision_relative()
+        R = self.parent().base_ring()
+        K = R.fraction_field()
+        V = self._moments.parent()
+        v = [K(0) for i in range(M)]
+        bern = [bernoulli(i) for i in range(0, M, 2)]
+        minhalf = ~K(-2)
+        for m in range(1, M):
+            scalar = K(self.moment(m) / m)
+            # bernoulli(1) = -1/2; the only nonzero odd Bernoulli number
+            v[m] += m * minhalf * scalar
+            for j in range(m - 1, M, 2):
+                v[j] += ZZ(j).binomial(m - 1) * bern[(j - m + 1) // 2] * scalar
+        p = self.parent().prime()
+        cdef Dist_vector ans
+        if p == 0:
+            if R in Fields():
+                ans = self._new_c()
+                ans.ordp = 0
+                ans._moments = V(v)
+            else:
+                newparent = self.parent().change_ring(K)
+                ans = newparent(v)
+        else:
+            ans = self._new_c()
+            try:
+                ans.ordp = min(a.valuation(p) for a in v)
+            except TypeError:
+                ans.ordp = 0
+            if ans.ordp < 0:
+                scalar = K(p) ** (-ans.ordp)
+                ans._moments = V([R(a * scalar) for a in v])
+            elif ans.ordp > 0:
+                scalar = K(p) ** ans.ordp
+                ans._moments = V([R(a // scalar) for a in v])
+            else:
+                ans._moments = V([R(a) for a in v])
+            v = ans._moments
+            N = len(ans._moments)
+            prec_loss = max([N - j - v[j].precision_absolute()
+                             for j in range(N)])
+            #            print("precision loss = ", prec_loss)
+            if prec_loss > 0:
+                ans._moments = ans._moments[:(N - prec_loss)]
+        return ans
+
+
+cdef class WeightKAction(Action):
+    r"""
+    Encode the action of the monoid `\Sigma_0(N)` on the space of distributions.
+
+    INPUT:
+
+    - ``Dk`` -- a space of distributions
+    - ``character`` -- data specifying a Dirichlet character to apply to
+      the top right corner, and a power of the determinant by which to scale.
+      See the documentation of
+      :class:`sage.modular.pollack_stevens.distributions.OverconvergentDistributions_factory`
+      for more details.
+    - ``adjuster`` -- a callable object that turns matrices into 4-tuples
+    - ``on_left`` -- whether this action should be on the left
+    - ``dettwist`` -- a power of the determinant to twist by
+    - ``padic`` -- if ``True``, define an action of `p`-adic matrices (not just integer ones)
+
+    EXAMPLES::
+
+        sage: D = OverconvergentDistributions(4,5,10,base = Qp(5,20)); D
+        Space of 5-adic distributions with k=4 action and precision cap 10
+        sage: D._act
+        Right action by Monoid Sigma0(5) with coefficients in 5-adic Field with capped relative precision 20 on Space of 5-adic distributions with k=4 action and precision cap 10
+    """
+    def __init__(self, Dk, character, adjuster, on_left, dettwist, padic=False):
+        r"""
+        Initialization.
+
+        TESTS::
+
+            sage: D = OverconvergentDistributions(4,5,10,base = Qp(5,20)); D # indirect doctest
+            Space of 5-adic distributions with k=4 action and precision cap 10
+            sage: D = Symk(10) # indirect doctest
+        """
+        self._k = Dk._k
+#        if self._k < 0: raise ValueError("k must not be negative")
+        self._adjuster = adjuster
+        self._character = character
+        self._dettwist = dettwist
+        self._p = Dk._p
+        self._symk = Dk.is_symk()
+        self._actmat = {}
+        self._maxprecs = {}
+        if character is None:
+            self._Np = ZZ(1)  # all of M2Z acts
+        else:
+            self._Np = character.modulus()
+        if not self._symk:
+            self._Np = self._Np.lcm(self._p)
+
+        if padic:
+            self._Sigma0 = Sigma0(self._Np, base_ring=Dk.base_ring(), adjuster=self._adjuster)
+        else:
+            self._Sigma0 = Sigma0(self._Np, base_ring=ZZ, adjuster=self._adjuster)
+        Action.__init__(self, self._Sigma0, Dk, on_left, operator.mul)
+
+    def clear_cache(self):
+        r"""
+        Clear the cached matrices which define the action of `U_p`
+        (these depend on the desired precision) and the
+        dictionary that stores the maximum precisions computed so far.
+
+        EXAMPLES::
+
+            sage: D = OverconvergentDistributions(4,5,4)
+            sage: D([1,2,5,3]) * D._act.actor()(Matrix(ZZ,2,2,[1,1,0,1]))
+            (1 + O(5^4), 3 + O(5^3), 2*5 + O(5^2), O(5))
+            sage: D._act.clear_cache()
+        """
+        self._actmat = {}
+        self._maxprecs = {}
+
+    cpdef acting_matrix(self, g, M):
+        r"""
+        The matrix defining the action of ``g`` at precision ``M``.
+
+        INPUT:
+
+        - ``g`` -- an instance of
+          :class:`sage.matrix.matrix_generic_dense.Matrix_generic_dense`
+
+        - ``M`` -- positive integer giving the precision at which
+          ``g`` should act
+
+        OUTPUT:
+
+        - An `M \times M` matrix so that the action of `g` on a
+          distribution with `M` moments is given by a vector-matrix
+          multiplication.
+
+        .. NOTE::
+
+            This function caches its results.  To clear the cache use
+            :meth:`clear_cache`.
+
+        EXAMPLES::
+
+            sage: D = Symk(3)
+            sage: v = D([5,2,7,1])
+            sage: g = Matrix(ZZ,2,2,[1,3,0,1])
+            sage: v * D._act.actor()(g) # indirect doctest
+            (5, 17, 64, 253)
+        """
+        g = g.matrix()
+        if g not in self._maxprecs:
+            A = self._compute_acting_matrix(g, M)
+            self._actmat[g] = {M: A}
+            self._maxprecs[g] = M
+            return A
+        else:
+            mats = self._actmat[g]
+            if M in mats:
+                return mats[M]
+            maxprec = self._maxprecs[g]
+            if M < maxprec:
+                A = mats[maxprec][:M, :M]  # submatrix; might want to reduce precisions
+                mats[M] = A
+                return A
+            if M < 30:  # This should not be hard-coded
+                maxprec = max([M, 2 * maxprec])  # This may be wasting memory
+            else:
+                maxprec = M
+            self._maxprecs[g] = maxprec
+            mats[maxprec] = self._compute_acting_matrix(g, maxprec)  # could lift from current maxprec
+            if M == maxprec:
+                return mats[maxprec]
+            A = mats[maxprec][:M, :M]  # submatrix; might want to reduce precisions
+            mats[M] = A
+            return A
+
+    cpdef _compute_acting_matrix(self, g, M):
+        r"""
+        Compute the matrix defining the action of ``g`` at precision ``M``.
+
+        INPUT:
+
+        - ``g`` -- a `2 \times 2` instance of
+          :class:`sage.matrices.matrix_integer_dense.Matrix_integer_dense`
+
+        - ``M`` -- positive integer giving the precision at which
+          ``g`` should act
+
+        OUTPUT:
+
+        - ``G`` -- an `M \times M` matrix. If `v `is the vector of moments of a
+          distribution `\mu`, then `v*G` is the vector of moments of `\mu|[a,b;c,d]`
+
+        EXAMPLES::
+
+            sage: D = Symk(3)
+            sage: v = D([5,2,7,1])
+            sage: g = Matrix(ZZ,2,2,[-2,1,-1,0])
+            sage: v * D._act.actor()(g) # indirect doctest
+            (-107, 35, -12, 5)
+        """
+        raise NotImplementedError
+
+
+cdef class WeightKAction_vector(WeightKAction):
+    cpdef _compute_acting_matrix(self, g, M):
+        r"""
+        Compute the matrix defining the action of ``g`` at precision ``M``.
+
+        INPUT:
+
+        - ``g`` -- a `2 \times 2` instance of
+          :class:`sage.matrix.matrix_generic_dense.Matrix_generic_dense`
+
+        - ``M`` -- positive integer giving the precision at which
+          ``g`` should act
+
+        OUTPUT:
+
+        - ``G`` -- an `M \times M` matrix. If `v` is the vector of moments of a
+          distribution `\mu`, then `v*G` is the vector of moments of `\mu|[a,b;c,d]`
+
+        EXAMPLES::
+
+            sage: D = Symk(3)
+            sage: v = D([5,2,7,1])
+            sage: g = Matrix(ZZ,2,2,[-2,1,-1,0])
+            sage: v * D._act.actor()(g) # indirect doctest
+            (-107, 35, -12, 5)
+        """
+        # tim = verbose("Starting")
+        a, b, c, d = self._adjuster(g)
+        # if g.parent().base_ring().is_exact():
+        #     self._check_mat(a, b, c, d)
+        k = self._k
+        if g.parent().base_ring() is ZZ:
+            if self._symk:
+                base_ring = QQ
+            else:
+                base_ring = Zmod(self._p ** M)
+        else:
+            base_ring = self.underlying_set().base_ring()
+        cdef Matrix B = matrix(base_ring, M, M)
+        if M == 0:
+            return B.change_ring(self.codomain().base_ring())
+        R = PowerSeriesRing(base_ring, 'y', default_prec=M)
+        y = R.gen()
+        # tim = verbose("Checked, made R",tim)
+        # special case for small precision, large weight
+        scale = (b + d * y) / (a + c * y)
+        t = (a + c * y) ** k  # will already have precision M
+        cdef long row, col
+        # tim = verbose("Made matrix",tim)
+        for col in range(M):
+            for row in range(M):
+                B.set_unsafe(row, col, t[row])
+            t *= scale
+        # verbose("Finished loop",tim)
+        # the changering here is annoying, but otherwise we have to
+        # change ring each time we multiply
+        B = B.change_ring(self.codomain().base_ring())
+        if self._character is not None:
+            B *= self._character(a)
+        if self._dettwist is not None:
+            B *= (a * d - b * c) ** (self._dettwist)
+        return B
+
+    cpdef _act_(self, g, _v):
+        r"""
+        The right action of ``g`` on a distribution.
+
+        INPUT:
+
+        - ``_v`` -- a :class:`Dist_vector` instance, the distribution
+          on which to act
+
+        - ``g`` -- a `2 \times 2` instance of
+          :class:`sage.matrix.matrix_integer_dense.Matrix_integer_dense`
+
+        OUTPUT: the distribution ``_v * g``
+
+        EXAMPLES::
+
+            sage: D = sage.modular.pollack_stevens.distributions.Symk(2)
+            sage: v = D([2,3,4])
+            sage: g = Matrix(ZZ,2,2,[3,-1,1,0])
+            sage: v * D._act.actor()(g) # indirect doctest
+            (40, -9, 2)
+        """
+        # if g is a matrix it needs to be immutable
+        # hashing on arithmetic_subgroup_elements is by str
+        if g == 1:
+            return _v
+        cdef Dist_vector v = <Dist_vector?>_v
+        cdef Dist_vector ans = v._new_c()
+
+        try:
+            g.set_immutable()
+        except AttributeError:
+            pass
+        v_moments = v._moments
+        ans._moments = v_moments * self.acting_matrix(g, len(v_moments))
+        ans.ordp = v.ordp
+        return ans