python-healpy 1.10.3-2 / usr / lib / python2.7 / dist-packages / healpy / sphtfunc.py

# 
#  This file is part of Healpy.
# 
#  Healpy 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.
# 
#  Healpy is distributed in the hope that it will be useful,
#  but WITHOUT ANY WARRANTY; without even the implied warranty of
#  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
#  GNU General Public License for more details.
# 
#  You should have received a copy of the GNU General Public License
#  along with Healpy; if not, write to the Free Software
#  Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
# 
#  For more information about Healpy, see http://code.google.com/p/healpy
# 
import warnings
import numpy as np
import six
pi = np.pi
import warnings
import astropy.io.fits as pf

from . import _healpy_sph_transform_lib as sphtlib
from . import _healpy_fitsio_lib as hfitslib
from . import _sphtools as _sphtools
from . import cookbook as cb

import os.path
from . import pixelfunc

from .pixelfunc import maptype, UNSEEN, ma_to_array, accept_ma

class FutureChangeWarning(UserWarning):
    pass

DATAPATH = os.path.join(os.path.dirname(os.path.abspath(__file__)), 'data')

# Spherical harmonics transformation
def anafast(map1, map2 = None, nspec = None, lmax = None, mmax = None, 
            iter = 3, alm = False, pol = True, use_weights = False,
            datapath = None):
    """Computes the power spectrum of an Healpix map, or the cross-spectrum
    between two maps if *map2* is given.
    No removal of monopole or dipole is performed.

    Parameters
    ----------
    map1 : float, array-like shape (Npix,) or (3, Npix)
      Either an array representing a map, or a sequence of 3 arrays
      representing I, Q, U maps
    map2 : float, array-like shape (Npix,) or (3, Npix)
      Either an array representing a map, or a sequence of 3 arrays
      representing I, Q, U maps
    nspec : None or int, optional
      The number of spectra to return. If None, returns all, otherwise
      returns cls[:nspec]
    lmax : int, scalar, optional
      Maximum l of the power spectrum (default: 3*nside-1)
    mmax : int, scalar, optional
      Maximum m of the alm (default: lmax)
    iter : int, scalar, optional
      Number of iteration (default: 3)
    alm : bool, scalar, optional
      If True, returns both cl and alm, otherwise only cl is returned
    pol : bool, optional
      If True, assumes input maps are TQU. Output will be TEB cl's and
      correlations (input must be 1 or 3 maps).
      If False, maps are assumed to be described by spin 0 spherical harmonics.
      (input can be any number of maps)
      If there is only one input map, it has no effect. Default: True.
    datapath : None or str, optional
      If given, the directory where to find the weights data.

    Returns
    -------
    res : array or sequence of arrays
      If *alm* is False, returns cl or a list of cl's (TT, EE, BB, TE, EB, TB for
      polarized input map)
      Otherwise, returns a tuple (cl, alm), where cl is as above and
      alm is the spherical harmonic transform or a list of almT, almE, almB
      for polarized input
    """

    map1 = ma_to_array(map1)
    alms1 = map2alm(map1, lmax = lmax, mmax = mmax, pol = pol, iter = iter, 
                    use_weights = use_weights,
                    datapath = datapath)
    if map2 is not None:
        map2 = ma_to_array(map2)
        alms2 = map2alm(map2, lmax = lmax, mmax = mmax, pol = pol,
                        iter = iter, use_weights = use_weights, 
                        datapath = datapath)
    else:
        alms2 = None
    
    cls = alm2cl(alms1, alms2 = alms2, lmax = lmax, mmax = mmax,
                 lmax_out = lmax, nspec = nspec)

    if alm:
        if map2 is not None:
            return (cls, alms1, alms2)
        else:
            return (cls, alms1)
    else:
        return cls

def map2alm(maps, lmax = None, mmax = None, iter = 3, pol = True, 
            use_weights = False, datapath = None):
    """Computes the alm of an Healpix map.
    
    Parameters
    ----------
    maps : array-like, shape (Npix,) or (n, Npix)
      The input map or a list of n input maps.
    lmax : int, scalar, optional
      Maximum l of the power spectrum. Default: 3*nside-1
    mmax : int, scalar, optional
      Maximum m of the alm. Default: lmax
    iter : int, scalar, optional
      Number of iteration (default: 3)
    pol : bool, optional
      If True, assumes input maps are TQU. Output will be TEB alm's.
      (input must be 1 or 3 maps)
      If False, apply spin 0 harmonic transform to each map.
      (input can be any number of maps)
      If there is only one input map, it has no effect. Default: True.
    use_weights: bool, scalar, optional
      If True, use the ring weighting. Default: False.
    datapath : None or str, optional
      If given, the directory where to find the weights data.
    
    Returns
    -------
    alms : array or tuple of array
      alm or a tuple of 3 alm (almT, almE, almB) if polarized input.
    
    Notes
    -----
    The pixels which have the special `UNSEEN` value are replaced by zeros
    before spherical harmonic transform. They are converted back to `UNSEEN`
    value, so that the input maps are not modified. Each map have its own, 
    independent mask.
    """
    maps = ma_to_array(maps)
    info = maptype(maps)
    if pol or info in (0, 1):
        alms = _sphtools.map2alm(maps, niter = iter, 
                                 datapath = datapath, use_weights = use_weights,
                                 lmax = lmax, mmax = mmax)
    else:
        # info >= 2 and pol is False : spin 0 spht for each map
        alms = [_sphtools.map2alm(mm, niter = iter,
                                  datapath = datapath, use_weights = use_weights,
                                  lmax = lmax, mmax = mmax)
               for mm in maps]
    return alms

def alm2map(alms, nside, lmax = None, mmax = None, pixwin = False,
            fwhm = 0.0, sigma = None,  pol = True,
            inplace = False, verbose=True):
    """Computes an Healpix map given the alm.

    The alm are given as a complex array. You can specify lmax
    and mmax, or they will be computed from array size (assuming
    lmax==mmax).

    Parameters
    ----------
    alms : complex, array or sequence of arrays
      A complex array or a sequence of complex arrays.
      Each array must have a size of the form: mmax * (2 * lmax + 1 - mmax) / 2 + lmax + 1
    nside : int, scalar
      The nside of the output map.
    lmax : None or int, scalar, optional
      Explicitly define lmax (needed if mmax!=lmax)
    mmax : None or int, scalar, optional
      Explicitly define mmax (needed if mmax!=lmax)
    pixwin : bool, optional
      Smooth the alm using the pixel window functions. Default: False.
    fwhm : float, scalar, optional
      The fwhm of the Gaussian used to smooth the map (applied on alm)
      [in radians]
    sigma : float, scalar, optional
      The sigma of the Gaussian used to smooth the map (applied on alm)
      [in radians]
    pol : bool, optional
      If True, assumes input alms are TEB. Output will be TQU maps.
      (input must be 1 or 3 alms)
      If False, apply spin 0 harmonic transform to each alm.
      (input can be any number of alms)
      If there is only one input alm, it has no effect. Default: True.
    inplace : bool, optional
      If True, input alms may be modified by pixel window function and beam
      smoothing (if alm(s) are complex128 contiguous arrays).
      Otherwise, input alms are not modified. A copy is made if needed to
      apply beam smoothing or pixel window.
      
    Returns
    -------
    maps : array or list of arrays
      An Healpix map in RING scheme at nside or a list of T,Q,U maps (if
      polarized input)
    """
    if not cb.is_seq(alms):
        raise TypeError("alms must be a sequence")

    alms = smoothalm(alms, fwhm = fwhm, sigma = sigma,  
                     pol = pol, inplace = inplace, verbose=verbose)

    if not cb.is_seq_of_seq(alms):
        alms = [alms]
        lonely = True
    else:
        lonely = False
    
    if pixwin:
        pw = globals()['pixwin'](nside,True)
        alms_new = []
        for ialm, alm in enumerate(alms):
            pixelwindow = pw[1] if ialm >= 1 and pol else pw[0]
            alms_new.append(almxfl(alm, pixelwindow, inplace = inplace))
    else:
        alms_new = alms

    if lmax is None:
        lmax = -1
    if mmax is None:
        mmax = -1
    if pol:
        output = sphtlib._alm2map(alms_new[0] if lonely else alms_new,
                                  nside, lmax = lmax, mmax = mmax)
        if lonely:
            output = [output]
    else:
        output = [sphtlib._alm2map(alm, nside, lmax = lmax, mmax = mmax)
                  for alm in alms_new]
    if lonely:
        return output[0]
    else:
        return output

def synalm(cls, lmax = None, mmax = None, new = False, verbose=True):
    """Generate a set of alm given cl.
    The cl are given as a float array. Corresponding alm are generated.
    If lmax is None, it is assumed lmax=cl.size-1
    If mmax is None, it is assumed mmax=lmax.

    Parameters
    ----------
    cls : float, array or tuple of arrays
      Either one cl (1D array) or a tuple of either 4 cl
      or of n*(n+1)/2 cl.
      Some of the cl may be None, implying no
      cross-correlation. See *new* parameter.
    lmax : int, scalar, optional
      The lmax (if None or <0, the largest size-1 of cls)
    mmax : int, scalar, optional
      The mmax (if None or <0, =lmax)
    new : bool, optional
      If True, use the new ordering of cl's, ie by diagonal
      (e.g. TT, EE, BB, TE, EB, TB or TT, EE, BB, TE if 4 cl as input).
      If False, use the old ordering, ie by row
      (e.g. TT, TE, TB, EE, EB, BB or TT, TE, EE, BB if 4 cl as input).      

    Returns
    -------
    alms : array or list of arrays
      the generated alm if one spectrum is given, or a list of n alms 
      (with n(n+1)/2 the number of input cl, or n=3 if there are 4 input cl).

    Notes
    -----
    The order of the spectra will change in a future release. The new= parameter
    help to make the transition smoother. You can start using the new order
    by setting new=True.
    In the next version of healpy, the default will be new=True.
    This change is done for consistency between the different tools
    (alm2cl, synfast, anafast).
    In the new order, the spectra are ordered by diagonal of the correlation
    matrix. Eg, if fields are T, E, B, the spectra are TT, EE, BB, TE, EB, TB
    with new=True, and TT, TE, TB, EE, EB, BB if new=False.
    """
    if (not new) and verbose:
        warnings.warn("The order of the input cl's will change in a future "
                      "release.\n"
                      "Use new=True keyword to start using the new order.\n"
                      "See documentation of healpy.synalm.",
                      category=FutureChangeWarning)
    if not cb.is_seq(cls):
        raise TypeError('cls must be an array or a sequence of arrays')

    if not cb.is_seq_of_seq(cls):
        # Only one spectrum
        if lmax is None or lmax < 0:
            lmax = cls.size-1
        if mmax is None or mmax < 0:
            mmax = lmax
        cls_list = [np.asarray(cls, dtype = np.float64)]
        szalm = Alm.getsize(lmax,mmax)
        alm = np.zeros(szalm,'D')
        alm.real = np.random.standard_normal(szalm)
        alm.imag = np.random.standard_normal(szalm)
        alms_list=[alm]
        sphtlib._synalm(cls_list,alms_list,lmax,mmax)
        return alm

    # From here, we interpret cls as a list of spectra
    cls_list = list(cls)
    maxsize = max([len(c) for c in cls])
    
    if lmax is None or lmax < 0:
        lmax = maxsize-1
    if mmax is None or mmax < 0:
        mmax = lmax
    
    Nspec = sphtlib._getn(len(cls_list))
    
    if Nspec <= 0:
        if len(cls_list) == 4:
            if new: ## new input order: TT EE BB TE -> TT EE BB TE 0 0
                cls_list = [cls[0], cls[1], cls[2], cls[3], None, None]
            else: ## old input order: TT TE EE BB -> TT TE 0 EE 0 BB
                cls_list = [cls[0], cls[1], None, cls[2], None, cls[3]]
            Nspec = 3
        else:
            raise TypeError("The sequence of arrays must have either 4 elements "
                            "or n(n+1)/2 elements (some may be None)")
    
    szalm = Alm.getsize(lmax,mmax)
    alms_list = []
    for i in six.moves.xrange(Nspec):
        alm = np.zeros(szalm,'D')
        alm.real = np.random.standard_normal(szalm)
        alm.imag = np.random.standard_normal(szalm)
        alms_list.append(alm)
    if new: # new input order: input given by diagonal, should be given by row
        cls_list = new_to_old_spectra_order(cls_list)
    # ensure cls are float64
    cls_list = [(np.asarray(cl, dtype = np.float64) if cl is not None else None)
                for cl in cls_list]
    sphtlib._synalm(cls_list, alms_list, lmax, mmax)
    return alms_list

def synfast(cls, nside, lmax = None, mmax = None, alm = False,
            pol = True, pixwin = False, fwhm = 0.0, sigma = None,
            new = False, verbose=True):
    """Create a map(s) from cl(s).

    Parameters
    ----------
    cls : array or tuple of array
      A cl or a list of cl (either 4 or 6, see :func:`synalm`)
    nside : int, scalar
      The nside of the output map(s)
    lmax : int, scalar, optional
      Maximum l for alm. Default: min of 3*nside-1 or length of the cls - 1
    mmax : int, scalar, optional
      Maximum m for alm.
    alm : bool, scalar, optional
      If True, return also alm(s). Default: False.
    pol : bool, optional
      If True, assumes input cls are TEB and correlation. Output will be TQU maps.
      (input must be 1, 4 or 6 cl's)
      If False, fields are assumed to be described by spin 0 spherical harmonics.
      (input can be any number of cl's)
      If there is only one input cl, it has no effect. Default: True.
    pixwin : bool, scalar, optional
      If True, convolve the alm by the pixel window function. Default: False.
    fwhm : float, scalar, optional
      The fwhm of the Gaussian used to smooth the map (applied on alm)
      [in radians]
    sigma : float, scalar, optional
      The sigma of the Gaussian used to smooth the map (applied on alm)
      [in radians]

    Returns
    -------
    maps : array or tuple of arrays
      The output map (possibly list of maps if polarized input).
      or, if alm is True, a tuple of (map,alm)
      (alm possibly a list of alm if polarized input)

    Notes
    -----
    The order of the spectra will change in a future release. The new= parameter
    help to make the transition smoother. You can start using the new order
    by setting new=True.
    In the next version of healpy, the default will be new=True.
    This change is done for consistency between the different tools
    (alm2cl, synfast, anafast).
    In the new order, the spectra are ordered by diagonal of the correlation
    matrix. Eg, if fields are T, E, B, the spectra are TT, EE, BB, TE, EB, TB
    with new=True, and TT, TE, TB, EE, EB, BB if new=False.
    """
    if not pixelfunc.isnsideok(nside):
        raise ValueError("Wrong nside value (must be a power of two).")
    cls_lmax = cb.len_array_or_arrays(cls) -1
    if lmax is None or lmax < 0:
        lmax = min(cls_lmax, 3 * nside - 1)
    alms = synalm(cls, lmax = lmax, mmax = mmax, new = new, verbose=verbose)
    maps = alm2map(alms, nside, lmax = lmax, mmax = mmax, pixwin = pixwin,
                   pol = pol, fwhm = fwhm, sigma = sigma, inplace = True, verbose=verbose)
    if alm:
        return maps, alms
    else:
        return maps
    
class Alm(object):
    """This class provides some static methods for alm index computation.

    Methods
    -------
    getlm
    getidx
    getsize
    getlmax
    """
    def __init__(self):
        pass

    @staticmethod
    def getlm(lmax,i=None):
        """Get the l and m from index and lmax.
        
        Parameters
        ----------
        lmax : int
          The maximum l defining the alm layout
        i : int or None
          The index for which to compute the l and m.
          If None, the function return l and m for i=0..Alm.getsize(lmax)
        """
        if i is None:
            i=np.arange(Alm.getsize(lmax))
        m=(np.ceil(((2*lmax+1)-np.sqrt((2*lmax+1)**2-8*(i-lmax)))/2)).astype(int)
        l = i-m*(2*lmax+1-m)//2
        return (l,m)

    @staticmethod
    def getidx(lmax,l,m):
        """Returns index corresponding to (l,m) in an array describing alm up to lmax.
        
        Parameters
        ----------
        lmax : int
          The maximum l, defines the alm layout
        l : int
          The l for which to get the index
        m : int
          The m for which to get the index

        Returns
        -------
        idx : int
          The index corresponding to (l,m)
        """
        return m*(2*lmax+1-m)//2+l

    @staticmethod
    def getsize(lmax,mmax = None):
        """Returns the size of the array needed to store alm up to *lmax* and *mmax*

        Parameters
        ----------
        lmax : int
          The maximum l, defines the alm layout
        mmax : int, optional
          The maximum m, defines the alm layout. Default: lmax.

        Returns
        -------
        size : int
          The size of the array needed to store alm up to lmax, mmax.
        """
        if mmax is None or mmax < 0 or mmax > lmax:
            mmax = lmax
        return mmax * (2 * lmax + 1 - mmax) // 2 + lmax + 1

    @staticmethod
    def getlmax(s, mmax = None):
        """Returns the lmax corresponding to a given array size.
        
        Parameters
        ----------
        s : int
          Size of the array
        mmax : None or int, optional
          The maximum m, defines the alm layout. Default: lmax.

        Returns
        -------
        lmax : int
          The maximum l of the array, or -1 if it is not a valid size.
        """
        if mmax is not None and mmax >= 0:
            x = (2 * s + mmax ** 2 - mmax - 2) / (2 * mmax + 2)
        else:
            x = (-3 + np.sqrt(1 + 8 * s)) / 2
        if x != np.floor(x):
            return -1
        else:
            return int(x)


def alm2cl(alms1, alms2 = None, lmax = None, mmax = None,
           lmax_out = None, nspec = None):
    """Computes (cross-)spectra from alm(s). If alm2 is given, cross-spectra between
    alm and alm2 are computed. If alm (and alm2 if provided) contains n alm,
    then n(n+1)/2 auto and cross-spectra are returned.

    Parameters
    ----------
    alm : complex, array or sequence of arrays
      The alm from which to compute the power spectrum. If n>=2 arrays are given,
      computes both auto- and cross-spectra.
    alms2 : complex, array or sequence of 3 arrays, optional
      If provided, computes cross-spectra between alm and alm2.
      Default: alm2=alm, so auto-spectra are computed.
    lmax : None or int, optional
      The maximum l of the input alm. Default: computed from size of alm
      and mmax_in
    mmax : None or int, optional
      The maximum m of the input alm. Default: assume mmax_in = lmax_in
    lmax_out : None or int, optional
      The maximum l of the returned spectra. By default: the lmax of the given
      alm(s).
    nspec : None or int, optional
      The number of spectra to return. None means all, otherwise returns cl[:nspec]

    Returns
    -------
    cl : array or tuple of n(n+1)/2 arrays
      the spectrum <*alm* x *alm2*> if *alm* (and *alm2*) is one alm, or 
      the auto- and cross-spectra <*alm*[i] x *alm2*[j]> if alm (and alm2)
      contains more than one spectra.
      If more than one spectrum is returned, they are ordered by diagonal.
      For example, if *alm* is almT, almE, almB, then the returned spectra are:
      TT, EE, BB, TE, EB, TB.
    """
    cls = _sphtools.alm2cl(alms1, alms2 = alms2, lmax = lmax,
                           mmax = mmax, lmax_out = lmax_out)
    if nspec is None:
        return cls
    else:
        return cls[:nspec]

    
def almxfl(alm, fl, mmax = None, inplace = False):
    """Multiply alm by a function of l. The function is assumed
    to be zero where not defined.

    Parameters
    ----------
    alm : array
      The alm to multiply
    fl : array
      The function (at l=0..fl.size-1) by which alm must be multiplied.
    mmax : None or int, optional
      The maximum m defining the alm layout. Default: lmax.
    inplace : bool, optional
      If True, modify the given alm, otherwise make a copy before multiplying.

    Returns
    -------
    alm : array
      The modified alm, either a new array or a reference to input alm, 
      if inplace is True.
    """
    almout = _sphtools.almxfl(alm, fl, mmax = mmax, inplace = inplace)
    return almout

def smoothalm(alms, fwhm = 0.0, sigma = None,  pol = True,
              mmax = None, verbose = True, inplace = True):
    """Smooth alm with a Gaussian symmetric beam function.

    Parameters
    ----------
    alms : array or sequence of 3 arrays
      Either an array representing one alm, or a sequence of arrays.
      See *pol* parameter.
    fwhm : float, optional
      The full width half max parameter of the Gaussian. Default:0.0
      [in radians]
    sigma : float, optional
      The sigma of the Gaussian. Override fwhm.
      [in radians]
    pol : bool, optional
      If True, assumes input alms are TEB. Output will be TQU maps.
      (input must be 1 or 3 alms)
      If False, apply spin 0 harmonic transform to each alm.
      (input can be any number of alms)
      If there is only one input alm, it has no effect. Default: True.
    mmax : None or int, optional
      The maximum m for alm. Default: mmax=lmax
    inplace : bool, optional
      If True, the alm's are modified inplace if they are contiguous arrays
      of type complex128. Otherwise, a copy of alm is made. Default: True.
    verbose : bool, optional
      If True prints diagnostic information. Default: True

    Returns
    -------
    alms : array or sequence of 3 arrays
      The smoothed alm. If alm[i] is a contiguous array of type complex128,
      and *inplace* is True the smoothing is applied inplace.
      Otherwise, a copy is made.
    """
    if sigma is None:
        sigma = fwhm / (2.*np.sqrt(2.*np.log(2.)))

    if verbose:
        print("Sigma is {0:f} arcmin ({1:f} rad) ".format(sigma*60*180/pi,sigma))
        print("-> fwhm is {0:f} arcmin".format(sigma*60*180/pi*(2.*np.sqrt(2.*np.log(2.)))))

    # Check alms
    if not cb.is_seq(alms):
        raise ValueError("alm must be a sequence")

    if sigma == 0:
        # nothing to be done
        return alms

    lonely = False
    if not cb.is_seq_of_seq(alms):
        alms = [alms]
        lonely = True
    
    # we have 3 alms -> apply smoothing to each map.
    # polarization has different B_l from temperature
    # exp{-[ell(ell+1) - s**2] * sigma**2/2}
    # with s the spin of spherical harmonics
    # s = 2 for pol, s=0 for temperature
    retalm = []
    for ialm, alm in enumerate(alms):
        lmax = Alm.getlmax(len(alm), mmax)
        if lmax < 0:
            raise TypeError('Wrong alm size for the given '
                            'mmax (len(alms[%d]) = %d).'%(ialm, len(alm)))
        ell = np.arange(lmax + 1.)
        s = 2 if ialm >= 1 and pol else 0
        fact = np.exp(-0.5 * (ell * (ell + 1) - s ** 2) * sigma ** 2)
        res = almxfl(alm, fact, mmax = mmax, inplace = inplace)
        retalm.append(res)
    # Test what to return (inplace/not inplace...)
    # Case 1: 1d input, return 1d output
    if lonely:
        return retalm[0]
    # case 2: 2d input, check if in-place smoothing for all alm's
    for i in six.moves.xrange(len(alms)):
        samearray = alms[i] is retalm[i]
        if not samearray:
            # Case 2a:
            # at least one of the alm could not be smoothed in place:
            # return the list of alm
            return retalm
    # Case 2b:
    # all smoothing have been performed in place:
    # return the input alms
    return alms

@accept_ma
def smoothing(map_in, fwhm = 0.0, sigma = None,  pol = True,
              iter = 3, lmax = None, mmax = None, use_weights = False,
              datapath = None, verbose = True):
    """Smooth a map with a Gaussian symmetric beam.

    No removal of monopole or dipole is performed.

    Parameters
    ----------
    map_in : array or sequence of 3 arrays
      Either an array representing one map, or a sequence of
      3 arrays representing 3 maps, accepts masked arrays
    fwhm : float, optional
      The full width half max parameter of the Gaussian [in 
      radians]. Default:0.0
    sigma : float, optional
      The sigma of the Gaussian [in radians]. Override fwhm.
    pol : bool, optional
      If True, assumes input maps are TQU. Output will be TQU maps.
      (input must be 1 or 3 alms)
      If False, each map is assumed to be a spin 0 map and is 
      treated independently (input can be any number of alms).
      If there is only one input map, it has no effect. Default: True.
    iter : int, scalar, optional
      Number of iteration (default: 3)
    lmax : int, scalar, optional
      Maximum l of the power spectrum. Default: 3*nside-1
    mmax : int, scalar, optional
      Maximum m of the alm. Default: lmax
    use_weights: bool, scalar, optional
      If True, use the ring weighting. Default: False.
    datapath : None or str, optional
      If given, the directory where to find the weights data.
    verbose : bool, optional
      If True prints diagnostic information. Default: True

    Returns
    -------
    maps : array or list of 3 arrays
      The smoothed map(s)
    """

    if not cb.is_seq(map_in):
        raise TypeError("map_in must be a sequence")

    # save the masks of inputs
    masks = pixelfunc.mask_bad(map_in) 

    if cb.is_seq_of_seq(map_in):
        nside = pixelfunc.npix2nside(len(map_in[0]))
        n_maps = len(map_in)
    else:
        nside = pixelfunc.npix2nside(len(map_in))
        n_maps = 0

    if pol or n_maps in (0, 1):
        # Treat the maps together (1 or 3 maps)
        alms = map2alm(map_in, lmax = lmax, mmax = mmax, iter = iter,
                       pol = pol, use_weights = use_weights,
                       datapath = datapath)
        smoothalm(alms, fwhm = fwhm, sigma = sigma, 
                  inplace = True, verbose = verbose)
        output_map = alm2map(alms, nside, pixwin = False, verbose=verbose)
    else:
        # Treat each map independently (any number)
        output_map = []
        for m in map_in:
            alm = map2alm(m, lmax = lmax, mmax = mmax, iter = iter, pol = pol,
                          use_weights = use_weights, datapath = datapath)
            smoothalm(alm, fwhm = fwhm, sigma = sigma, 
                      inplace = True, verbose = verbose)
            output_map.append(alm2map(alm, nside, pixwin = False, verbose=verbose))
    if pixelfunc.maptype(output_map) == 0:
        output_map[masks.flatten()] = UNSEEN
    else:
        for m, mask in zip(output_map, masks):
            m[mask] = UNSEEN
        
    return output_map

def pixwin(nside, pol = False):
    """Return the pixel window function for the given nside.

    Parameters
    ----------
    nside : int
      The nside for which to return the pixel window function
    pol : bool, optional
      If True, return also the polar pixel window. Default: False

    Returns
    -------
    pw or pwT,pwP : array or tuple of 2 arrays
      The temperature pixel window function, or a tuple with both
      temperature and polarisation pixel window functions.
    """
    datapath = DATAPATH
    if not pixelfunc.isnsideok(nside):
        raise ValueError("Wrong nside value (must be a power of two).")
    fname = os.path.join(datapath, 'pixel_window_n%04d.fits'%nside)
    if not os.path.isfile(fname):
        raise ValueError("No pixel window for this nside "
                         "or data files missing")
    # return hfitslib._pixwin(nside,datapath,pol)  ## BROKEN -> seg fault...
    pw = pf.getdata(fname)
    pw_temp, pw_pol = pw.field(0), pw.field(1)
    if pol:
        return pw_temp, pw_pol
    else:
        return pw_temp

def alm2map_der1(alm, nside, lmax = None, mmax = None):
   """Computes an Healpix map and its first derivatives given the alm.

   The alm are given as a complex array. You can specify lmax
   and mmax, or they will be computed from array size (assuming
   lmax==mmax).

   Parameters
   ----------
   alm : array, complex
     A complex array of alm. Size must be of the form mmax(lmax-mmax+1)/2+lmax
   nside : int
     The nside of the output map.
   lmax : None or int, optional
     Explicitly define lmax (needed if mmax!=lmax)
   mmax : None or int, optional
     Explicitly define mmax (needed if mmax!=lmax)

   Returns
   -------
   m, d_theta, d_phi : tuple of arrays
     The maps correponding to alm, and its derivatives with respect to
     theta and phi. d_phi is already divided by sin(theta)
   """
   if lmax is None:
       lmax = -1
   if mmax is None:
       mmax = -1
   return sphtlib._alm2map_der1(alm,nside,lmax=lmax,mmax=mmax)

def new_to_old_spectra_order(cls_new_order):
    """Reorder the cls from new order (by diagonal) to old order (by row).
    For example : TT, EE, BB, TE, EB, BB => TT, TE, TB, EE, EB, BB
    """
    Nspec = sphtlib._getn(len(cls_new_order))
    if Nspec < 0:
        raise ValueError("Input must be a list of n(n+1)/2 arrays")
    cls_old_order = []
    for i in six.moves.xrange(Nspec):
        for j in six.moves.xrange(i, Nspec):
            p = j - i
            q = i
            idx_new = p * (2 * Nspec + 1 - p) // 2 + q
            cls_old_order.append(cls_new_order[idx_new])
    return cls_old_order

def load_sample_spectra():
    """Read a sample power spectra for testing and demo purpose.
    Based on LambdaCDM. Gives TT, EE, BB, TE.

    Returns
    -------
    ell, f, cls : arrays
      ell is the array of ell values (from 0 to lmax)
      f is the factor ell*(ell+1)/2pi (in general, plots show f * cl)
      cls is a sequence of the power spectra TT, EE, BB and TE
    """
    cls = np.loadtxt(os.path.join(DATAPATH, 'totcls.dat'), unpack = True)
    ell = cls[0]
    f = ell * (ell + 1) / 2 / np.pi
    cls[1:, 1:] /= f[1:]
    return ell, f, cls[1:]

def gauss_beam(fwhm, lmax=512, pol=False):
    """Gaussian beam window function

    Computes the spherical transform of an axisimmetric gaussian beam

    For a sky of underlying power spectrum C(l) observed with beam of  
    given FWHM, the measured power spectrum will be 
    C(l)_meas = C(l) B(l)^2
    where B(l) is given by gaussbeam(Fwhm,Lmax). 
    The polarization beam is also provided (when pol = True ) assuming 
    a perfectly co-polarized beam 
    (e.g., Challinor et al 2000, astro-ph/0008228)

    Parameters
    ----------
    fwhm : float
        full width half max in radians
    lmax : integer
        ell max
    pol : bool
        if False, output has size (lmax+1) and is temperature beam
        if True output has size (lmax+1, 4) with components:
        * temperature beam
        * grad/electric polarization beam
        * curl/magnetic polarization beam
        * temperature * grad beam

    Returns
    -------
    beam : array
        beam window function [0, lmax] if dim not specified
        otherwise (lmax+1, 4) contains polarized beam
    """

    sigma = fwhm / np.sqrt(8. * np.log(2.))
    ell = np.arange(lmax + 1)
    sigma2 = sigma ** 2
    g = np.exp(-.5 * ell * (ell + 1) * sigma2)

    if not pol: # temperature-only beam
        return g
    else: # polarization beam
        # polarization factors [1, 2 sigma^2, 2 sigma^2, sigma^2]
        pol_factor = np.exp([0., 2*sigma2, 2*sigma2, sigma2])
        return g[:, np.newaxis] * pol_factor