#...............................................................................
""" new colormaps from old: stack, truncate builtin cmaps / files / numpy arrays

What's a colormap or cmap in matplotlib ?
Mainly a bar or array of 256 colors, rgb or rgba values 0 .. 1,
used in
    pl.imshow( a 2d numpy array, cmap=cmap, ... )
    pl.colorbar()
Cmaps can be indexed with () like
    cmap( .25 ),  cmap( [0, .25, .5] ),  cmap( np.linspace( ... ))
to get rgb values.

The functions below return cmaps:
    get_cmap(): "Blues" ... builtins / filename / numpy array
    array_cmap(): a numpy array, n x 3 or 4  ints 0..255 or floats 0..1
    truncate_colormap(): subset
    stack_colormap(): A B -> bottom half, A, top half B.
    band_colormap(): e.g. 10 bands

See also
    http://matplotlib.org/api/colors_api.html   $matplotlib/colors.py
    http://matplotlib.org/api/cm_api.html       $matplotlib/cm.py
    http://en.wikipedia.org/wiki/Indexed_color
"""

from __future__ import division
import numpy as np
from matplotlib import pyplot as pl, cm, colors

__version__ = "2013-12-19 dec denis"


def truncate_colormap( cmap, minval=0.0, maxval=1.0, n=256 ):
    """ mycolormap = truncate_colormap(
            cmap name or file or ndarray,
            minval=0.2, maxval=0.8 ): subset
            minval=1, maxval=0 )    : reverse
    by unutbu http://stackoverflow.com/questions/18926031/how-to-extract-a-subset-of-a-colormap-as-a-new-colormap-in-matplotlib
    """
    cmap = get_cmap( cmap )
    name = "%s-trunc-%.2g-%.2g" % (cmap.name, minval, maxval)
    return colors.LinearSegmentedColormap.from_list(
        name, cmap( np.linspace( minval, maxval, n )))

def stack_colormap( A, B, n=256 ):
    """ low half -> A colors, high half -> B colors """
    A = get_cmap( A )
    B = get_cmap( B )
    name = "%s-%s" % (A.name, B.name)
    lin = np.linspace( 0, 1, n )
    return array_cmap( np.vstack(( A(lin), B(lin) )), name, n=n )

def get_cmap( cmap, name=None, n=256 ):
    """ in: a name "Blues" "BuGn_r" ... of a builtin cmap (case-sensitive)
        or a filename, np.loadtxt() n x 3 or 4  ints 0..255 or floats 0..1
        or a cmap already
        or a numpy array.
        See http://wiki.scipy.org/Cookbook/Matplotlib/Show_colormaps
        or in IPython, pl.cm.<tab>
    """
    if isinstance( cmap, colors.Colormap ):
        return cmap
    if isinstance( cmap, basestring ):
        if cmap in cm.cmap_d:
            return pl.get_cmap( cmap )  # "Blues" ...
        A = np.loadtxt( cmap, delimiter=None )  # None: white space
        name = name or cmap.split("/")[-1] .split(".")[0]  # .../xx.csv -> xx
    else:
        A = cmap  # numpy array or array-like
    return array_cmap( A, name, n=n )

def array_cmap( A, name=None, n=256 ):
    """ numpy array -> a cmap, matplotlib.colors.Colormap
        n x 3 or 4  ints 0 .. 255 or floats 0 ..1
    """
    A = np.asanyarray( A )
    assert A.ndim == 2  and A.shape[1] in (3, 4), \
            "array must be n x 3 or 4, not %s" % str(A.shape)
    Amin, Amax = A.min(), A.max()
    if A.dtype.kind == "i":
        assert 0 <= Amin < Amax <= 255, "Amin %d  Amax %d must be in 0 .. 255" % (Amin, Amax)
        A = A / 255.  # not /=
    else:
        assert 0 <= Amin < Amax <= 1, "Amin %g  Amax %g must be in 0 .. 1" % (Amin, Amax)
    return colors.LinearSegmentedColormap.from_list( name or "noname", A, N=n )

def save_cmap( outfile, cmap ):
    """ -> a file of 256 x 4 ints 0 .. 255
        to load it, np.loadtxt() or get_cmap( filename )
    """
    cmap = get_cmap( cmap )
    A = cmap( np.linspace( 0, 1, 256 ))
    np.savetxt( outfile, A * 255, fmt="%4.0f",
            header="colormap %s" % cmap.name )  # From ...

def band_colormap( cmap, nband=10 ):
    """ -> a colormap with e.g. 10 bands """
    cmap = get_cmap( cmap )
    h = .5 / nband
    A = cmap( np.linspace( h, 1 - h, nband ))
    name = "%s-band-%d" % (cmap.name, nband)
    return array_cmap( A, name, n=nband )

#...............................................................................
cmap_brown = truncate_colormap( pl.cm.PuOr, minval=.5, maxval=0 )  # left half, flipped
cmap_bluebrown = stack_colormap( "Blues_r", cmap_brown )
cmap_bluebrown10 = band_colormap( cmap_bluebrown, 10 )
    # Tufte, Envisioning info p. 91: bathymetric


#...............................................................................
if __name__ == "__main__":
    import sys

    cmap = cmap_bluebrown10
    bw = array_cmap( [ [0.,0,0], [1,1,1] ], name="bw", n=2 )
    plot = 0

        # run this.py a=1 b=None c=[3] 'd = expr' ...  in sh or ipython
    exec( "\n".join( sys.argv[1:] ))
    np.set_printoptions( 2, threshold=100, edgeitems=10, linewidth=100, suppress=True )

    print cmap.name, "\n", cmap( np.arange( 120, 136 ) / 256 ).T
    save_cmap( cmap.name + ".tmp", cmap )

    if plot:
        A = np.arange(8**2) .reshape((8,8))
        im = pl.imshow( A, cmap=cmap, interpolation="nearest" )  # nearest: big squares
            # imshow_mouse_z
        pl.colorbar( im )
        pl.show()