"""
=========================================
Creating a Full Sun Map with AIA and EUVI
=========================================

With SDO/AIA and STEREO/A and STEREO/B, it is possible (for specific dates)
to combine combine three EUV images from these satellites to produce a nearly
full latitude / longitude map of the Sun.

You will need `reproject <https://reproject.readthedocs.io/en/stable/>`__ v0.6 or higher installed.
"""
# sphinx_gallery_thumbnail_number = 4

import matplotlib.pyplot as plt
import numpy as np
from reproject import reproject_interp
from reproject.mosaicking import reproject_and_coadd

import astropy.units as u
from astropy.coordinates import SkyCoord
from astropy.wcs import WCS

import sunpy.map
import sunpy.sun
from sunpy.coordinates import get_body_heliographic_stonyhurst
from sunpy.data.sample import AIA_193_JUN2012, STEREO_A_195_JUN2012, STEREO_B_195_JUN2012

######################################################################
# First create a sunpy map for each of the files.

maps = sunpy.map.Map(sorted([AIA_193_JUN2012, STEREO_A_195_JUN2012, STEREO_B_195_JUN2012]))

######################################################################
# To reduce memory consumption we also downsample these maps before continuing,
# you can disable this.

maps = [m.resample((1024, 1024)*u.pix) for m in maps]

######################################################################
# When combining these images all three need to assume the same radius of
# the Sun for the data. The AIA images specify a slightly different value
# than the IAU 2015 constant. To avoid coordinate transformation issues we
# reset this here.

maps[0].meta['rsun_ref'] = sunpy.sun.constants.radius.to_value(u.m)

######################################################################
# Next we will plot the locations of the three spacecraft with respect to
# the Sun so we can easily see the relative separations.

earth = get_body_heliographic_stonyhurst('earth', maps[0].date)

fig = plt.figure(figsize=(8, 8))
ax = fig.add_subplot(projection='polar')
circle = plt.Circle((0.0, 0.0), (10*u.Rsun).to_value(u.AU),
                    transform=ax.transProjectionAffine + ax.transAxes, color="yellow",
                    alpha=1, label="Sun")
ax.add_artist(circle)
ax.text(earth.lon.to_value("rad")+0.05, earth.radius.to_value(u.AU), "Earth")

for this_satellite, this_coord in [(m.observatory, m.observer_coordinate) for m in maps]:
    ax.plot(this_coord.lon.to('rad'), this_coord.radius.to(u.AU), 'o', label=this_satellite)

ax.set_theta_zero_location("S")
ax.set_rlim(0, 1.3)
ax.legend()

plt.show()

######################################################################
# The next step is to calculate the output coordinate system for the combined
# map. We select a heliographic Stonyhurst frame, and a Plate Carree (CAR)
# projection, and generate a header using `sunpy.map.header_helper.make_fitswcs_header` and
# then construct a World Coordinate System (WCS) object for that header.

shape_out = (180, 360)  # This is set deliberately low to reduce memory consumption
header = sunpy.map.make_fitswcs_header(shape_out,
                                       SkyCoord(0, 0, unit=u.deg,
                                                frame="heliographic_stonyhurst",
                                                obstime=maps[0].date),
                                       scale=[360 / shape_out[1],
                                              180 / shape_out[0]] * u.deg / u.pix,
                                       wavelength=int(maps[0].meta['wavelnth']) * u.AA,
                                       projection_code="CAR")
out_wcs = WCS(header)

######################################################################
# Next we call the `reproject.mosaicking.reproject_and_coadd` function, which
# takes a list of maps, and the desired output WCS and array shape.

array, footprint = reproject_and_coadd(maps, out_wcs, shape_out,
                                       reproject_function=reproject_interp)

######################################################################
# To display the output we construct a new map using the new array and our
# generated header. We also borrow the plot settings from the AIA map.

outmap = sunpy.map.Map((array, header))
outmap.plot_settings = maps[0].plot_settings

plt.figure()
outmap.plot()

plt.show()

######################################################################
# Improving the Output
# --------------------
#
# As you can see this leaves a little to be desired. To reduce the obvious
# warping towards the points which are close to the limb in the input
# images, we can define a set of weights to use when co-adding the output
# arrays. To reduce this warping we want to calculate an set of weights
# which highly weigh points close to the centre of the disk in the input
# image.
#
# We can achieve this by using sunpy's coordinate framework. First we
# calculate all the world coordinates for all the pixels in all three
# input maps.

coordinates = tuple(map(sunpy.map.all_coordinates_from_map, maps))

######################################################################
# To get a weighting which is high close to disk centre and low towards
# the limb, we can use the Z coordinate in the heliocentric frame. This
# coordinate is the distance of the sphere from the centre of the Sun
# towards the observer.

weights = [coord.transform_to("heliocentric").z.value for coord in coordinates]

######################################################################
# These weights are good, but they are better if the ramp down is a little
# smoother, and more biased to the centre. Also we can scale them to the
# range 0-1, and set any off disk (NaN) regions to 0.

weights = [(w / np.nanmax(w)) ** 3 for w in weights]
for w in weights:
    w[np.isnan(w)] = 0

plt.figure()
plt.imshow(weights[0])
plt.colorbar()

plt.show()

######################################################################
# Now we can rerun the reprojection. This time we also set
# ``match_background=True`` which scales the images by a single scaling
# factor so they are of similar brightness. We also set
# ``background_reference=0`` which uses the AIA map as the reference for
# the background scaling.
#
# Here we are using the fastest but least accurate method of reprojection,
# `reproject.reproject_interp`, a more accurate but slower method is
# `reproject.reproject_adaptive`.

array, _ = reproject_and_coadd(maps, out_wcs, shape_out,
                               input_weights=weights,
                               reproject_function=reproject_interp,
                               match_background=True,
                               background_reference=0)

######################################################################
# Once again we create a new map, and this time we customise the plot a
# little.

outmap = sunpy.map.Map((array, header))
outmap.plot_settings = maps[0].plot_settings
outmap.nickname = 'AIA + EUVI/A + EUVI/B'

fig = plt.figure(figsize=(10, 5))
ax = fig.add_subplot(projection=out_wcs)
im = outmap.plot(axes=ax, vmin=400)

lon, lat = ax.coords
lon.set_coord_type("longitude")
lon.coord_wrap = 180
lon.set_format_unit(u.deg)
lat.set_coord_type("latitude")
lat.set_format_unit(u.deg)

lon.set_axislabel('Heliographic Longitude', minpad=0.8)
lat.set_axislabel('Heliographic Latitude', minpad=0.9)
lon.set_ticks(spacing=25*u.deg, color='k')
lat.set_ticks(spacing=15*u.deg, color='k')

plt.colorbar(im, ax=ax)

# Reset the view to pixel centers
_ = ax.axis((0, shape_out[1], 0, shape_out[0]))

plt.show()