import numpy as np
import scipy.spatial
import crl.color as color
import crl.tables as tables

class CRIError(Exception):
	pass

class CRIResult:
	def __init__(self, method_name):
		self.method_name = method_name
		self.SCRIs = []
		self.tcs = []
		self.tcs_ref = []

	def __str__(self):
		ret = f"CRI result for {self.method_name}\n"
		ret += f"\tXYZ = {self.XYZ}\n"
		ret += f"\tCCT = {self.CCT} K (delta uv = {self.duv})\n"
		SCRIs = ", ".join("%.1f" % SCRI for SCRI in self.SCRIs)
		ret += f"\tSCRIs = [{SCRIs}]\n"
		ret += f"\tCRI = {self.CRI}"
		return ret

def CAT(sample, ref, test):
	def c(u, v, _):
		return (4 - u - 10 * v) / v
	def d(u, v, _):
		return (1.708 * v - 1.481 * u + 0.404) / v

	cs, ds = c(*sample), d(*sample)
	cr, dr = c(*ref), d(*ref)
	ct, dt = c(*test), d(*test)

	u = (10.872 + 0.404 * cr / ct * cs - 4 * dr / dt * ds) \
	    / (16.518 + 1.481 * cr / ct * cs - dr / dt * ds)
	v = 5.52 / (16.518 + 1.481 * cr / ct * cs - dr / dt * ds)

	return color.uvY(u, v, sample[2])

def cri_Ra(S, observer=color.Observer_2deg, ignore_errors=False):
	res = CRIResult("CRI Ra")
	res.S = S.normed()
	res.XYZ = color.XYZ(res.S, observer=observer)
	res.CCT, res.duv = color.cct(res.XYZ)

	if res.duv > 5e-2 and not ignore_errors:
		raise CRIError(f"spectrum not white enough with delta uv = {res.duv}")

	if res.CCT < 5000:
		white = color.BlackBodySpectrum(res.CCT).normed()
	else:
		if res.CCT > 25000 and ignore_errors:
			res.CCT = 25000
		white = color.CIEDaylightSpectrum(res.CCT).normed()

	res.XYZ_ref = color.XYZ(white, observer=observer)

	class UVWstarRef(color.UVWstar):
		white_uvY = color.uvY(white, observer=observer)

	class UVWstar(color.UVWstar):
		white_uvY = color.uvY(S, observer=observer)

	for i in range(14):
		tcs = color.Spectrum(tables.tcs_ra[:, 0], tables.tcs_ra[:, 1 + i])

		lit = color.uvY(color.CombinedSpectrum(tcs, res.S))
		lit = CAT(lit, UVWstarRef.white_uvY, UVWstar.white_uvY)
		lit = UVWstarRef(lit)
		lit_ref = UVWstarRef(color.CombinedSpectrum(tcs, white), observer=observer)

		res.tcs.append(lit)
		res.tcs_ref.append(lit_ref)

		dE = np.linalg.norm(lit.array - lit_ref.array)
		res.SCRIs.append(100 - 4.6 * dE)

	res.CRI = np.mean(res.SCRIs[:8])
	res.ECRI = np.mean(res.SCRIs)
	return res

def cri_GAI(spec, observer=color.Observer_2deg):
	points = []
	for i in range(8):
		tcs = color.Spectrum(color.tables.tcs_ra[:, 0], color.tables.tcs_ra[:, 1 + i])
		u, v, _ = color.uvY76(color.CombinedSpectrum(tcs, spec), observer=observer)
		points.append([u, v])
	points = np.array(points)
	hull = scipy.spatial.ConvexHull(points)

	# The magic number is the hull.volume of the E illuminant
	GAI = hull.volume * 100 / 0.007351716795637625

	return GAI, hull.points[hull.vertices]

def cri_FSI(spec):
	wvl = np.arange(380, 730, 1)
	Y = spec.Yi(wvl)
	dwvl = wvl[1] - wvl[0]
	Y = spec.Yi(wvl)
	Y /= np.sum(Y * dwvl)

	CEE = (wvl - 380) / (730 - 380)

	Is = []
	for i, _ in enumerate(wvl):
		C = np.cumsum(Y * dwvl)
		D = (C - CEE) ** 2
		Is.append(sum(D * dwvl))
		Y = np.roll(Y, 1)

	FSI = np.mean(Is)
	return FSI

def cri_FSCI(spec):
	return 100 - 5.1 * cri_FSI(spec)