pyzebra/pyzebra/app/plot_hkl.py

392 lines
15 KiB
Python

import base64
import io
import os
import numpy as np
from bokeh.layouts import column, row
from bokeh.models import (
Arrow,
Button,
CheckboxGroup,
ColumnDataSource,
Div,
FileInput,
Legend,
LegendItem,
NormalHead,
NumericInput,
RadioGroup,
Range1d,
Spacer,
Spinner,
TextAreaInput,
TextInput,
)
from bokeh.palettes import Dark2
from bokeh.plotting import figure
from scipy.integrate import simpson, trapezoid
import pyzebra
class PlotHKL:
def __init__(self):
measured_data_div = Div(text="Measured data:")
measured_data = FileInput(accept=".ccl", multiple=True, width=200)
min_grid_x = -10
max_grid_x = 10
min_grid_y = -5
max_grid_y = 5
cmap = Dark2[8]
syms = ["circle", "inverted_triangle", "square", "diamond", "star", "triangle"]
def plot_file_callback():
orth_dir = list(map(float, hkl_normal.value.split()))
cut_tol = hkl_delta.value
cut_or = hkl_cut.value
x_dir = list(map(float, hkl_in_plane_x.value.split()))
y_dir = list(map(float, hkl_in_plane_y.value.split()))
k = np.array(k_vectors.value.split()).astype(float).reshape(-1, 3)
tol_k = tol_k_ni.value
# Plotting options
grid_flag = 1
grid_minor_flag = 1
grid_div = 2 # Number of minor division lines per unit
# different symbols based on file number
file_flag = 0 in disting_opt_cb.active
# scale marker size according to intensity
intensity_flag = 1 in disting_opt_cb.active
# use color to mark different propagation vectors
prop_legend_flag = 2 in disting_opt_cb.active
# use resolution ellipsis
res_flag = disting_opt_rb.active
# multiplier for resolution function (in case of samples with large mosaicity)
res_mult = res_mult_ni.value
md_fnames = measured_data.filename
md_fdata = measured_data.value
# Load first data file, read angles and define matrices to perform conversion to cartesian
# coordinates and back
with io.StringIO(base64.b64decode(md_fdata[0]).decode()) as file:
_, ext = os.path.splitext(md_fnames[0])
try:
file_data = pyzebra.parse_1D(file, ext)
except:
print(f"Error loading {md_fnames[0]}")
return
alpha = file_data[0]["alpha_cell"] * np.pi / 180.0
beta = file_data[0]["beta_cell"] * np.pi / 180.0
gamma = file_data[0]["gamma_cell"] * np.pi / 180.0
# reciprocal angle parameters
beta_star = np.arccos(
(np.cos(alpha) * np.cos(gamma) - np.cos(beta)) / (np.sin(alpha) * np.sin(gamma))
)
gamma_star = np.arccos(
(np.cos(alpha) * np.cos(beta) - np.cos(gamma)) / (np.sin(alpha) * np.sin(beta))
)
# conversion matrix
M = np.array(
[
[1, np.cos(gamma_star), np.cos(beta_star)],
[0, np.sin(gamma_star), -np.sin(beta_star) * np.cos(alpha)],
[0, 0, np.sin(beta_star) * np.sin(alpha)],
]
)
# Calculate in-plane y-direction
x_c = M @ x_dir
y_c = M @ y_dir
o_c = M @ orth_dir
# Normalize all directions
y_c = y_c / np.linalg.norm(y_c)
x_c = x_c / np.linalg.norm(x_c)
o_c = o_c / np.linalg.norm(o_c)
# Read all data
hkl_coord = []
intensity_vec = []
k_flag_vec = []
file_flag_vec = []
res_vec_x = []
res_vec_y = []
res_N = 10
for j, md_fname in enumerate(md_fnames):
with io.StringIO(base64.b64decode(md_fdata[j]).decode()) as file:
_, ext = os.path.splitext(md_fname)
try:
file_data = pyzebra.parse_1D(file, ext)
except:
print(f"Error loading {md_fname}")
return
pyzebra.normalize_dataset(file_data)
# Loop throguh all data
for scan in file_data:
om = scan["omega"]
gammad = scan["twotheta"]
chi = scan["chi"]
phi = scan["phi"]
nud = 0 # 1d detector
ub_inv = np.linalg.inv(scan["ub"])
counts = scan["counts"]
wave = scan["wavelength"]
# Calculate resolution in degrees
expr = np.tan(gammad / 2 * np.pi / 180)
res = (
np.sqrt(0.4639 * expr**2 - 0.4452 * expr + 0.1506) * res_mult
) # res in deg
# convert to resolution in hkl along scan line
ang2hkl_1d = pyzebra.ang2hkl_1d
res_x = []
res_y = []
for _om in np.linspace(om[0], om[-1], num=res_N):
expr1 = ang2hkl_1d(wave, gammad, _om + res / 2, chi, phi, nud, ub_inv)
expr2 = ang2hkl_1d(wave, gammad, _om - res / 2, chi, phi, nud, ub_inv)
hkl_temp = M @ (np.abs(expr1 - expr2) / 2)
res_x.append(hkl_temp[0])
res_y.append(hkl_temp[1])
# Get first and final hkl
hkl1 = ang2hkl_1d(wave, gammad, om[0], chi, phi, nud, ub_inv)
hkl2 = ang2hkl_1d(wave, gammad, om[-1], chi, phi, nud, ub_inv)
# Get hkl at best intensity
hkl_m = ang2hkl_1d(wave, gammad, om[np.argmax(counts)], chi, phi, nud, ub_inv)
# Estimate intensity for marker size scaling
y_bkg = [counts[0], counts[-1]]
x_bkg = [om[0], om[-1]]
c = int(simpson(counts, x=om) - trapezoid(y_bkg, x=x_bkg))
# Recognize k_flag_vec
reduced_hkl_m = np.minimum(1 - hkl_m % 1, hkl_m % 1)
for ind, _k in enumerate(k):
if all(np.abs(reduced_hkl_m - _k) < tol_k):
k_flag_vec.append(ind)
break
else:
# not required
continue
# Save data
hkl_coord.append([hkl1, hkl2, hkl_m])
intensity_vec.append(c)
file_flag_vec.append(j)
res_vec_x.append(res_x)
res_vec_y.append(res_y)
plot.x_range.start = plot.x_range.reset_start = -2
plot.x_range.end = plot.x_range.reset_end = 5
plot.y_range.start = plot.y_range.reset_start = -4
plot.y_range.end = plot.y_range.reset_end = 3.5
# Plot grid lines
xs, ys = [], []
xs_minor, ys_minor = [], []
if grid_flag:
for yy in np.arange(min_grid_y, max_grid_y, 1):
hkl1 = M @ [0, yy, 0]
xs.append([min_grid_y, max_grid_y])
ys.append([hkl1[1], hkl1[1]])
for xx in np.arange(min_grid_x, max_grid_x, 1):
hkl1 = M @ [xx, min_grid_x, 0]
hkl2 = M @ [xx, max_grid_x, 0]
xs.append([hkl1[0], hkl2[0]])
ys.append([hkl1[1], hkl2[1]])
if grid_minor_flag:
for yy in np.arange(min_grid_y, max_grid_y, 1 / grid_div):
hkl1 = M @ [0, yy, 0]
xs_minor.append([min_grid_y, max_grid_y])
ys_minor.append([hkl1[1], hkl1[1]])
for xx in np.arange(min_grid_x, max_grid_x, 1 / grid_div):
hkl1 = M @ [xx, min_grid_x, 0]
hkl2 = M @ [xx, max_grid_x, 0]
xs_minor.append([hkl1[0], hkl2[0]])
ys_minor.append([hkl1[1], hkl2[1]])
grid_source.data.update(xs=xs, ys=ys)
minor_grid_source.data.update(xs=xs_minor, ys=ys_minor)
el_x, el_y, el_w, el_h, el_c = [], [], [], [], []
scan_xs, scan_ys, scan_x, scan_y = [], [], [], []
scan_m, scan_s, scan_c, scan_l = [], [], [], []
for j in range(len(hkl_coord)):
# Get middle hkl from list
hklm = M @ hkl_coord[j][2]
# Decide if point is in the cut
proj = np.dot(hklm, o_c)
if abs(proj - cut_or) >= cut_tol:
continue
hkl1 = M @ hkl_coord[j][0]
hkl2 = M @ hkl_coord[j][1]
if intensity_flag:
markersize = max(1, int(intensity_vec[j] / max(intensity_vec) * 20))
else:
markersize = 4
if file_flag:
plot_symbol = syms[file_flag_vec[j]]
else:
plot_symbol = "circle"
if prop_legend_flag:
col_value = cmap[k_flag_vec[j]]
else:
col_value = "black"
if res_flag:
# Generate series of ellipses along scan line
el_x.extend(np.linspace(hkl1[0], hkl2[0], num=res_N))
el_y.extend(np.linspace(hkl1[1], hkl2[1], num=res_N))
el_w.extend(np.array(res_vec_x[j]) * 2)
el_h.extend(np.array(res_vec_y[j]) * 2)
el_c.extend([col_value] * res_N)
else:
# Plot scan line
scan_xs.append([hkl1[0], hkl2[0]])
scan_ys.append([hkl1[1], hkl2[1]])
# Plot middle point of scan
scan_x.append(hklm[0])
scan_y.append(hklm[1])
scan_m.append(plot_symbol)
scan_s.append(markersize)
# Color and legend label
scan_c.append(col_value)
scan_l.append(md_fnames[file_flag_vec[j]])
ellipse_source.data.update(x=el_x, y=el_y, width=el_w, height=el_h, c=el_c)
scan_source.data.update(
xs=scan_xs, ys=scan_ys, x=scan_x, y=scan_y, m=scan_m, s=scan_s, c=scan_c, l=scan_l
)
arrow1.visible = True
arrow1.x_end = x_c[0]
arrow1.y_end = x_c[1]
arrow2.visible = True
arrow2.x_end = y_c[0]
arrow2.y_end = y_c[1]
kvect_source.data.update(
x=[x_c[0] / 2, y_c[0] / 2 - 0.1], y=[x_c[1] - 0.1, y_c[1] / 2], text=["h", "k"]
)
# Legend items for different file entries (symbol)
legend_items = []
if not res_flag and file_flag:
labels, inds = np.unique(scan_source.data["l"], return_index=True)
for label, ind in zip(labels, inds):
legend_items.append(LegendItem(label=label, renderers=[scatter], index=ind))
# Legend items for propagation vector (color)
if prop_legend_flag:
if res_flag:
source, render = ellipse_source, ellipse
else:
source, render = scan_source, mline
labels, inds = np.unique(source.data["c"], return_index=True)
for label, ind in zip(labels, inds):
label = f"k={k[cmap.index(label)]}"
legend_items.append(LegendItem(label=label, renderers=[render], index=ind))
plot.legend.items = legend_items
plot_file = Button(label="Plot selected file(s)", button_type="primary", width=200)
plot_file.on_click(plot_file_callback)
plot = figure(
x_range=Range1d(),
y_range=Range1d(),
plot_height=450,
plot_width=450 + 32,
tools="pan,wheel_zoom,reset",
)
plot.toolbar.logo = None
arrow1 = Arrow(
x_start=0, y_start=0, x_end=0, y_end=0, end=NormalHead(size=10), visible=False
)
plot.add_layout(arrow1)
arrow2 = Arrow(
x_start=0, y_start=0, x_end=0, y_end=0, end=NormalHead(size=10), visible=False
)
plot.add_layout(arrow2)
kvect_source = ColumnDataSource(dict(x=[], y=[], text=[]))
plot.text(source=kvect_source)
grid_source = ColumnDataSource(dict(xs=[], ys=[]))
plot.multi_line(source=grid_source, line_color="gray")
minor_grid_source = ColumnDataSource(dict(xs=[], ys=[]))
plot.multi_line(source=minor_grid_source, line_color="gray", line_dash="dotted")
ellipse_source = ColumnDataSource(dict(x=[], y=[], width=[], height=[], c=[]))
ellipse = plot.ellipse(source=ellipse_source, fill_color="c", line_color="c")
scan_source = ColumnDataSource(dict(xs=[], ys=[], x=[], y=[], m=[], s=[], c=[], l=[]))
mline = plot.multi_line(source=scan_source, line_color="c")
scatter = plot.scatter(
source=scan_source, marker="m", size="s", fill_color="c", line_color="c"
)
plot.add_layout(Legend(items=[], location="top_left", click_policy="hide"))
hkl_div = Div(text="HKL:", margin=(5, 5, 0, 5))
hkl_normal = TextInput(title="normal", value="0 0 1", width=70)
hkl_cut = Spinner(title="cut", value=0, step=0.1, width=70)
hkl_delta = NumericInput(title="delta", value=0.1, mode="float", width=70)
hkl_in_plane_x = TextInput(title="in-plane X", value="1 0 0", width=70)
hkl_in_plane_y = TextInput(title="in-plane Y", value="0 1 0", width=70)
disting_opt_div = Div(text="Distinguish options:", margin=(5, 5, 0, 5))
disting_opt_cb = CheckboxGroup(
labels=["files (symbols)", "intensities (size)", "k vectors nucl/magn (colors)"],
active=[0, 1, 2],
width=200,
)
disting_opt_rb = RadioGroup(
labels=["scan direction", "resolution ellipsoid"], active=0, width=200
)
k_vectors = TextAreaInput(
title="k vectors:", value="0.0 0.0 0.0\n0.5 0.0 0.0\n0.5 0.5 0.0", width=150
)
res_mult_ni = NumericInput(title="Resolution mult:", value=10, mode="int", width=100)
tol_k_ni = NumericInput(title="k tolerance:", value=0.01, mode="float", width=100)
hkl_layout = column(
hkl_div,
row(hkl_normal, hkl_cut, hkl_delta, Spacer(width=10), hkl_in_plane_x, hkl_in_plane_y),
)
disting_layout = column(disting_opt_div, row(disting_opt_cb, disting_opt_rb))
layout = column(
row(measured_data_div, measured_data, plot_file),
plot,
row(hkl_layout, k_vectors),
row(disting_layout, column(tol_k_ni, res_mult_ni)),
)
self.layout = layout