Files
crystfel_tools/reduction_tools/stream_select_res.py
2025-05-08 15:38:23 +02:00

329 lines
10 KiB
Python

#!/usr/bin/env python3
# author J.Beale
"""
# aim
analyses and selects crystals based on their reported resolution from crystfel
# usage
python stream_random.py -s <path to stream>
-o output file names
-p plots histogram of images resolution
-r selects all images with higher resolution than value
# output
either
- histogram of images by resolution
- .stream file of selected images
"""
# modules
import re
import argparse
import pandas as pd
import numpy as np
import os
import matplotlib.pyplot as plt
def extract_chunks( input_file ):
# setup
chunk_df = pd.DataFrame()
image_no = []
chunks = []
hits = []
collect_lines = False
# Open the input file for reading
with open(input_file, 'r') as f:
for line in f:
# Check for the start condition
if line.startswith('----- Begin chunk -----'):
hit = False
collect_lines = True
chunk_lines = []
if collect_lines:
chunk_lines.append(line)
# find image_no
if line.startswith( "Event:" ):
image_search = re.findall( r"Event: //(\d+)", line )
image = int(image_search[0])
image_no.append( image )
# is there a hit in chunk
if line.startswith( "Cell parameters" ):
hit = True
if line.startswith('----- End chunk -----'):
collect_lines = False # Stop collecting lines
chunks.append( chunk_lines )
hits.append( hit )
chunk_df[ "chunks" ] = chunks
chunk_df[ "image_no" ] = image_no
chunk_df[ "hit" ] = hits
return chunk_df
def scrub_res( line ):
# get resolution
try:
pattern = r"diffraction_resolution_limit\s=\s\d\.\d+\snm\^-1\sor\s(\d+\.\d+)\sA"
res = re.search( pattern, line ).group(1)
except AttributeError as e:
res = np.nan
return float( res )
def extract_xtals( chunk ):
# setup
xtals = []
resolutions = []
collect_crystal_lines = False
# Open the input file for reading
for line in chunk:
# Check for the xtals start condition
if line.startswith('--- Begin crystal'):
collect_crystal_lines = True
xtal_lines = []
if collect_crystal_lines:
xtal_lines.append(line)
if line.startswith('--- End crystal\n'):
collect_crystal_lines = False # Stop collecting lines
xtals.append( xtal_lines )
if line.startswith( "diffraction_resolution_limit" ):
res = scrub_res( line )
resolutions.append( res )
return xtals, resolutions
def extract_header( chunk ):
# setup
header = []
collect_header_lines = False
# Open the input file for reading
for line in chunk:
# Check for the xtals start condition
if line.startswith('----- Begin chunk -----'):
collect_header_lines = True
header_lines = []
if collect_header_lines:
header_lines.append(line)
if line.startswith('End of peak list'):
collect_header_lines = False # Stop collecting lines
header.append( header_lines )
return header
def get_header( header, input_file ):
if header == "geom":
start_keyword = "----- Begin geometry file -----"
end_keyword = "----- End geometry file -----"
if header == "cell":
start_keyword = "----- Begin unit cell -----"
end_keyword = "----- End unit cell -----"
# setup
collect_lines = False
headers = []
# Open the input file for reading
with open(input_file, 'r') as f:
for line in f:
# Check for the start condition
if line.strip() == start_keyword:
collect_lines = True
headers_lines = []
# Collect lines between start and end conditions
if collect_lines:
headers_lines.append(line)
# Check for the end condition
if line.strip() == end_keyword:
collect_lines = False # Stop collecting lines
headers.append(headers_lines)
return headers[0]
def write_to_file( geom, cell, chunk_header, crystals, output_file ):
# Write sections with matching cell parameters to the output file
with open(output_file, 'w') as out_file:
out_file.write('CrystFEL stream format 2.3\n')
out_file.write('Generated by CrystFEL 0.10.2\n')
out_file.writelines(geom)
out_file.writelines(cell)
for crystal, header in zip( crystals, chunk_header ):
out_file.writelines( header )
out_file.writelines( crystal )
out_file.writelines( "----- End chunk -----\n" )
def sort_xtals( chunk_df ):
# extract xtals
xtal_df = pd.DataFrame()
counter = 0
for index, row in chunk_df.iterrows():
chunk, hit, image_no = row[ "chunks" ], row[ "hit" ], row[ "image_no" ]
if hit:
# find xtals and header
header = extract_header( chunk )
xtals, resolutions = extract_xtals( chunk )
# make header same length as xtals
header = header*len(xtals)
# concat results
xtal_df_1 = pd.DataFrame()
xtal_df_1[ "header" ] = header
xtal_df_1[ "xtals" ] = xtals
xtal_df_1[ "image_no" ] = image_no
xtal_df_1[ "resolution" ] = resolutions
xtal_df = pd.concat( ( xtal_df, xtal_df_1 ) )
# add count and print every 1000s
counter = counter + len(xtals)
if counter % 1000 == 0:
print( counter, end='\r' )
print( "done" )
# sort by image no and reindex
xtal_df = xtal_df.sort_values( by=[ "image_no" ] )
xtal_df = xtal_df.reset_index( drop=True )
return xtal_df
def plot_res_histogram( res_df, res_median, res_q1, res_q3 ):
# calculate relative numbers of bins
bin_range = 0.25
res_min = res_df.min().values[0]
res_max = res_df.max().values[0]
q1_bins = round( ( res_q1 - res_min )/bin_range )
q2_bins = round( ( res_median - res_q1 )/bin_range )
q3_bins = round( ( res_q3 - res_median )/bin_range )
q4_bins = round( ( res_max - res_q3 )/bin_range )
# cut data by quantile
df_q1 = res_df[ res_df.resolution <= res_q1 ]
df_q2 = res_df[ ( res_df.resolution > res_q1 ) & ( res_df.resolution <= res_median ) ]
df_q3 = res_df[ ( res_df.resolution > res_median ) & ( res_df.resolution <= res_q3 ) ]
df_q4 = res_df[ res_df.resolution > res_q3 ]
# plot histogram of resolution
fig, axs = plt.subplots()
axs.hist( df_q1, bins=q1_bins, rwidth=1, stacked=True, color="blue", label="x<q1" )
axs.hist( df_q2, bins=q2_bins, rwidth=1, stacked=True, color="red", label="q1<x<q2")
axs.hist( df_q3, bins=q3_bins, rwidth=1, stacked=True, color="green", label="q2<x<q3" )
axs.hist( df_q4, bins=q4_bins, rwidth=1, stacked=True, color="purple", label="q3<x<q4" )
axs.axvline( x=res_median, color="black", linestyle="dashed", label="median = {0}".format( res_median ) )
axs.set_xlabel( "resolution" )
axs.set_ylabel( "frequency" )
axs.legend()
bin_size = round( ( len( df_q1 ) + len( df_q2 ) + len( df_q3 ) + len( df_q4 ) )/4 )
axs.text( 0.98, 0.7,
"images per quartile = {0}".format( bin_size ),
ha="right",
va="top",
transform=axs.transAxes
)
fig.tight_layout()
plt.show()
def main( input_file, output, plotter, resolution ):
# get geom and cell file headers
print( "getting header info from .stream file" )
geom = get_header( "geom", input_file )
cell = get_header( "cell", input_file )
print( "done" )
# extract chunks
print( "finding chucks" )
chunk_df = extract_chunks( input_file )
# display no. of chunks
print( "found {0} chunks".format( len(chunk_df) ) )
# remove rows without xtals
chunk_df = chunk_df.loc[chunk_df.hit, :]
print( "found {0} hits (not including multiples)".format( len(chunk_df) ) )
print( "done" )
print( "sorting xtals from chunks" )
xtal_df = sort_xtals( chunk_df )
print( "done" )
print( "calculate stats" )
res_df = xtal_df[ [ "resolution" ] ]
res_median = res_df.median().values[0]
res_q1 = res_df.quantile( 0.25 ).values[0]
res_q3 = res_df.quantile( 0.75 ).values[0]
print( "median resolution and range = {0} ({1}-{2})".format( res_median, res_q1, res_q3 ) )
print( "done" )
if plotter == True:
print( "plot image resolution histogram" )
plot_res_histogram( res_df, res_median, res_q1, res_q3 )
if resolution:
print( "finding images with resolution greater than {0}".format( resolution ) )
select_df = xtal_df[ xtal_df[ "resolution" ] <= resolution ]
print( "done" )
print( "writing {0} to output file".format( len( select_df ) ) )
crystals = select_df.xtals.to_list()
chunk_header = select_df.header.to_list()
output_file = "{0}.stream".format( output )
write_to_file( geom, cell, chunk_header, crystals, output_file )
print( "done" )
else:
print( "no output. please use either the plot-histogram or resolution functions of the script" )
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument(
"-s",
"--stream",
help="input stream file",
required=True,
type=os.path.abspath
)
parser.add_argument(
"-o",
"--output",
help="output stream file name. '.stream will be added'",
type=str,
default="selected"
)
parser.add_argument(
"-p",
"--plot_histogram",
help="plots a histogram of the crystfel calculated resolutions for inspection",
type=bool,
default=False
)
parser.add_argument(
"-r",
"--resolution",
help="upper resolution limit. Will take all images below this.",
type=float
)
args = parser.parse_args()
# run main
main( args.stream, args.output, args.plot_histogram, args.resolution )