QSpaceConverter.py 55.8 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
# coding: utf-8
# /*##########################################################################
#
# Copyright (c) 2015-2016 European Synchrotron Radiation Facility
#
# Permission is hereby granted, free of charge, to any person obtaining a copy
# of this software and associated documentation files (the "Software"), to deal
# in the Software without restriction, including without limitation the rights
# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
# copies of the Software, and to permit persons to whom the Software is
# furnished to do so, subject to the following conditions:
#
# The above copyright notice and this permission notice shall be included in
# all copies or substantial portions of the Software.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
# THE SOFTWARE.
#
# ###########################################################################*/

26 27
from __future__ import absolute_import

28 29 30 31 32
__authors__ = ["D. Naudet"]
__license__ = "MIT"
__date__ = "01/03/2016"


33
import logging
34 35 36 37 38 39 40 41 42 43
import os
import time
import ctypes
from threading import Thread
import multiprocessing as mp
import multiprocessing.sharedctypes as mp_sharedctypes

import numpy as np
import xrayutilities as xu

44
from ...util.medianfilter import medfilt2d
45 46
from ...util.histogramnd_lut import histogramnd_get_lut, histogramnd_from_lut
# from silx.math import histogramnd
Damien Naudet's avatar
Damien Naudet committed
47
from ...io import XsocsH5, QSpaceH5, ShiftH5
48

49 50 51 52

logger = logging.getLogger(__name__)


53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78
disp_times = False


class QSpaceConverter(object):
    (READY, RUNNING, DONE,
     ERROR, CANCELED, UNKNOWN) = __STATUSES = range(6)
    """ Available status codes """

    status = property(lambda self: self.__status)
    """ Current status code of this instance """

    status_msg = property(lambda self: self.__status_msg)
    """ Status message if any, or None """

    results = property(lambda self: self.__results)
    """ Parse results. KmapParseResults instance. """

    xsocsH5_f = property(lambda self: self.__xsocsH5_f)
    """ Input file name. """

    output_f = property(lambda self: self.__output_f)
    """ Output file name. """

    qspace_dims = property(lambda self: self.__params['qspace_dims'])
    """ dimensions of the Q Space (i.e : number of bins). """

Damien Naudet's avatar
Damien Naudet committed
79
    medfilt_dims = property(lambda self: self.__params['medfilt_dims'])
80 81
    """ Median filter applied to the images after masking and
    before conversion. """
Damien Naudet's avatar
Damien Naudet committed
82

83 84 85 86 87 88 89 90 91
    beam_energy = property(lambda self: self.__params['beam_energy'])
    """Beam energy or None to read it from entries"""

    direct_beam = property(lambda self: self.__params['direct_beam'])
    """Direct beam calibration position or None to read it from entries"""

    channels_per_degree = property(lambda self: self.__params['channels_per_degree'])
    """Channels per degree calibration or None to read it from entries"""

92 93 94 95 96 97
    sample_indices = property(lambda self: self.__params['sample_indices'])
    """ Indices of sample positions that will be converted. """

    n_proc = property(lambda self: self.__n_proc)
    """ Number of processes to use. Will use cpu_count() if None or 0. """

Damien Naudet's avatar
Damien Naudet committed
98 99
    roi = property(lambda self: self.__params['roi'])
    """ Selected ROI in sample coordinates : [xmin, xmax, ymin, ymax] """
100

101 102 103
    normalizer = property(lambda self: self.__params['normalizer'])
    """ Selected normalizer name in measurement group (str) or None """

104 105 106 107 108 109
    mask = property(lambda self: self.__params['mask'])
    """ Mask to apply on images (2D numpy.ndarray) or None.

    A non-zero value means that the pixel is masked.
    """

110 111 112
    def __init__(self,
                 xsocsH5_f,
                 qspace_dims=None,
Damien Naudet's avatar
Damien Naudet committed
113
                 medfilt_dims=None,
114
                 output_f=None,
Damien Naudet's avatar
Damien Naudet committed
115
                 roi=None,
116
                 entries=None,
Damien Naudet's avatar
Damien Naudet committed
117 118
                 callback=None,
                 shiftH5_f=None):
119 120 121 122 123
        """
        Merger for the Kmap SPEC and EDF files. This loads a spech5 file,
             converts it to HDF5 and then tries to match scans and edf image
             files.
        :param xsocsH5_f: path to the input XsocsH5 file.
124
        :param qspace_dims: dimensions of the qspace volume
Damien Naudet's avatar
Damien Naudet committed
125 126 127
        :param medfilt_dims: dimensions of the median filter kernel
            to apply to the images before conversion. The filter is not
            applied if this keyword is set to None (this is the default).
128
            The median filter is applied after the mask (if any).
129
        :param output_f: path to the output file that will be created.
130 131 132
        :param roi: Roi in sample coordinates (xMin, xMax, yMin, yMax)
        :param entries: a list of entry names to convert to qspace. If None,
            all entries found in the xsocsH5File will be used.
133
        :param callback: callback to call when the parsing is done.
Damien Naudet's avatar
Damien Naudet committed
134
        :param shiftH5_f: a ShiftH5 file name to use if applying shift.
135 136 137 138 139 140 141 142 143 144
        """
        super(QSpaceConverter, self).__init__()

        self.__status = None

        self.__set_status(self.UNKNOWN, 'Init')

        self.__xsocsH5_f = xsocsH5_f
        self.__output_f = output_f

Damien Naudet's avatar
Damien Naudet committed
145 146 147 148 149 150 151
        if shiftH5_f:
            shiftH5 = ShiftH5.ShiftH5(shiftH5_f)
        else:
            shiftH5 = None

        self.__shiftH5 = shiftH5

152 153 154 155 156 157 158 159 160 161 162
        xsocsH5 = XsocsH5.XsocsH5(xsocsH5_f)
        # checking entries
        if entries is None:
            entries = xsocsH5.entries()
        else:
            diff = set(entries) - set(xsocsH5.entries())
            if len(diff) > 0:
                raise ValueError('The following entries were not found in '
                                 'the input file :\n - {0}'
                                 ''.format('\n -'.join(diff)))

163
        self.__params = {'qspace_dims': None,
164
                         'mask': None,
165
                         'normalizer': None,
166
                         'sample_indices': None,
167
                         'roi': None,
168 169 170 171
                         'entries': sorted(entries),
                         'beam_energy': None,
                         'direct_beam': None,
                         'channels_per_degree': None}
172

173 174 175
        self.maxipix_correction = False
        """Whether or not to apply Maxipix module edges correction (bool)"""

176 177 178 179 180 181 182 183 184 185
        self.__callback = callback
        self.__n_proc = None
        self.__overwrite = False

        self.__shared_progress = None
        self.__results = None
        self.__term_evt = None

        self.__thread = None

Damien Naudet's avatar
Damien Naudet committed
186
        self.medfilt_dims = medfilt_dims
187
        self.qspace_dims = qspace_dims
Damien Naudet's avatar
Damien Naudet committed
188
        self.roi = roi
189 190 191 192 193

        self.__set_status(self.READY)

    def __get_scans(self):
        """
194
        Returns the entries that will be converted.
195
        """
196
        return self.__params['entries']
197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241

    scans = property(__get_scans)
    """ Returns the scans found in the input file. """

    def __set_status(self, status, msg=None):
        """
        Sets the status of this instance.
        :param status:
        :param msg:
        :return:
        """
        assert status in self.__STATUSES
        self.__status = status
        self.__status_msg = msg

    def convert(self,
                overwrite=False,
                blocking=True,
                callback=None,
                check_consistency=True):
        """
        Starts the conversion.
        :param overwrite: if False raises an exception if some files already
        exist.
        :param blocking: if False, the merge will be done in a separate
         thread and this method will return immediately.
        :param callback: callback that will be called when the merging is done.
        It overwrites the one passed the constructor.
        :param check_consistency: set to False to ignore any incensitencies
        in the input entries (e.g : different counters, ...).
        :return:
        """

        if self.is_running():
            raise RuntimeError('This QSpaceConverter instance is already '
                               'parsing.')

        self.__set_status(self.RUNNING)

        errors = self.check_parameters()

        if len(errors) > 0:
            msg = 'Invalid parameters.\n{0}'.format('\n'.join(errors))
            raise ValueError(msg)

242 243 244 245
        errors = self.check_consistency(
            beam_energy_check=self.beam_energy is None,
            direct_beam_check=self.direct_beam is None,
            channels_per_degree_check=self.channels_per_degree is None)
246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305

        if len(errors) > 0:
            msg = 'Inconsistent input data.\n{0}'.format('\n'.join(errors))

            if check_consistency:
                raise ValueError(msg)
            else:
                print('==============.')
                print('==============.')
                print('WARNING.')
                print(msg)

        output_f = self.__output_f
        if output_f is None:
            self.__set_status(self.ERROR)
            raise ValueError('Output file name (output_f) has not been set.')

        output_dir = os.path.dirname(output_f)
        if not os.path.exists(output_dir):
            os.makedirs(output_dir)

        if not overwrite:
            if len(self.check_overwrite()):
                    self.__set_status(self.ERROR)
                    raise RuntimeError('Some files already exist. Use the '
                                       'overwrite keyword to ignore this '
                                       'warning.')

        self.__results = None
        self.__overwrite = overwrite

        if callback is not None:
            self.__callback = callback

        if blocking:
            self.__run_convert()
        else:
            thread = self.__thread = Thread(target=self.__run_convert)
            thread.start()

    @qspace_dims.setter
    def qspace_dims(self, qspace_dims):
        """
        Sets the dimensions of the qspace volume (i.e. number of bins).
        """

        if qspace_dims is None or None in qspace_dims:
            self.__params['qspace_dims'] = None
            return

        qspace_dims = np.array(qspace_dims, ndmin=1).astype(np.int32)

        if qspace_dims.ndim != 1 or qspace_dims.size != 3:
            raise ValueError('qspace_dims must be a three elements array.')

        if not np.all(qspace_dims > 1):
            raise ValueError('<qspace_dims> values must be strictly'
                             ' greater than one.')
        self.__params['qspace_dims'] = qspace_dims

306 307 308 309 310 311
    @normalizer.setter
    def normalizer(self, normalizer):
        """Name of dataset in measurement to use for normalization"""
        if normalizer is not None:
             normalizer = str(normalizer)

312
             # Check for valid input in all entries
313 314 315 316 317 318 319
             with XsocsH5.XsocsH5(self.__xsocsH5_f) as xsocsH5:
                 for entry in xsocsH5.entries():
                     if xsocsH5.measurement(
                             entry=entry, measurement=normalizer) is None:
                         raise ValueError(
                             'normalizer %s is not available in measurement group of entry %s' %
                             normalizer, entry)
320 321 322

        self.__params['normalizer'] = normalizer

323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338
    @mask.setter
    def mask(self, mask):
        """Mask array or None to mask pixels in input images"""
        if mask is not None:
            mask = np.array(mask)
            if mask.ndim != 2:
                raise ValueError('Mask is not an image')

            # TODO this might be a problem when saving a subset of the images
            with XsocsH5.XsocsH5(self.__xsocsH5_f) as xsocsH5:
                image_size = xsocsH5.image_size()
            if image_size != mask.shape:
                raise ValueError('Mask has not the size of the images')

        self.__params['mask'] = mask

Damien Naudet's avatar
Damien Naudet committed
339 340 341
    @medfilt_dims.setter
    def medfilt_dims(self, medfilt_dims):
        """
342
        Median filter applied to the image after the mask (if any) and
Damien Naudet's avatar
Damien Naudet committed
343 344 345 346
        before converting to qspace.
        """
        err = False
        if medfilt_dims is None:
347
            self.__params['medfilt_dims'] = (1, 1)
Damien Naudet's avatar
Damien Naudet committed
348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364
            return

        medfilt_dims_int = None
        if len(medfilt_dims) != 2:
            raise ValueError('medfilt_dims must be a two elements array.')
        if None in medfilt_dims:
            err = True
        else:
            medfilt_dims_int = [int(medfilt_dims[0]), int(medfilt_dims[1])]
            if min(medfilt_dims_int) <= 0:
                err = True
        if err:
            raise ValueError('<medfilt_dims> values must be strictly'
                             ' positive integers.')
        self.__params['medfilt_dims'] = np.array(medfilt_dims_int,
                                                 dtype=np.int32)

365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385
    @beam_energy.setter
    def beam_energy(self, beam_energy):
        """Beam energy setter"""
        beam_energy = float(beam_energy) if beam_energy is not None else None
        self.__params['beam_energy'] = beam_energy

    @direct_beam.setter
    def direct_beam(self, direct_beam):
        """Direct beam calibration position"""
        if direct_beam is not None:
            direct_beam = float(direct_beam[0]), float(direct_beam[1])
        self.__params['direct_beam'] = direct_beam

    @channels_per_degree.setter
    def channels_per_degree(self, channels_per_degree):
        """Channels per degree calibration"""
        if channels_per_degree is not None:
            channels_per_degree = (float(channels_per_degree[0]),
                                   float(channels_per_degree[1]))
        self.__params['channels_per_degree'] = channels_per_degree

Damien Naudet's avatar
Damien Naudet committed
386 387 388 389 390 391 392 393 394 395
    @roi.setter
    def roi(self, roi):
        """
        Sets the roi. Set to None to unset it. To change an already set roi
        the previous one has to be unset first.
        :param roi: roi coordinates in sample coordinates.
            Four elements array : (xmin, xmax, ymin, ymax)
        :return:
        """
        if self.roi is False:
396 397
            raise ValueError('Cannot set a rectangular ROI, pos_indices are '
                             'already set, remove them first.')
Damien Naudet's avatar
Damien Naudet committed
398
        self.__params['roi'] = roi
399 400 401 402 403 404 405 406 407 408 409
        self.__params['sample_indices'] = self.__indices_from_roi()

    def __indices_from_roi(self):
        # TODO : check all positions
        # at the moment using only the first scan's positions
        with XsocsH5.XsocsH5(self.__xsocsH5_f) as xsocsH5:
            entries = xsocsH5.entries()
            positions = xsocsH5.scan_positions(entries[0])
            x_pos = positions.pos_0
            y_pos = positions.pos_1

Damien Naudet's avatar
Damien Naudet committed
410 411 412 413 414 415
        if self.__shiftH5:
            with self.__shiftH5:
                shifted_idx = self.__shiftH5.shifted_indices(entries[0])
        else:
            shifted_idx = None

Damien Naudet's avatar
Damien Naudet committed
416 417
        roi = self.roi
        if self.roi is None:
Damien Naudet's avatar
Damien Naudet committed
418 419
            if shifted_idx is not None and shifted_idx.size != 0:
                return np.arange(shifted_idx.size)
420 421
            return np.arange(len(x_pos))

Damien Naudet's avatar
Damien Naudet committed
422 423 424 425
        if shifted_idx is not None and shifted_idx.size != 0:
            x_pos = x_pos[shifted_idx]
            y_pos = y_pos[shifted_idx]

Damien Naudet's avatar
Damien Naudet committed
426 427 428 429
        x_min = roi[0]
        x_max = roi[1]
        y_min = roi[2]
        y_max = roi[3]
430 431 432 433 434

        # we cant do this because the points arent perfectly aligned!
        # we could end up with non rectangular rois
        pos_indices = np.where((x_pos >= x_min) & (x_pos <= x_max) &
                               (y_pos >= y_min) & (y_pos <= y_max))[0]
Damien Naudet's avatar
Damien Naudet committed
435

436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479
        # # TODO : rework this
        # n_x = scan_params['motor_0_steps']
        # n_y = scan_params['motor_1_steps']
        # steps_0 = scan_params['motor_0_steps']
        # steps_1 = scan_params['motor_1_steps']
        # x = np.linspace(scan_params['motor_0_start'],
        #                 scan_params['motor_0_end'], steps_0, endpoint=False)
        # y = np.linspace(scan_params['motor_1_start'],
        #                 scan_params['motor_1_end'], steps_1, endpoint=False)


        # x_pos = x_pos[]
        #
        # x_pos.shape = (n_y, n_x)
        # y_pos.shape = (n_y, n_x)
        # pos_indices_2d = np.where((x_pos >= x_min) & (x_pos <= x_max) &
        #                           (y_pos >= y_min) & (y_pos <= y_max))[0]
        return pos_indices  # pos_indices_2d.shape

    def check_overwrite(self):
        """
        Checks if the output file(s) already exist(s).
        """
        output_f = self.__output_f
        if output_f is not None and os.path.exists(output_f):
            return [self.__output_f]
        return []

    def summary(self):
        """
        Gives an summary of what will be done.
        """
        # TODO : finish
        files = [self.output_f]
        return files

    def check_parameters(self):
        """
        Checks if the RecipSpaceConverter parameters are valid.
        Returns a list of strings describing those errors, if any,
        or an empty list.
        """
        errors = []

480
        qspace_dims = self.qspace_dims
481 482 483
        if (qspace_dims is None
                or None in qspace_dims
                or len(qspace_dims) != 3
Damien Naudet's avatar
Damien Naudet committed
484
                or min(qspace_dims) <= 0):
485 486
            errors.append('- "qspace size" must be an array of three'
                          ' strictly positive integers.')
487

488 489
        return errors

490 491 492 493
    def check_consistency(self,
                          beam_energy_check=True,
                          direct_beam_check=True,
                          channels_per_degree_check=True):
494 495 496 497 498 499
        """
        Check if all entries have the same values plus some other
        MINIMAL checks.
        This does not check if the parameter values are valid.
        Returns a list of strings describing those errors, if any,
        or an empty list.
500 501 502 503

        :param bool beam_energy_check: Toggle beam_energy check
        :param bool direct_beam_check: Toggle direct_beam check
        :param bool channels_per_degree_check: Toggle channels_per_degree check
504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522
        """
        errors = []

        params = _get_all_params(self.__xsocsH5_f)

        def check_values(dic, key, description):
            values = [dic[scan][key] for scan in sorted(dic.keys())]
            if isinstance(values[0], (list, tuple)):
                values = [tuple(val) for val in values]
            values_set = set(values)
            if len(values_set) != 1:
                errors.append('Parameter inconsistency : '
                              '"{0}" : {1}.'
                              ''.format(description, '; '.join(str(m)
                                        for m in values_set)))

        check_values(params, 'n_images', 'Number of images')
        check_values(params, 'n_positions', 'Number of X/Y positions')
        check_values(params, 'img_size', 'Images size')
523 524 525 526 527 528
        if beam_energy_check:
            check_values(params, 'beam_energy', 'Beam energy')
        if channels_per_degree_check:
            check_values(params, 'chan_per_deg', 'Chan. per deg.')
        if direct_beam_check:
            check_values(params, 'center_chan', 'Center channel')
529

Damien Naudet's avatar
Damien Naudet committed
530 531 532
        keys = list(params.keys())
        n_images = params[keys[0]]['n_images']
        n_positions = params[keys[0]]['n_positions']
533 534 535 536 537 538 539 540
        if n_images != n_positions:
            errors.append('number of images != number of X/Y coordinates '
                          'on sample : '
                          '{0} != {1}'.format(n_images, n_positions))

        return errors

    def scan_params(self, scan):
Damien Naudet's avatar
Damien Naudet committed
541
        """ Returns the scan parameters (filled during acquisition). """
542 543 544 545 546 547 548 549 550 551 552
        params = _get_all_params(self.__xsocsH5_f)
        return params[scan]

    def __run_convert(self):
        """
        Performs the conversion.
        :return:
        """

        self.__set_status(self.RUNNING)

553
        normalizer = self.normalizer
Damien Naudet's avatar
Damien Naudet committed
554
        medfilt_dims = self.medfilt_dims
555 556 557
        qspace_dims = self.qspace_dims
        xsocsH5_f = self.xsocsH5_f
        output_f = self.output_f
558
        sample_roi = self.__params['roi']
559 560 561
        beam_energy = self.__params['beam_energy']
        center_chan = self.__params['direct_beam']
        chan_per_deg = self.__params['channels_per_degree']
562 563

        try:
Damien Naudet's avatar
Damien Naudet committed
564 565 566 567 568 569 570 571 572 573 574 575 576 577
            # setting medfilt_dims to None if it is equal to [1, 1]
            if medfilt_dims is not None:
                medfilt_dims = np.array(medfilt_dims, ndmin=1)
                if medfilt_dims.ndim != 1:
                    raise ValueError('medfilt_dims must be a 1D array.')
                if medfilt_dims.size > 2:
                    raise ValueError(
                        'medfilt_dims must be a one or two elements'
                        ' array.')
                if medfilt_dims.size == 1:
                    medfilt_dims = np.repeat(medfilt_dims, 2)
                if np.any(np.less(medfilt_dims, [1, 1])):
                    raise ValueError('medfilt_dims values must be >= 1.')

578 579 580 581
            ta = time.time()

            params = _get_all_params(xsocsH5_f)

582
            entries = self.__get_scans()
583 584 585 586
            n_entries = len(entries)

            first_param = params[entries[0]]

587 588
            if beam_energy is None:  # Load it from first entry
                beam_energy = first_param['beam_energy']
589 590 591 592
            if beam_energy is None:
                raise ValueError('Invalid/missing beam energy : {0}.'
                                 ''.format(beam_energy))

593 594 595
            if chan_per_deg is None:  # Load it from first entry
                chan_per_deg = first_param['chan_per_deg']
            if chan_per_deg is None or len(chan_per_deg) != 2:
596 597 598
                raise ValueError('Invalid/missing chan_per_deg value : {0}.'
                                 ''.format(chan_per_deg))

599 600 601
            if center_chan is None:  # Load it from first entry
                center_chan = first_param['center_chan']
            if center_chan is None or len(center_chan) != 2:
602 603 604
                raise ValueError('Invalid/missing center_chan value : {0}.'
                                 ''.format(center_chan))

605 606 607
            # Load image ROI from first entry
            image_roi_offset = first_param['image_roi_offset']

608 609 610 611 612 613 614 615 616 617 618
            n_images = first_param['n_images']
            if n_images is None or n_images == 0:
                raise ValueError(
                    'Data does not contain any images (n_images={0}).'
                    ''.format(n_images))

            img_size = first_param['img_size']
            if img_size is None or 0 in img_size:
                raise ValueError('Invalid image size (img_size={0}).'
                                 ''.format(img_size))

619 620 621 622 623 624
            mask = self.mask
            if mask is not None:
                if np.count_nonzero(mask) == mask.size:
                    # Mask is empty, disable mask
                    mask = None
                else:
625 626 627 628 629 630
                    if (image_roi_offset is not None and
                            image_roi_offset != (0, 0)):
                        # Apply image ROI to mask
                        row, column = image_roi_offset
                        mask = mask[row[0]:row[0]+img_size[0],
                                    column[1]:column[1]+img_size[1]]
631 632
                    if mask.shape != img_size:
                        raise ValueError('Invalid mask size')
633

634 635 636 637 638
            # TODO image ROI correction
            if self.maxipix_correction and (img_size[0] > 516 or img_size[1] > 516):
                # Maxipix correction for something else
                raise ValueError('Invalid image size for Maxipix correction')

Damien Naudet's avatar
Damien Naudet committed
639 640 641 642 643 644 645 646 647 648 649
            shiftH5 = self.__shiftH5

            if shiftH5:
                shifted_idx = shiftH5.shifted_indices(entries[0])
                if shifted_idx is not None and shifted_idx.size > 0:
                    n_images = shifted_idx.size
                else:
                    shifted_idx = None
            else:
                shifted_idx = None

650 651 652 653 654 655 656 657 658 659
            # TODO value testing
            sample_indices = self.sample_indices
            if sample_indices is None:
                sample_indices = np.arange(n_images)
            else:
                n_images = len(sample_indices)

            n_xy = len(sample_indices)

            print('Parameters :')
660 661 662 663
            print('\t- beam energy        : {0}'.format(beam_energy))
            print('\t- center channel     : {0}'.format(center_chan))
            print('\t- image roi offset   : {0}'.format(image_roi_offset))
            print('\t- channel per degree : {0}'.format(chan_per_deg))
664
            print('\t- maxipix correction : {0}'.format(self.maxipix_correction))
665
            print('\t- mask               : {0}'.format(
666
                'Yes' if mask is not None else 'No'))
667 668 669 670 671 672 673 674
            print('\t- normalizer         : {0}'.format(normalizer))
            print('\t- medfilt dims       : {0}'.format(medfilt_dims))
            print('\t- qspace size        : {0}'.format(qspace_dims))

            # Offset center_chan with image roi offset if any
            if image_roi_offset is not None:
                center_chan = (center_chan[0] - image_roi_offset[0],
                               center_chan[1] - image_roi_offset[1])
675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698

            # TODO : make this editable?
            nx, ny, nz = qspace_dims
            qconv = xu.experiment.QConversion(['y-', 'z-'],
                                              ['z-', 'y-'],
                                              [1, 0, 0])

            # convention for coordinate system:
            # x downstream
            # z upwards
            # y to the "outside"
            # (righthanded)
            hxrd = xu.HXRD([1, 0, 0],
                           [0, 0, 1],
                           en=beam_energy,
                           qconv=qconv)

            hxrd.Ang2Q.init_area('z-',
                                 'y+',
                                 cch1=center_chan[0],
                                 cch2=center_chan[1],
                                 Nch1=img_size[0],
                                 Nch2=img_size[1],
                                 chpdeg1=chan_per_deg[0],
699
                                 chpdeg2=chan_per_deg[1])
700 701 702

            # shape of the array that will store the qx/qy/qz for all
            # rocking angles
703
            q_shape = n_entries, img_size[0] * img_size[1], 3
704 705 706 707 708 709 710 711 712 713

            # then the array
            q_ar = np.zeros(q_shape, dtype=np.float64)

            img_dtype = None

            with XsocsH5.XsocsH5(xsocsH5_f, mode='r') as master_h5:

                entry_files = []

714 715
                all_entries = set(master_h5.entries())

716 717 718 719
                positions = master_h5.scan_positions(entries[0])
                sample_x = positions.pos_0
                sample_y = positions.pos_1

Damien Naudet's avatar
Damien Naudet committed
720 721 722 723
                if shifted_idx is not None:
                    sample_x = sample_x[shifted_idx]
                    sample_y = sample_y[shifted_idx]

724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751
                for entry_idx, entry in enumerate(entries):
                    entry_file = master_h5.entry_filename(entry)
                    if not os.path.isabs(entry_file):
                        base_dir = os.path.dirname(xsocsH5_f)
                        entry_file = os.path.abspath(os.path.join(base_dir,
                                                                  entry_file))
                    entry_files.append(entry_file)

                    phi = np.float64(master_h5.positioner(entry, 'phi'))
                    eta = np.float64(master_h5.positioner(entry, 'eta'))
                    nu = np.float64(master_h5.positioner(entry, 'nu'))
                    delta = np.float64(master_h5.positioner(entry, 'del'))

                    qx, qy, qz = hxrd.Ang2Q.area(eta, phi, nu, delta)
                    q_ar[entry_idx, :, 0] = qx.reshape(-1)
                    q_ar[entry_idx, :, 1] = qy.reshape(-1)
                    q_ar[entry_idx, :, 2] = qz.reshape(-1)

                    entry_dtype = master_h5.image_dtype(entry=entry)

                    if img_dtype is None:
                        img_dtype = entry_dtype
                    elif img_dtype != entry_dtype:
                        raise TypeError(
                            'All images in the input HDF5 files should '
                            'be of the same type. Found {0} and {1}.'
                            ''.format(img_dtype, entry_dtype))

752 753 754
            if mask is not None:
                # Mark masked pixels (i.e., non zero in the mask) with NaN
                q_ar[:, mask.reshape(-1) != 0, :] = np.nan
755

Damien Naudet's avatar
Damien Naudet committed
756 757
            # custom bins range to have the same histo as
            # xrayutilities.gridder3d
758 759 760 761 762 763
            # bins centered around the qx, qy, qz
            # bins will be like :
            # bin_1 = [min - step/2, min + step/2[
            # bin_2 = [min - step/2, min + 3*step/2]
            # ...
            # bin_N = [max - step/2, max + step/2]
764 765 766 767 768 769
            qx_min = np.nanmin(q_ar[:, :, 0])
            qy_min = np.nanmin(q_ar[:, :, 1])
            qz_min = np.nanmin(q_ar[:, :, 2])
            qx_max = np.nanmax(q_ar[:, :, 0])
            qy_max = np.nanmax(q_ar[:, :, 1])
            qz_max = np.nanmax(q_ar[:, :, 2])
770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844

            step_x = (qx_max - qx_min) / (nx - 1.)
            step_y = (qy_max - qy_min) / (ny - 1.)
            step_z = (qz_max - qz_min) / (nz - 1.)

            bins_rng_x = ([qx_min - step_x / 2., qx_min +
                           (qx_max - qx_min + step_x) - step_x / 2.])
            bins_rng_y = ([qy_min - step_y / 2., qy_min +
                           (qy_max - qy_min + step_y) - step_y / 2.])
            bins_rng_z = ([qz_min - step_z / 2., qz_min +
                           (qz_max - qz_min + step_z) - step_z / 2.])
            bins_rng = [bins_rng_x, bins_rng_y, bins_rng_z]

            qx_idx = qx_min + step_x * np.arange(0, nx, dtype=np.float64)
            qy_idx = qy_min + step_y * np.arange(0, ny, dtype=np.float64)
            qz_idx = qz_min + step_z * np.arange(0, nz, dtype=np.float64)

            # TODO : on windows we may be forced to use shared memory
            # TODO : find why we use more memory when using shared arrays
            #        this shouldnt be the case
            #        (use the same amount as non shared mem)
            # on linux apparently we dont because when fork() is called data is
            # only copied on write.
            # shared histo used by all processes
            # histo_shared = mp_sharedctypes.RawArray(ctypes.c_int32,
            #                                         nx * ny * nz)
            # histo = np.frombuffer(histo_shared, dtype='int32')
            # histo.shape = nx, ny, nz
            # histo[:] = 0
            histo = np.zeros(qspace_dims, dtype=np.int32)

            # shared LUT used by all processes
            # h_lut = None
            # h_lut_shared = None
            h_lut = []
            lut = None
            for h_idx in range(n_entries):
                lut = histogramnd_get_lut(q_ar[h_idx, ...],
                                          bins_rng,
                                          [nx, ny, nz],
                                          last_bin_closed=True)

                # if h_lut_shared is None:
                #     lut_dtype = lut[0].dtype
                #     if lut_dtype == np.int16:
                #         lut_ctype = ctypes.c_int16
                #     elif lut_dtype == np.int32:
                #         lut_ctype = ctypes.c_int32
                #     elif lut_dtype == np.int64:
                #         lut_ctype == ctypes.c_int64
                #     else:
                #         raise TypeError('Unknown type returned by '
                #                         'histogramnd_get_lut : {0}.'
                #                         ''.format(lut.dtype))
                #     h_lut_shared = mp_sharedctypes.RawArray(lut_ctype,
                #                                       n_images * lut[0].size)
                #     h_lut = np.frombuffer(h_lut_shared, dtype=lut_dtype)
                #     h_lut.shape = (n_images, -1)
                #
                # h_lut[h_idx, ...] = lut[0]
                h_lut.append(lut[0])
                histo += lut[1]

            del lut
            del q_ar

            # TODO : split the output file into several files? speedup?
            output_shape = histo.shape

            chunks = (1,
                      max(output_shape[0] // 4, 1),
                      max(output_shape[1] // 4, 1),
                      max(output_shape[2] // 4, 1),)
            qspace_sum_chunks = max(n_images // 10, 1),

845 846
            discarded_entries = sorted(all_entries - set(entries))

847 848
            _create_result_file(output_f,
                                output_shape,
Damien Naudet's avatar
Damien Naudet committed
849
                                medfilt_dims,
850
                                sample_roi,
851 852 853 854 855 856
                                sample_x[sample_indices],
                                sample_y[sample_indices],
                                qx_idx,
                                qy_idx,
                                qz_idx,
                                histo,
857 858
                                selected_entries=entries,
                                discarded_entries=discarded_entries,
859 860 861
                                compression='lzf',
                                qspace_chunks=chunks,
                                qspace_sum_chunks=qspace_sum_chunks,
Damien Naudet's avatar
Damien Naudet committed
862
                                overwrite=self.__overwrite,
863 864 865
                                shiftH5=shiftH5,
                                beam_energy=beam_energy,
                                direct_beam=center_chan,
866
                                channels_per_degree=chan_per_deg,
867
                                normalizer=normalizer,
868 869
                                mask=mask,
                                maxipix_correction=self.maxipix_correction)
870 871 872 873 874 875 876 877 878 879 880 881 882 883 884

            manager = mp.Manager()
            self.__term_evt = term_evt = manager.Event()

            write_lock = manager.Lock()
            idx_queue = manager.Queue()

            n_proc = self.n_proc
            if n_proc is None or n_proc <= 0:
                n_proc = mp.cpu_count()

            self.__shared_progress = mp_sharedctypes.RawArray(ctypes.c_int32,
                                                              n_proc)
            np.frombuffer(self.__shared_progress, dtype='int32')[:] = 0

Damien Naudet's avatar
Damien Naudet committed
885 886 887 888 889 890 891 892 893 894 895 896 897 898 899
            if shiftH5 is not None and shifted_idx is not None:
                n_shifted = n_entries * shifted_idx.size
                shared_shifted = mp_sharedctypes.RawArray(ctypes.c_int32,
                                                          n_shifted)
                shifted_np = np.frombuffer(shared_shifted, dtype='int32')
                shifted_np.shape = n_entries, shifted_idx.size
                shared_shifted_shape = shifted_np.shape

                for i_entry, entry in enumerate(entries):
                    shifted_indices = shiftH5.shifted_indices(entry)
                    shifted_np[i_entry, :] = shifted_indices
            else:
                shared_shifted = None
                shared_shifted_shape = None

900 901 902 903 904 905 906 907 908 909 910
            pool = mp.Pool(n_proc,
                           initializer=_init_thread,
                           initargs=(idx_queue,
                                     write_lock,
                                     bins_rng,
                                     qspace_dims,
                                     h_lut,  # _shared,
                                     None,  # lut_dtype,
                                     n_xy,
                                     histo,  # _shared,))
                                     self.__shared_progress,
Damien Naudet's avatar
Damien Naudet committed
911 912 913
                                     term_evt,
                                     shared_shifted,
                                     shared_shifted_shape,))
914 915 916 917 918 919 920 921 922

            if disp_times:
                class myTimes(object):
                    def __init__(self):
                        self.t_histo = 0.
                        self.t_sum = 0.
                        self.t_mask = 0.
                        self.t_read = 0.
                        self.t_context = 0.
923
                        self.t_preprocess = 0.
924 925 926 927 928
                        self.t_medfilt = 0.
                        self.t_write = 0.
                        self.t_w_lock = 0.

                    def update(self, arg):
929
                        (t_read_, t_context_, t_preprocess_, t_medfilt_, t_histo_,
930 931 932 933 934 935
                         t_mask_, t_sum_, t_write_, t_w_lock_) = arg[2]
                        self.t_histo += t_histo_
                        self.t_sum += t_sum_
                        self.t_mask += t_mask_
                        self.t_read += t_read_
                        self.t_context += t_context_
936
                        self.t_preprocess += t_preprocess_
937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953
                        self.t_medfilt += t_medfilt_
                        self.t_write += t_write_
                        self.t_w_lock += t_w_lock_

                res_times = myTimes()
                callback = res_times.update
            else:
                callback = None

            # creating the processes
            results = []
            for th_idx in range(n_proc):
                arg_list = (th_idx,
                            entry_files,
                            entries,
                            img_size,
                            output_f,
954
                            self.maxipix_correction,
955
                            normalizer,
Damien Naudet's avatar
Damien Naudet committed
956
                            medfilt_dims,
957 958 959 960 961 962 963 964
                            img_dtype)
                res = pool.apply_async(_to_qspace, args=arg_list,
                                       callback=callback)
                results.append(res)
            # sending the image indices
            for result_idx, pos_idx in enumerate(sample_indices):
                idx_queue.put((result_idx, pos_idx))

Damien Naudet's avatar
Damien Naudet committed
965 966
            # sending the None value to let the threads know that they
            # should return
967 968 969 970 971 972 973 974 975 976 977 978
            for th_idx in range(n_proc):
                idx_queue.put(None)

            pool.close()
            pool.join()

            tb = time.time()

            if disp_times:
                print('TOTAL {0}'.format(tb - ta))
                print('Read {0}'.format(res_times.t_read))
                print('Context {0}'.format(res_times.t_context))
979
                print('Preprocess {0}'.format(res_times.t_preprocess))
980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028
                print('Medfilt {0}'.format(res_times.t_medfilt))
                print('Histo {0}'.format(res_times.t_histo))
                print('Mask {0}'.format(res_times.t_mask))
                print('Sum {0}'.format(res_times.t_sum))
                print('Write {0}'.format(res_times.t_write))
                print('(lock : {0})'.format(res_times.t_w_lock))

            proc_results = [result.get() for result in results]
            proc_codes = np.array([proc_result[0]
                                   for proc_result in proc_results])

            rc = self.DONE
            if not np.all(proc_codes == self.DONE):
                if self.ERROR in proc_codes:
                    rc = self.ERROR
                elif self.CANCELED in proc_codes:
                    rc = self.CANCELED
                else:
                    raise ValueError('Unknown return code.')

            if rc != self.DONE:
                errMsg = 'Conversion failed. Process status :'
                for th_idx, result in enumerate(proc_results):
                    errMsg += ('\n- Proc {0} : rc={1}; {2}'
                               ''.format(th_idx, result[0], result[1]))
                self.__set_status(rc, errMsg)
            else:
                self.__set_status(rc)

        except Exception as ex:
            self.__set_status(self.ERROR, str(ex))
        else:
            self.__results = self.output_f

        # TODO : catch exception?
        if self.__callback:
            self.__callback()

        return self.__results

    def wait(self):
        """
        Waits until parsing is done, or returns if it is not running.
        :return:
        """
        if self.__thread:
            self.__thread.join()

    def __running_exception(self):
Damien Naudet's avatar
Damien Naudet committed
1029
        """ Raises an exception if a conversion is in progress. """
1030 1031 1032 1033 1034
        if self.is_running():
            raise RuntimeError('Operation not permitted while '
                               'a parse or merge in running.')

    def is_running(self):
Damien Naudet's avatar
Damien Naudet committed
1035
        """ Returns True if a conversion is in progress. """
1036 1037 1038 1039 1040
        return self.status == QSpaceConverter.RUNNING
        #self.__thread and self.__thread.is_alive()

    @output_f.setter
    def output_f(self, output_f):
Damien Naudet's avatar
Damien Naudet committed
1041
        """ Sets the output file. """
1042 1043 1044 1045 1046 1047 1048
        if not isinstance(output_f, str):
            raise TypeError('output_f must be a string. Received {0}'
                            ''.format(type(output_f)))
        self.__output_f = output_f

    @n_proc.setter
    def n_proc(self, n_proc):
Damien Naudet's avatar
Damien Naudet committed
1049 1050 1051 1052
        """ Sets the number of processes to use. If None or 0 the number of
            processes used will be the number returned by
            multiprocessing.cpu_count.
        """
1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063
        if n_proc is None:
            self.__n_proc = None
            return

        n_proc = int(n_proc)
        if n_proc <= 0:
            self.__n_proc = None
        else:
            self.__n_proc = n_proc

    def abort(self, wait=True):
Damien Naudet's avatar
Damien Naudet committed
1064 1065 1066 1067 1068 1069
        """
        Aborts the current conversion, if any.
        :param wait: set to False to return immediatly without waiting for the
        processes to return.
        :return:
        """
1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094
        if self.is_running():
            self.__term_evt.set()
            if wait:
                self.wait()

    def progress(self):
        """
        Returns the progress of the conversion.
        :return:
        """
        if self.__shared_progress:
            progress = np.frombuffer(self.__shared_progress, dtype='int32')
            return progress.max()
        return 0


def _init_thread(idx_queue_,
                 write_lock_,
                 bins_rng_,
                 qspace_size_,
                 h_lut_shared_,
                 h_lut_dtype_,
                 n_xy_,
                 histo_shared_,
                 shared_progress_,
Damien Naudet's avatar
Damien Naudet committed
1095 1096 1097
                 term_evt_,
                 shared_shifted_,
                 shared_shifted_shape_):
1098 1099 1100 1101 1102 1103 1104 1105 1106
    global idx_queue, \
        write_lock, \
        bins_rng, \
        qspace_size, \
        h_lut_shared, \
        h_lut_dtype, \
        n_xy, \
        histo_shared, \
        shared_progress, \
Damien Naudet's avatar
Damien Naudet committed
1107 1108 1109
        term_evt, \
        shared_shifted, \
        shared_shifted_shape
1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120

    idx_queue = idx_queue_
    write_lock = write_lock_
    bins_rng = bins_rng_
    qspace_size = qspace_size_
    h_lut_shared = h_lut_shared_
    h_lut_dtype = h_lut_dtype_
    n_xy = n_xy_
    histo_shared = histo_shared_
    shared_progress = shared_progress_
    term_evt = term_evt_
Damien Naudet's avatar
Damien Naudet committed
1121 1122
    shared_shifted = shared_shifted_
    shared_shifted_shape = shared_shifted_shape_
1123 1124 1125


def _create_result_file(h5_fn,
1126
                        qspace_dims,
Damien Naudet's avatar
Damien Naudet committed
1127
                        medfilt_dims,
1128
                        sample_roi,
1129 1130
                        pos_x,
                        pos_y,
1131 1132 1133
                        q_x,
                        q_y,
                        q_z,
1134
                        histo,
1135 1136
                        selected_entries,
                        discarded_entries=None,
1137 1138 1139
                        compression='lzf',
                        qspace_chunks=None,
                        qspace_sum_chunks=None,
Damien Naudet's avatar
Damien Naudet committed
1140
                        overwrite=False,
1141 1142 1143
                        shiftH5=None,
                        beam_energy=None,
                        direct_beam=None,
1144
                        channels_per_degree=None,
1145
                        normalizer='',
1146 1147
                        mask=None,
                        maxipix_correction=False):
1148 1149 1150 1151
    """
    Initializes the output file.
    :param h5_fn: name of the file to initialize
    :param qspace_dims: dimensions of the q space
Damien Naudet's avatar
Damien Naudet committed
1152
    :param medfilt_dims: dimensions of the median filter applied to the image
1153
        after the mask (if any).
1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165
    :param pos_x: sample X positions (one for each qspace cube)
    :param pos_y: sample Y positions (one for each qspace cube)
    :param q_x: X coordinates of the qspace cube
    :param q_y: Y coordinates of the qspace cube
    :param q_z: Z coordinates of the qspace cube
    :param histo: histogram (number of hits per element of the qspace elements)
    :param selected_entries: list of input entries used for the conversion
    :param discarded_entries: list of input entries discarded, or None
    :param compression: datasets compression
    :param qspace_chunks: qspace chunking
    :param qspace_sum_chunks:
    :param overwrite: True to force overwriting the file if it already exists.
Damien Naudet's avatar
Damien Naudet committed
1166
    :param shiftH5: file containing the shifts applied to the selected entries
1167 1168 1169
    :param beam_energy: Beam energy in eV
    :param direct_beam: Direct beam calibration position
    :param channels_per_degree: Channels per degree calibration
1170 1171
    :param str normalizer:
        Name of measurement group dataset used for normalization
1172 1173
    :param Union[numpy.ndarray, None] mask:
        Mask used to discard pixels in images
1174 1175
    :param bool maxipix_correction:
        Whether or not to apply a correction for maxipix module edges
1176 1177
    """

1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188
    if not overwrite:
        mode = 'w-'
    else:
        mode = 'w'

    dir_name = os.path.dirname(h5_fn)
    if len(dir_name) > 0 and not os.path.exists(dir_name):
        os.makedirs(dir_name)

    qspace_h5 = QSpaceH5.QSpaceH5Writer(h5_fn, mode=mode)
    qspace_h5.init_file(len(pos_x),
1189
                        qspace_dims,
1190 1191 1192 1193 1194 1195
                        qspace_chunks=qspace_chunks,
                        qspace_sum_chunks=qspace_sum_chunks,
                        compression=compression)
    qspace_h5.set_histo(histo)
    qspace_h5.set_sample_x(pos_x)
    qspace_h5.set_sample_y(pos_y)
1196 1197 1198
    qspace_h5.set_qx(q_x)
    qspace_h5.set_qy(q_y)
    qspace_h5.set_qz(q_z)
Damien Naudet's avatar
Damien Naudet committed
1199
    qspace_h5.set_medfilt_dims(medfilt_dims)
1200
    qspace_h5.set_sample_roi(sample_roi)
1201 1202 1203
    qspace_h5.set_beam_energy(beam_energy)
    qspace_h5.set_direct_beam(direct_beam)
    qspace_h5.set_channels_per_degree(channels_per_degree)
1204
    qspace_h5.set_maxipix_correction(maxipix_correction)
1205

1206 1207 1208 1209
    if normalizer is None:
        normalizer = ''
    qspace_h5.set_image_normalizer(normalizer)

1210 1211 1212
    if mask is not None:
        qspace_h5.set_image_mask(mask)

Damien Naudet's avatar
Damien Naudet committed
1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235
    if shiftH5:
        sample_shifts = []
        grid_shifts = []
        with shiftH5:
            for entry in selected_entries:
                shift = shiftH5.shift(entry)

                if shift is not None:
                    sample_shifts.append([shift['shift_x'], shift['shift_y']])
                    if shiftH5.is_snapped_to_grid():
                        grid_shifts.append(shift['grid_shift'])

        if len(grid_shifts) == 0:
            grid_shifts = None
    else:
        sample_shifts = None
        grid_shifts = None

    qspace_h5.set_entries(selected_entries,
                          discarded=discarded_entries,
                          sample_shifts=sample_shifts,
                          grid_shifts=grid_shifts)

1236 1237 1238 1239 1240 1241

def _to_qspace(th_idx,
               entry_files,
               entries,
               img_size,
               output_fn,
1242
               maxipix_correction,
1243
               normalizer,
Damien Naudet's avatar
Damien Naudet committed
1244
               medfilt_dims,
1245
               img_dtype):
Damien Naudet's avatar
Damien Naudet committed
1246 1247 1248 1249 1250 1251 1252
    """
    Fonction running in a process. Performs the conversion.
    :param th_idx:
    :param entry_files:
    :param entries:
    :param img_size:
    :param output_fn:
1253 1254
    :param bool maxipix_correction:
        Whether to apply maxipix correction or not
1255 1256
    :param str normalizer:
       Name of measurement group dataset to use for normalization
Damien Naudet's avatar
Damien Naudet committed
1257
    :param medfilt_dims:
Damien Naudet's avatar
Damien Naudet committed
1258 1259 1260
    :param img_dtype:
    :return:
    """
Damien Naudet's avatar
typo  
Damien Naudet committed
1261
    print('Process {0} started.'.format(th_idx))
1262 1263 1264 1265 1266

    t_histo = 0.
    t_mask = 0.
    t_sum = 0.
    t_read = 0.
1267
    t_preprocess = 0.
1268 1269 1270 1271 1272
    t_medfilt = 0.
    t_write = 0.
    t_w_lock = 0.
    t_context = 0.

1273
    write_lock.acquire()
1274
    output_h5 = QSpaceH5.QSpaceH5Writer(output_fn, mode='r+')
1275
    write_lock.release()
1276 1277 1278 1279 1280 1281 1282

    if shared_progress is not None:
        progress_np = np.frombuffer(shared_progress, dtype='int32')
        progress_np[th_idx] = 0
    else:
        progress_np = None

Damien Naudet's avatar
Damien Naudet committed
1283 1284 1285 1286 1287 1288
    if shared_shifted is not None:
        shifted_np = np.frombuffer(shared_shifted, dtype='int32')
        shifted_np.shape = shared_shifted_shape
    else:
        shifted_np = None

1289 1290 1291 1292 1293 1294 1295 1296 1297
    # histo = np.frombuffer(histo_shared, dtype='int32')
    # histo.shape = qspace_size
    histo = histo_shared
    mask = histo > 0

    # h_lut = np.frombuffer(h_lut_shared, dtype=h_lut_dtype)
    # h_lut.shape = (n_xy, -1)
    h_lut = h_lut_shared

1298
    if (maxipix_correction or normalizer) and img_dtype.kind != 'f':
1299 1300 1301 1302
        # Force the type to float64
        logger.info('Using float64 to perform normalization')
        img_dtype = np.float64

1303 1304
    img = np.ascontiguousarray(np.zeros(img_size), dtype=img_dtype)

1305 1306 1307
    # Set OpenMP to use a single thread (for median filter)
    os.environ["OMP_NUM_THREADS"] = "1"

1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327
    rc = None
    errMsg = None
    try:
        while True:
            if term_evt.is_set():  # noqa
                rc = QSpaceConverter.CANCELED
                raise Exception('conversion aborted')

            next_data = idx_queue.get()
            if next_data is None:
                rc = QSpaceConverter.DONE
                break

            result_idx, image_idx = next_data
            if result_idx % 100 == 0:
                print('#{0}/{1}'.format(result_idx, n_xy))

            cumul = None
            # histo = None

Damien Naudet's avatar
Damien Naudet committed
1328 1329 1330 1331 1332
            if shifted_np is not None:
                image_indices = shifted_np[:, image_idx]
            else:
                image_indices = None

1333
            for entry_idx, entry in enumerate(entries):
1334
                xsocsH5 = XsocsH5.XsocsH5(entry_files[entry_idx], mode='r')
1335 1336 1337

                t0 = time.time()

Damien Naudet's avatar
Damien Naudet committed
1338 1339 1340
                if image_indices is not None:
                    image_idx = image_indices[entry_idx]

1341 1342 1343 1344 1345 1346
                try:
                    # TODO : there s room for improvement here maybe
                    # (recreating a XsocsH5 instance each time slows down
                    # slows down things a big, not much tho)
                    # TODO : add a lock on the files if there is no SWMR
                    # test if it slows down things much
1347
                    with xsocsH5.image_dset_ctx() as img_data:
1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362
                        t1 = time.time()
                        img_data.read_direct(img,
                                             source_sel=np.s_[image_idx],
                                             dest_sel=None)
                        t_context = time.time() - t1
                        # img = img_data[image_idx].astype(np.float64)
                except Exception as ex:
                    raise RuntimeError('Error in proc {0} while reading '
                                       'img {1} from entry {2} ({3}) : {4}.'
                                       ''.format(th_idx, image_idx, entry_idx,
                                                 entry, ex))

                t_read += time.time() - t0
                t0 = time.time()

1363 1364 1365 1366 1367 1368 1369 1370 1371
                # Apply maxipix correction

                if maxipix_correction:
                    # TODO handle image ROI
                    img[255:258] = img[255] / 3
                    img[258:261] = img[260] / 3
                    img[:, 255:258] = (img[:, 255] / 3)[:, None]
                    img[:, 258:261] = (img[:, 260] / 3)[:, None]

1372 1373 1374 1375 1376 1377 1378 1379
                # Apply normalization

                if normalizer:
                    normalization = xsocsH5.measurement(entry, normalizer)
                    # Make sure to use float to perform division
                    assert img.dtype.kind == 'f'
                    img /= normalization[image_idx]

1380
                intensity = img
1381

1382
                t_preprocess += time.time() - t0
1383 1384 1385
                t0 = time.time()

                # intensity = medfilt2d(intensity, 3)
Damien Naudet's avatar
Damien Naudet committed
1386
                if medfilt_dims[0] != 1 or medfilt_dims[1] != 1:
1387 1388 1389 1390
                    intensity = medfilt2d(intensity,
                                          medfilt_dims,
                                          mode='constant',
                                          cval=0)
1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441

                t_medfilt += time.time() - t0
                t0 = time.time()

                try:
                    cumul = histogramnd_from_lut(intensity.reshape(-1),
                                                 h_lut[entry_idx],
                                                 shape=qspace_size,
                                                 weighted_histo=cumul,
                                                 dtype=np.float64)
                except Exception as ex:
                    print('EX2 {0}'.format(str(ex)))
                    raise ex

                t_histo += time.time() - t0

            t0 = time.time()
            cumul_sum = cumul.sum(dtype=np.float64)
            t_sum += time.time() - t0

            t0 = time.time()
            # cumul[mask] = cumul[mask]/histo[mask]
            t_mask += time.time() - t0

            t0 = time.time()
            write_lock.acquire()
            t_w_lock += time.time() - t0
            t0 = time.time()
            try:
                output_h5.set_position_data(result_idx, cumul, cumul_sum)
            except Exception as ex:
                raise RuntimeError('Error in proc {0} while writing result '
                                   'for img {1} (idx = {3}) : {2}.)'
                                   ''.format(th_idx, image_idx, ex, result_idx))
            write_lock.release()

            if progress_np is not None:
                progress_np[th_idx] = round(100. * (result_idx + 1.) / n_xy)

            t_write += time.time() - t0
    except Exception as ex:
        if rc is None:
            rc = QSpaceConverter.ERROR
        errMsg = 'In thread {0} : {1}.'.format(th_idx, str(ex))
        term_evt.set()

    if rc is None:
        rc = QSpaceConverter.DONE

    if disp_times:
        print('Thread {0} is done. Times={1}'
1442
              ''.format(th_idx, (t_read, t_context, t_preprocess,