if debug and not flat_phase:
    gu.multislices_plot(
        phase,
        sum_frames=False,
        plot_colorbar=True,
        width_z=200,
        width_y=200,
        width_x=200,
        vmin=-phase_range,
        vmax=phase_range,
        cmap=my_cmap,
        title="Phase before wrapping\n",
    )

phase = util.wrap(phase, start_angle=-np.pi, range_angle=2 * np.pi)

support[abs(amp) < support_threshold * abs(amp).max()] = 0
del amp

volume = support.sum() * voxel_size**3  # in nm3
print("estimated volume", volume, " nm3")

phase[support == 0] = 0

if ref_axis_outplane == "x":
    _, _, strain = np.gradient(
        planar_dist / (2 * np.pi) * phase,
        voxel_size)  # q is along x after rotating the crystal
elif ref_axis_outplane == "y":
    _, strain, _ = np.gradient(
Example #2
0
def run(prm):
    """
    Run the postprocessing.

    :param prm: the parsed parameters
    """
    pretty = pprint.PrettyPrinter(indent=4)

    ################################
    # assign often used parameters #
    ################################
    bragg_peak = prm.get("bragg_peak")
    debug = prm.get("debug", False)
    comment = prm.get("comment", "")
    centering_method = prm.get("centering_method", "max_com")
    original_size = prm.get("original_size")
    phasing_binning = prm.get("phasing_binning", [1, 1, 1])
    preprocessing_binning = prm.get("preprocessing_binning", [1, 1, 1])
    ref_axis_q = prm.get("ref_axis_q", "y")
    fix_voxel = prm.get("fix_voxel")
    save = prm.get("save", True)
    tick_spacing = prm.get("tick_spacing", 50)
    tick_direction = prm.get("tick_direction", "inout")
    tick_length = prm.get("tick_length", 10)
    tick_width = prm.get("tick_width", 2)
    invert_phase = prm.get("invert_phase", True)
    correct_refraction = prm.get("correct_refraction", False)
    threshold_unwrap_refraction = prm.get("threshold_unwrap_refraction", 0.05)
    threshold_gradient = prm.get("threshold_gradient", 1.0)
    offset_method = prm.get("offset_method", "mean")
    phase_offset = prm.get("phase_offset", 0)
    offset_origin = prm.get("phase_offset_origin")
    sort_method = prm.get("sort_method", "variance/mean")
    correlation_threshold = prm.get("correlation_threshold", 0.90)
    roi_detector = create_roi(dic=prm)

    # parameters below must be provided
    try:
        detector_name = prm["detector"]
        beamline_name = prm["beamline"]
        rocking_angle = prm["rocking_angle"]
        isosurface_strain = prm["isosurface_strain"]
        output_size = prm["output_size"]
        save_frame = prm["save_frame"]
        data_frame = prm["data_frame"]
        scan = prm["scan"]
        sample_name = prm["sample_name"]
        root_folder = prm["root_folder"]
    except KeyError as ex:
        print("Required parameter not defined")
        raise ex

    prm["sample"] = (f"{sample_name}+{scan}",)
    #########################
    # Check some parameters #
    #########################
    if not prm.get("backend"):
        prm["backend"] = "Qt5Agg"
    matplotlib.use(prm["backend"])
    if prm["simulation"]:
        invert_phase = False
        correct_refraction = 0
    if invert_phase:
        phase_fieldname = "disp"
    else:
        phase_fieldname = "phase"

    if data_frame == "detector":
        is_orthogonal = False
    else:
        is_orthogonal = True

    if data_frame == "crystal" and save_frame != "crystal":
        print(
            "data already in the crystal frame before phase retrieval,"
            " it is impossible to come back to the laboratory "
            "frame, parameter 'save_frame' defaulted to 'crystal'"
        )
        save_frame = "crystal"

    axis_to_array_xyz = {
        "x": np.array([1, 0, 0]),
        "y": np.array([0, 1, 0]),
        "z": np.array([0, 0, 1]),
    }  # in xyz order

    ###############
    # Set backend #
    ###############
    if prm.get("backend") is not None:
        try:
            plt.switch_backend(prm["backend"])
        except ModuleNotFoundError:
            print(f"{prm['backend']} backend is not supported.")

    ###################
    # define colormap #
    ###################
    if prm.get("grey_background"):
        bad_color = "0.7"
    else:
        bad_color = "1.0"  # white background
    colormap = gu.Colormap(bad_color=bad_color)
    my_cmap = colormap.cmap

    #######################
    # Initialize detector #
    #######################
    detector = create_detector(
        name=detector_name,
        template_imagefile=prm.get("template_imagefile"),
        roi=roi_detector,
        binning=phasing_binning,
        preprocessing_binning=preprocessing_binning,
        pixel_size=prm.get("pixel_size"),
    )

    ####################################
    # define the experimental geometry #
    ####################################
    setup = Setup(
        beamline=beamline_name,
        detector=detector,
        energy=prm.get("energy"),
        outofplane_angle=prm.get("outofplane_angle"),
        inplane_angle=prm.get("inplane_angle"),
        tilt_angle=prm.get("tilt_angle"),
        rocking_angle=rocking_angle,
        distance=prm.get("sdd"),
        sample_offsets=prm.get("sample_offsets"),
        actuators=prm.get("actuators"),
        custom_scan=prm.get("custom_scan", False),
        custom_motors=prm.get("custom_motors"),
        dirbeam_detector_angles=prm.get("dirbeam_detector_angles"),
        direct_beam=prm.get("direct_beam"),
        is_series=prm.get("is_series", False),
    )

    ########################################
    # Initialize the paths and the logfile #
    ########################################
    setup.init_paths(
        sample_name=sample_name,
        scan_number=scan,
        root_folder=root_folder,
        data_dir=prm.get("data_dir"),
        save_dir=prm.get("save_dir"),
        specfile_name=prm.get("specfile_name"),
        template_imagefile=prm.get("template_imagefile"),
    )

    setup.create_logfile(
        scan_number=scan, root_folder=root_folder, filename=detector.specfile
    )

    # load the goniometer positions needed in the calculation
    # of the transformation matrix
    setup.read_logfile(scan_number=scan)

    ###################
    # print instances #
    ###################
    print(f'{"#"*(5+len(str(scan)))}\nScan {scan}\n{"#"*(5+len(str(scan)))}')
    print("\n##############\nSetup instance\n##############")
    pretty.pprint(setup.params)
    print("\n#################\nDetector instance\n#################")
    pretty.pprint(detector.params)

    ################
    # preload data #
    ################
    if prm.get("reconstruction_file") is not None:
        file_path = (prm["reconstruction_file"],)
    else:
        root = tk.Tk()
        root.withdraw()
        file_path = filedialog.askopenfilenames(
            initialdir=detector.scandir
            if prm.get("data_dir") is None
            else detector.datadir,
            filetypes=[
                ("NPZ", "*.npz"),
                ("NPY", "*.npy"),
                ("CXI", "*.cxi"),
                ("HDF5", "*.h5"),
            ],
        )

    nbfiles = len(file_path)
    plt.ion()

    obj, extension = util.load_file(file_path[0])
    if extension == ".h5":
        comment = comment + "_mode"

    print("\n###############\nProcessing data\n###############")
    nz, ny, nx = obj.shape
    print("Initial data size: (", nz, ",", ny, ",", nx, ")")
    if not original_size:
        original_size = obj.shape
    print("FFT size before accounting for phasing_binning", original_size)
    original_size = tuple(
        [
            original_size[index] // phasing_binning[index]
            for index in range(len(phasing_binning))
        ]
    )
    print("Binning used during phasing:", detector.binning)
    print("Padding back to original FFT size", original_size)
    obj = util.crop_pad(array=obj, output_shape=original_size)

    ###########################################################################
    # define range for orthogonalization and plotting - speed up calculations #
    ###########################################################################
    zrange, yrange, xrange = pu.find_datarange(
        array=obj, amplitude_threshold=0.05, keep_size=prm.get("keep_size", False)
    )

    numz = zrange * 2
    numy = yrange * 2
    numx = xrange * 2
    print(
        f"Data shape used for orthogonalization and plotting: ({numz}, {numy}, {numx})"
    )

    ####################################################################################
    # find the best reconstruction from the list, based on mean amplitude and variance #
    ####################################################################################
    if nbfiles > 1:
        print("\nTrying to find the best reconstruction\nSorting by ", sort_method)
        sorted_obj = pu.sort_reconstruction(
            file_path=file_path,
            amplitude_threshold=isosurface_strain,
            data_range=(zrange, yrange, xrange),
            sort_method=sort_method,
        )
    else:
        sorted_obj = [0]

    #######################################
    # load reconstructions and average it #
    #######################################
    avg_obj = np.zeros((numz, numy, numx))
    ref_obj = np.zeros((numz, numy, numx))
    avg_counter = 1
    print("\nAveraging using", nbfiles, "candidate reconstructions")
    for counter, value in enumerate(sorted_obj):
        obj, extension = util.load_file(file_path[value])
        print("\nOpening ", file_path[value])
        prm[f"from_file_{counter}"] = file_path[value]

        if prm.get("flip_reconstruction", False):
            obj = pu.flip_reconstruction(obj, debugging=True)

        if extension == ".h5":
            centering_method = "do_nothing"  # do not center, data is already cropped
            # just on support for mode decomposition
            # correct a roll after the decomposition into modes in PyNX
            obj = np.roll(obj, prm.get("roll_modes", [0, 0, 0]), axis=(0, 1, 2))
            fig, _, _ = gu.multislices_plot(
                abs(obj),
                sum_frames=True,
                plot_colorbar=True,
                title="1st mode after centering",
            )

        # use the range of interest defined above
        obj = util.crop_pad(obj, [2 * zrange, 2 * yrange, 2 * xrange], debugging=False)

        # align with average reconstruction
        if counter == 0:  # the fist array loaded will serve as reference object
            print("This reconstruction will be used as reference.")
            ref_obj = obj

        avg_obj, flag_avg = reg.average_arrays(
            avg_obj=avg_obj,
            ref_obj=ref_obj,
            obj=obj,
            support_threshold=0.25,
            correlation_threshold=correlation_threshold,
            aligning_option="dft",
            space=prm.get("averaging_space", "reciprocal_space"),
            reciprocal_space=False,
            is_orthogonal=is_orthogonal,
            debugging=debug,
        )
        avg_counter = avg_counter + flag_avg

    avg_obj = avg_obj / avg_counter
    if avg_counter > 1:
        print("\nAverage performed over ", avg_counter, "reconstructions\n")
    del obj, ref_obj
    gc.collect()

    ################
    # unwrap phase #
    ################
    phase, extent_phase = pu.unwrap(
        avg_obj,
        support_threshold=threshold_unwrap_refraction,
        debugging=debug,
        reciprocal_space=False,
        is_orthogonal=is_orthogonal,
    )

    print(
        "Extent of the phase over an extended support (ceil(phase range)) ~ ",
        int(extent_phase),
        "(rad)",
    )
    phase = util.wrap(phase, start_angle=-extent_phase / 2, range_angle=extent_phase)
    if debug:
        gu.multislices_plot(
            phase,
            width_z=2 * zrange,
            width_y=2 * yrange,
            width_x=2 * xrange,
            plot_colorbar=True,
            title="Phase after unwrap + wrap",
            reciprocal_space=False,
            is_orthogonal=is_orthogonal,
        )

    #############################################
    # phase ramp removal before phase filtering #
    #############################################
    amp, phase, rampz, rampy, rampx = pu.remove_ramp(
        amp=abs(avg_obj),
        phase=phase,
        initial_shape=original_size,
        method="gradient",
        amplitude_threshold=isosurface_strain,
        threshold_gradient=threshold_gradient,
    )
    del avg_obj
    gc.collect()

    if debug:
        gu.multislices_plot(
            phase,
            width_z=2 * zrange,
            width_y=2 * yrange,
            width_x=2 * xrange,
            plot_colorbar=True,
            title="Phase after ramp removal",
            reciprocal_space=False,
            is_orthogonal=is_orthogonal,
        )

    ########################
    # phase offset removal #
    ########################
    support = np.zeros(amp.shape)
    support[amp > isosurface_strain * amp.max()] = 1
    phase = pu.remove_offset(
        array=phase,
        support=support,
        offset_method=offset_method,
        phase_offset=phase_offset,
        offset_origin=offset_origin,
        title="Phase",
        debugging=debug,
    )
    del support
    gc.collect()

    phase = util.wrap(
        obj=phase, start_angle=-extent_phase / 2, range_angle=extent_phase
    )

    ##############################################################################
    # average the phase over a window or apodize to reduce noise in strain plots #
    ##############################################################################
    half_width_avg_phase = prm.get("half_width_avg_phase", 0)
    if half_width_avg_phase != 0:
        bulk = pu.find_bulk(
            amp=amp, support_threshold=isosurface_strain, method="threshold"
        )
        # the phase should be averaged only in the support defined by the isosurface
        phase = pu.mean_filter(
            array=phase, support=bulk, half_width=half_width_avg_phase
        )
        del bulk
        gc.collect()

    if half_width_avg_phase != 0:
        comment = comment + "_avg" + str(2 * half_width_avg_phase + 1)

    gridz, gridy, gridx = np.meshgrid(
        np.arange(0, numz, 1),
        np.arange(0, numy, 1),
        np.arange(0, numx, 1),
        indexing="ij",
    )

    phase = (
        phase + gridz * rampz + gridy * rampy + gridx * rampx
    )  # put back the phase ramp otherwise the diffraction
    # pattern will be shifted and the prtf messed up

    if prm.get("apodize", False):
        amp, phase = pu.apodize(
            amp=amp,
            phase=phase,
            initial_shape=original_size,
            window_type=prm.get("apodization_window", "blackman"),
            sigma=prm.get("apodization_sigma", [0.30, 0.30, 0.30]),
            mu=prm.get("apodization_mu", [0.0, 0.0, 0.0]),
            alpha=prm.get("apodization_alpha", [1.0, 1.0, 1.0]),
            is_orthogonal=is_orthogonal,
            debugging=True,
        )
        comment = comment + "_apodize_" + prm.get("apodization_window", "blackman")

    ################################################################
    # save the phase with the ramp for PRTF calculations,          #
    # otherwise the object will be misaligned with the measurement #
    ################################################################
    np.savez_compressed(
        detector.savedir + "S" + str(scan) + "_avg_obj_prtf" + comment,
        obj=amp * np.exp(1j * phase),
    )

    ####################################################
    # remove again phase ramp before orthogonalization #
    ####################################################
    phase = phase - gridz * rampz - gridy * rampy - gridx * rampx

    avg_obj = amp * np.exp(1j * phase)  # here the phase is again wrapped in [-pi pi[

    del amp, phase, gridz, gridy, gridx, rampz, rampy, rampx
    gc.collect()

    ######################
    # centering of array #
    ######################
    if centering_method == "max":
        avg_obj = pu.center_max(avg_obj)
        # shift based on max value,
        # required if it spans across the edge of the array before COM
    elif centering_method == "com":
        avg_obj = pu.center_com(avg_obj)
    elif centering_method == "max_com":
        avg_obj = pu.center_max(avg_obj)
        avg_obj = pu.center_com(avg_obj)

    #######################
    #  save support & vti #
    #######################
    if prm.get("save_support", False):
        # to be used as starting support in phasing, hence still in the detector frame
        support = np.zeros((numz, numy, numx))
        support[abs(avg_obj) / abs(avg_obj).max() > 0.01] = 1
        # low threshold because support will be cropped by shrinkwrap during phasing
        np.savez_compressed(
            detector.savedir + "S" + str(scan) + "_support" + comment, obj=support
        )
        del support
        gc.collect()

    if prm.get("save_rawdata", False):
        np.savez_compressed(
            detector.savedir + "S" + str(scan) + "_raw_amp-phase" + comment,
            amp=abs(avg_obj),
            phase=np.angle(avg_obj),
        )

        # voxel sizes in the detector frame
        voxel_z, voxel_y, voxel_x = setup.voxel_sizes_detector(
            array_shape=original_size,
            tilt_angle=(
                prm.get("tilt_angle")
                * detector.preprocessing_binning[0]
                * detector.binning[0]
            ),
            pixel_x=detector.pixelsize_x,
            pixel_y=detector.pixelsize_y,
            verbose=True,
        )
        # save raw amp & phase to VTK
        # in VTK, x is downstream, y vertical, z inboard,
        # thus need to flip the last axis
        gu.save_to_vti(
            filename=os.path.join(
                detector.savedir, "S" + str(scan) + "_raw_amp-phase" + comment + ".vti"
            ),
            voxel_size=(voxel_z, voxel_y, voxel_x),
            tuple_array=(abs(avg_obj), np.angle(avg_obj)),
            tuple_fieldnames=("amp", "phase"),
            amplitude_threshold=0.01,
        )

    #########################################################
    # calculate q of the Bragg peak in the laboratory frame #
    #########################################################
    q_lab = (
        setup.q_laboratory
    )  # (1/A), in the laboratory frame z downstream, y vertical, x outboard
    qnorm = np.linalg.norm(q_lab)
    q_lab = q_lab / qnorm

    angle = simu.angle_vectors(
        ref_vector=[q_lab[2], q_lab[1], q_lab[0]],
        test_vector=axis_to_array_xyz[ref_axis_q],
    )
    print(
        f"\nNormalized diffusion vector in the laboratory frame (z*, y*, x*): "
        f"({q_lab[0]:.4f} 1/A, {q_lab[1]:.4f} 1/A, {q_lab[2]:.4f} 1/A)"
    )

    planar_dist = 2 * np.pi / qnorm  # qnorm should be in angstroms
    print(f"Wavevector transfer: {qnorm:.4f} 1/A")
    print(f"Atomic planar distance: {planar_dist:.4f} A")
    print(f"\nAngle between q_lab and {ref_axis_q} = {angle:.2f} deg")
    if debug:
        print(
            "Angle with y in zy plane = "
            f"{np.arctan(q_lab[0]/q_lab[1])*180/np.pi:.2f} deg"
        )
        print(
            "Angle with y in xy plane = "
            f"{np.arctan(-q_lab[2]/q_lab[1])*180/np.pi:.2f} deg"
        )
        print(
            "Angle with z in xz plane = "
            f"{180+np.arctan(q_lab[2]/q_lab[0])*180/np.pi:.2f} deg\n"
        )

    planar_dist = planar_dist / 10  # switch to nm

    #######################
    #  orthogonalize data #
    #######################
    print("\nShape before orthogonalization", avg_obj.shape, "\n")
    if data_frame == "detector":
        if debug:
            phase, _ = pu.unwrap(
                avg_obj,
                support_threshold=threshold_unwrap_refraction,
                debugging=True,
                reciprocal_space=False,
                is_orthogonal=False,
            )
            gu.multislices_plot(
                phase,
                width_z=2 * zrange,
                width_y=2 * yrange,
                width_x=2 * xrange,
                sum_frames=False,
                plot_colorbar=True,
                reciprocal_space=False,
                is_orthogonal=False,
                title="unwrapped phase before orthogonalization",
            )
            del phase
            gc.collect()

        if not prm.get("outofplane_angle") and not prm.get("inplane_angle"):
            print("Trying to correct detector angles using the direct beam")
            # corrected detector angles not provided
            if bragg_peak is None and detector.template_imagefile is not None:
                # Bragg peak position not provided, find it from the data
                data, _, _, _ = setup.diffractometer.load_check_dataset(
                    scan_number=scan,
                    detector=detector,
                    setup=setup,
                    frames_pattern=prm.get("frames_pattern"),
                    bin_during_loading=False,
                    flatfield=prm.get("flatfield"),
                    hotpixels=prm.get("hotpix_array"),
                    background=prm.get("background"),
                    normalize=prm.get("normalize_flux", "skip"),
                )
                bragg_peak = bu.find_bragg(
                    data=data,
                    peak_method="maxcom",
                    roi=detector.roi,
                    binning=None,
                )
                roi_center = (
                    bragg_peak[0],
                    bragg_peak[1] - detector.roi[0],  # no binning as in bu.find_bragg
                    bragg_peak[2] - detector.roi[2],  # no binning as in bu.find_bragg
                )
                bu.show_rocking_curve(
                    data,
                    roi_center=roi_center,
                    tilt_values=setup.incident_angles,
                    savedir=detector.savedir,
                )
            setup.correct_detector_angles(bragg_peak_position=bragg_peak)
            prm["outofplane_angle"] = setup.outofplane_angle
            prm["inplane_angle"] = setup.inplane_angle

        obj_ortho, voxel_size, transfer_matrix = setup.ortho_directspace(
            arrays=avg_obj,
            q_com=np.array([q_lab[2], q_lab[1], q_lab[0]]),
            initial_shape=original_size,
            voxel_size=fix_voxel,
            reference_axis=axis_to_array_xyz[ref_axis_q],
            fill_value=0,
            debugging=True,
            title="amplitude",
        )
        prm["transformation_matrix"] = transfer_matrix
    else:  # data already orthogonalized using xrayutilities
        # or the linearized transformation matrix
        obj_ortho = avg_obj
        try:
            print("Select the file containing QxQzQy")
            file_path = filedialog.askopenfilename(
                title="Select the file containing QxQzQy",
                initialdir=detector.savedir,
                filetypes=[("NPZ", "*.npz")],
            )
            npzfile = np.load(file_path)
            qx = npzfile["qx"]
            qy = npzfile["qy"]
            qz = npzfile["qz"]
        except FileNotFoundError:
            raise FileNotFoundError(
                "q values not provided, the voxel size cannot be calculated"
            )
        dy_real = (
            2 * np.pi / abs(qz.max() - qz.min()) / 10
        )  # in nm qz=y in nexus convention
        dx_real = (
            2 * np.pi / abs(qy.max() - qy.min()) / 10
        )  # in nm qy=x in nexus convention
        dz_real = (
            2 * np.pi / abs(qx.max() - qx.min()) / 10
        )  # in nm qx=z in nexus convention
        print(
            f"direct space voxel size from q values: ({dz_real:.2f} nm,"
            f" {dy_real:.2f} nm, {dx_real:.2f} nm)"
        )
        if fix_voxel:
            voxel_size = fix_voxel
            print(f"Direct space pixel size for the interpolation: {voxel_size} (nm)")
            print("Interpolating...\n")
            obj_ortho = pu.regrid(
                array=obj_ortho,
                old_voxelsize=(dz_real, dy_real, dx_real),
                new_voxelsize=voxel_size,
            )
        else:
            # no need to interpolate
            voxel_size = dz_real, dy_real, dx_real  # in nm

        if (
            data_frame == "laboratory"
        ):  # the object must be rotated into the crystal frame
            # before the strain calculation
            print("Rotating the object in the crystal frame for the strain calculation")

            amp, phase = util.rotate_crystal(
                arrays=(abs(obj_ortho), np.angle(obj_ortho)),
                is_orthogonal=True,
                reciprocal_space=False,
                voxel_size=voxel_size,
                debugging=(True, False),
                axis_to_align=q_lab[::-1],
                reference_axis=axis_to_array_xyz[ref_axis_q],
                title=("amp", "phase"),
            )

            obj_ortho = amp * np.exp(
                1j * phase
            )  # here the phase is again wrapped in [-pi pi[
            del amp, phase

    del avg_obj
    gc.collect()

    ######################################################
    # center the object (centering based on the modulus) #
    ######################################################
    print("\nCentering the crystal")
    obj_ortho = pu.center_com(obj_ortho)

    ####################
    # Phase unwrapping #
    ####################
    print("\nPhase unwrapping")
    phase, extent_phase = pu.unwrap(
        obj_ortho,
        support_threshold=threshold_unwrap_refraction,
        debugging=True,
        reciprocal_space=False,
        is_orthogonal=True,
    )
    amp = abs(obj_ortho)
    del obj_ortho
    gc.collect()

    #############################################
    # invert phase: -1*phase = displacement * q #
    #############################################
    if invert_phase:
        phase = -1 * phase

    ########################################
    # refraction and absorption correction #
    ########################################
    if correct_refraction:  # or correct_absorption:
        bulk = pu.find_bulk(
            amp=amp,
            support_threshold=threshold_unwrap_refraction,
            method=prm.get("optical_path_method", "threshold"),
            debugging=debug,
        )

        kin = setup.incident_wavevector
        kout = setup.exit_wavevector
        # kin and kout were calculated in the laboratory frame,
        # but after the geometric transformation of the crystal, this
        # latter is always in the crystal frame (for simpler strain calculation).
        # We need to transform kin and kout back
        # into the crystal frame (also, xrayutilities output is in crystal frame)
        kin = util.rotate_vector(
            vectors=[kin[2], kin[1], kin[0]],
            axis_to_align=axis_to_array_xyz[ref_axis_q],
            reference_axis=[q_lab[2], q_lab[1], q_lab[0]],
        )
        kout = util.rotate_vector(
            vectors=[kout[2], kout[1], kout[0]],
            axis_to_align=axis_to_array_xyz[ref_axis_q],
            reference_axis=[q_lab[2], q_lab[1], q_lab[0]],
        )

        # calculate the optical path of the incoming wavevector
        path_in = pu.get_opticalpath(
            support=bulk, direction="in", k=kin, debugging=debug
        )  # path_in already in nm

        # calculate the optical path of the outgoing wavevector
        path_out = pu.get_opticalpath(
            support=bulk, direction="out", k=kout, debugging=debug
        )  # path_our already in nm

        optical_path = path_in + path_out
        del path_in, path_out
        gc.collect()

        if correct_refraction:
            phase_correction = (
                2 * np.pi / (1e9 * setup.wavelength) * prm["dispersion"] * optical_path
            )
            phase = phase + phase_correction

            gu.multislices_plot(
                np.multiply(phase_correction, bulk),
                width_z=2 * zrange,
                width_y=2 * yrange,
                width_x=2 * xrange,
                sum_frames=False,
                plot_colorbar=True,
                vmin=0,
                vmax=np.nan,
                title="Refraction correction on the support",
                is_orthogonal=True,
                reciprocal_space=False,
            )
        correct_absorption = False
        if correct_absorption:
            amp_correction = np.exp(
                2 * np.pi / (1e9 * setup.wavelength) * prm["absorption"] * optical_path
            )
            amp = amp * amp_correction

            gu.multislices_plot(
                np.multiply(amp_correction, bulk),
                width_z=2 * zrange,
                width_y=2 * yrange,
                width_x=2 * xrange,
                sum_frames=False,
                plot_colorbar=True,
                vmin=1,
                vmax=1.1,
                title="Absorption correction on the support",
                is_orthogonal=True,
                reciprocal_space=False,
            )

        del bulk, optical_path
        gc.collect()

    ##############################################
    # phase ramp and offset removal (mean value) #
    ##############################################
    print("\nPhase ramp removal")
    amp, phase, _, _, _ = pu.remove_ramp(
        amp=amp,
        phase=phase,
        initial_shape=original_size,
        method=prm.get("phase_ramp_removal", "gradient"),
        amplitude_threshold=isosurface_strain,
        threshold_gradient=threshold_gradient,
        debugging=debug,
    )

    ########################
    # phase offset removal #
    ########################
    print("\nPhase offset removal")
    support = np.zeros(amp.shape)
    support[amp > isosurface_strain * amp.max()] = 1
    phase = pu.remove_offset(
        array=phase,
        support=support,
        offset_method=offset_method,
        phase_offset=phase_offset,
        offset_origin=offset_origin,
        title="Orthogonal phase",
        debugging=debug,
        reciprocal_space=False,
        is_orthogonal=True,
    )
    del support
    gc.collect()
    # Wrap the phase around 0 (no more offset)
    phase = util.wrap(
        obj=phase, start_angle=-extent_phase / 2, range_angle=extent_phase
    )

    ################################################################
    # calculate the strain depending on which axis q is aligned on #
    ################################################################
    print(f"\nCalculation of the strain along {ref_axis_q}")
    strain = pu.get_strain(
        phase=phase,
        planar_distance=planar_dist,
        voxel_size=voxel_size,
        reference_axis=ref_axis_q,
        extent_phase=extent_phase,
        method=prm.get("strain_method", "default"),
        debugging=debug,
    )

    ################################################
    # optionally rotates back the crystal into the #
    # laboratory frame (for debugging purpose)     #
    ################################################
    q_final = None
    if save_frame in {"laboratory", "lab_flat_sample"}:
        comment = comment + "_labframe"
        print("\nRotating back the crystal in laboratory frame")
        amp, phase, strain = util.rotate_crystal(
            arrays=(amp, phase, strain),
            axis_to_align=axis_to_array_xyz[ref_axis_q],
            voxel_size=voxel_size,
            is_orthogonal=True,
            reciprocal_space=False,
            reference_axis=[q_lab[2], q_lab[1], q_lab[0]],
            debugging=(True, False, False),
            title=("amp", "phase", "strain"),
        )
        # q_lab is already in the laboratory frame
        q_final = q_lab

    if save_frame == "lab_flat_sample":
        comment = comment + "_flat"
        print("\nSending sample stage circles to 0")
        (amp, phase, strain), q_final = setup.diffractometer.flatten_sample(
            arrays=(amp, phase, strain),
            voxel_size=voxel_size,
            q_com=q_lab[::-1],  # q_com needs to be in xyz order
            is_orthogonal=True,
            reciprocal_space=False,
            rocking_angle=setup.rocking_angle,
            debugging=(True, False, False),
            title=("amp", "phase", "strain"),
        )
    if save_frame == "crystal":
        # rotate also q_lab to have it along ref_axis_q,
        # as a cross-checkm, vectors needs to be in xyz order
        comment = comment + "_crystalframe"
        q_final = util.rotate_vector(
            vectors=q_lab[::-1],
            axis_to_align=axis_to_array_xyz[ref_axis_q],
            reference_axis=q_lab[::-1],
        )

    ###############################################
    # rotates the crystal e.g. for easier slicing #
    # of the result along a particular direction  #
    ###############################################
    # typically this is an inplane rotation, q should stay aligned with the axis
    # along which the strain was calculated
    if prm.get("align_axis", False):
        print("\nRotating arrays for visualization")
        amp, phase, strain = util.rotate_crystal(
            arrays=(amp, phase, strain),
            reference_axis=axis_to_array_xyz[prm["ref_axis"]],
            axis_to_align=prm["axis_to_align"],
            voxel_size=voxel_size,
            debugging=(True, False, False),
            is_orthogonal=True,
            reciprocal_space=False,
            title=("amp", "phase", "strain"),
        )
        # rotate q accordingly, vectors needs to be in xyz order
        q_final = util.rotate_vector(
            vectors=q_final[::-1],
            axis_to_align=axis_to_array_xyz[prm["ref_axis"]],
            reference_axis=prm["axis_to_align"],
        )

    q_final = q_final * qnorm
    print(
        f"\nq_final = ({q_final[0]:.4f} 1/A,"
        f" {q_final[1]:.4f} 1/A, {q_final[2]:.4f} 1/A)"
    )

    ##############################################
    # pad array to fit the output_size parameter #
    ##############################################
    if output_size is not None:
        amp = util.crop_pad(array=amp, output_shape=output_size)
        phase = util.crop_pad(array=phase, output_shape=output_size)
        strain = util.crop_pad(array=strain, output_shape=output_size)
    print(f"\nFinal data shape: {amp.shape}")

    ######################
    # save result to vtk #
    ######################
    print(
        f"\nVoxel size: ({voxel_size[0]:.2f} nm, {voxel_size[1]:.2f} nm,"
        f" {voxel_size[2]:.2f} nm)"
    )
    bulk = pu.find_bulk(
        amp=amp, support_threshold=isosurface_strain, method="threshold"
    )
    if save:
        prm["comment"] = comment
        np.savez_compressed(
            f"{detector.savedir}S{scan}_amp{phase_fieldname}strain{comment}",
            amp=amp,
            phase=phase,
            bulk=bulk,
            strain=strain,
            q_com=q_final,
            voxel_sizes=voxel_size,
            detector=detector.params,
            setup=setup.params,
            params=prm,
        )

        # save results in hdf5 file
        with h5py.File(
            f"{detector.savedir}S{scan}_amp{phase_fieldname}strain{comment}.h5", "w"
        ) as hf:
            out = hf.create_group("output")
            par = hf.create_group("params")
            out.create_dataset("amp", data=amp)
            out.create_dataset("bulk", data=bulk)
            out.create_dataset("phase", data=phase)
            out.create_dataset("strain", data=strain)
            out.create_dataset("q_com", data=q_final)
            out.create_dataset("voxel_sizes", data=voxel_size)
            par.create_dataset("detector", data=str(detector.params))
            par.create_dataset("setup", data=str(setup.params))
            par.create_dataset("parameters", data=str(prm))

        # save amp & phase to VTK
        # in VTK, x is downstream, y vertical, z inboard,
        # thus need to flip the last axis
        gu.save_to_vti(
            filename=os.path.join(
                detector.savedir,
                "S"
                + str(scan)
                + "_amp-"
                + phase_fieldname
                + "-strain"
                + comment
                + ".vti",
            ),
            voxel_size=voxel_size,
            tuple_array=(amp, bulk, phase, strain),
            tuple_fieldnames=("amp", "bulk", phase_fieldname, "strain"),
            amplitude_threshold=0.01,
        )

    ######################################
    # estimate the volume of the crystal #
    ######################################
    amp = amp / amp.max()
    temp_amp = np.copy(amp)
    temp_amp[amp < isosurface_strain] = 0
    temp_amp[np.nonzero(temp_amp)] = 1
    volume = temp_amp.sum() * reduce(lambda x, y: x * y, voxel_size)  # in nm3
    del temp_amp
    gc.collect()

    ##############################
    # plot slices of the results #
    ##############################
    pixel_spacing = [tick_spacing / vox for vox in voxel_size]
    print(
        "\nPhase extent without / with thresholding the modulus "
        f"(threshold={isosurface_strain}): {phase.max()-phase.min():.2f} rad, "
        f"{phase[np.nonzero(bulk)].max()-phase[np.nonzero(bulk)].min():.2f} rad"
    )
    piz, piy, pix = np.unravel_index(phase.argmax(), phase.shape)
    print(
        f"phase.max() = {phase[np.nonzero(bulk)].max():.2f} "
        f"at voxel ({piz}, {piy}, {pix})"
    )
    strain[bulk == 0] = np.nan
    phase[bulk == 0] = np.nan

    # plot the slice at the maximum phase
    gu.combined_plots(
        (phase[piz, :, :], phase[:, piy, :], phase[:, :, pix]),
        tuple_sum_frames=False,
        tuple_sum_axis=0,
        tuple_width_v=None,
        tuple_width_h=None,
        tuple_colorbar=True,
        tuple_vmin=np.nan,
        tuple_vmax=np.nan,
        tuple_title=("phase at max in xy", "phase at max in xz", "phase at max in yz"),
        tuple_scale="linear",
        cmap=my_cmap,
        is_orthogonal=True,
        reciprocal_space=False,
    )

    # bulk support
    fig, _, _ = gu.multislices_plot(
        bulk,
        sum_frames=False,
        title="Orthogonal bulk",
        vmin=0,
        vmax=1,
        is_orthogonal=True,
        reciprocal_space=False,
    )
    fig.text(0.60, 0.45, "Scan " + str(scan), size=20)
    fig.text(
        0.60,
        0.40,
        "Bulk - isosurface=" + str("{:.2f}".format(isosurface_strain)),
        size=20,
    )
    plt.pause(0.1)
    if save:
        plt.savefig(detector.savedir + "S" + str(scan) + "_bulk" + comment + ".png")

    # amplitude
    fig, _, _ = gu.multislices_plot(
        amp,
        sum_frames=False,
        title="Normalized orthogonal amp",
        vmin=0,
        vmax=1,
        tick_direction=tick_direction,
        tick_width=tick_width,
        tick_length=tick_length,
        pixel_spacing=pixel_spacing,
        plot_colorbar=True,
        is_orthogonal=True,
        reciprocal_space=False,
    )
    fig.text(0.60, 0.45, f"Scan {scan}", size=20)
    fig.text(
        0.60,
        0.40,
        f"Voxel size=({voxel_size[0]:.1f}, {voxel_size[1]:.1f}, "
        f"{voxel_size[2]:.1f}) (nm)",
        size=20,
    )
    fig.text(0.60, 0.35, f"Ticks spacing={tick_spacing} nm", size=20)
    fig.text(0.60, 0.30, f"Volume={int(volume)} nm3", size=20)
    fig.text(0.60, 0.25, "Sorted by " + sort_method, size=20)
    fig.text(0.60, 0.20, f"correlation threshold={correlation_threshold}", size=20)
    fig.text(0.60, 0.15, f"average over {avg_counter} reconstruction(s)", size=20)
    fig.text(0.60, 0.10, f"Planar distance={planar_dist:.5f} nm", size=20)
    if prm.get("get_temperature", False):
        temperature = pu.bragg_temperature(
            spacing=planar_dist * 10,
            reflection=prm["reflection"],
            spacing_ref=prm.get("reference_spacing"),
            temperature_ref=prm.get("reference_temperature"),
            use_q=False,
            material="Pt",
        )
        fig.text(0.60, 0.05, f"Estimated T={temperature} C", size=20)
    if save:
        plt.savefig(detector.savedir + f"S{scan}_amp" + comment + ".png")

    # amplitude histogram
    fig, ax = plt.subplots(1, 1)
    ax.hist(amp[amp > 0.05 * amp.max()].flatten(), bins=250)
    ax.set_ylim(bottom=1)
    ax.tick_params(
        labelbottom=True,
        labelleft=True,
        direction="out",
        length=tick_length,
        width=tick_width,
    )
    ax.spines["right"].set_linewidth(1.5)
    ax.spines["left"].set_linewidth(1.5)
    ax.spines["top"].set_linewidth(1.5)
    ax.spines["bottom"].set_linewidth(1.5)
    fig.savefig(detector.savedir + f"S{scan}_histo_amp" + comment + ".png")

    # phase
    fig, _, _ = gu.multislices_plot(
        phase,
        sum_frames=False,
        title="Orthogonal displacement",
        vmin=-prm.get("phase_range", np.pi / 2),
        vmax=prm.get("phase_range", np.pi / 2),
        tick_direction=tick_direction,
        cmap=my_cmap,
        tick_width=tick_width,
        tick_length=tick_length,
        pixel_spacing=pixel_spacing,
        plot_colorbar=True,
        is_orthogonal=True,
        reciprocal_space=False,
    )
    fig.text(0.60, 0.30, f"Scan {scan}", size=20)
    fig.text(
        0.60,
        0.25,
        f"Voxel size=({voxel_size[0]:.1f}, {voxel_size[1]:.1f}, "
        f"{voxel_size[2]:.1f}) (nm)",
        size=20,
    )
    fig.text(0.60, 0.20, f"Ticks spacing={tick_spacing} nm", size=20)
    fig.text(0.60, 0.15, f"average over {avg_counter} reconstruction(s)", size=20)
    if half_width_avg_phase > 0:
        fig.text(
            0.60, 0.10, f"Averaging over {2*half_width_avg_phase+1} pixels", size=20
        )
    else:
        fig.text(0.60, 0.10, "No phase averaging", size=20)
    if save:
        plt.savefig(detector.savedir + f"S{scan}_displacement" + comment + ".png")

    # strain
    fig, _, _ = gu.multislices_plot(
        strain,
        sum_frames=False,
        title="Orthogonal strain",
        vmin=-prm.get("strain_range", 0.002),
        vmax=prm.get("strain_range", 0.002),
        tick_direction=tick_direction,
        tick_width=tick_width,
        tick_length=tick_length,
        plot_colorbar=True,
        cmap=my_cmap,
        pixel_spacing=pixel_spacing,
        is_orthogonal=True,
        reciprocal_space=False,
    )
    fig.text(0.60, 0.30, f"Scan {scan}", size=20)
    fig.text(
        0.60,
        0.25,
        f"Voxel size=({voxel_size[0]:.1f}, "
        f"{voxel_size[1]:.1f}, {voxel_size[2]:.1f}) (nm)",
        size=20,
    )
    fig.text(0.60, 0.20, f"Ticks spacing={tick_spacing} nm", size=20)
    fig.text(0.60, 0.15, f"average over {avg_counter} reconstruction(s)", size=20)
    if half_width_avg_phase > 0:
        fig.text(
            0.60, 0.10, f"Averaging over {2*half_width_avg_phase+1} pixels", size=20
        )
    else:
        fig.text(0.60, 0.10, "No phase averaging", size=20)
    if save:
        plt.savefig(detector.savedir + f"S{scan}_strain" + comment + ".png")
Example #3
0
def check_cdi_angle(data, mask, cdi_angle, frames_logical, debugging=False):
    """
    Check for overlaps of the sample rotation motor position in forward CDI experiment.

    It checks if there is no overlap in the measurement angles, and crops it otherwise.
    Flip the rotation direction to convert sample angles into detector angles. Update
    data, mask and frames_logical accordingly.

    :param data: 3D forward CDI dataset before gridding.
    :param mask: 3D mask
    :param cdi_angle: array of measurement sample angles in degrees
    :param frames_logical: array of initial length the number of measured frames.
     In case of padding the length changes. A frame whose index is set to 1 means
     that it is used, 0 means not used, -1 means padded (added) frame.
    :param debugging: True to have more printed comments
    :return: updated data, mask, detector cdi_angle, frames_logical
    """
    valid.valid_ndarray(arrays=(data, mask), ndim=3)
    detector_angle = np.zeros(len(cdi_angle))
    # flip the rotation axis in order to compensate the rotation of the Ewald sphere
    # due to sample rotation
    print(
        "Reverse the rotation direction to compensate the rotation of the Ewald sphere"
    )
    for idx, item in enumerate(cdi_angle):
        detector_angle[idx] = cdi_angle[0] - (item - cdi_angle[0])

    wrap_angle = util.wrap(obj=detector_angle,
                           start_angle=detector_angle.min(),
                           range_angle=180)
    for idx, item in enumerate(wrap_angle):
        duplicate = np.isclose(wrap_angle[:idx], item, rtol=1e-06,
                               atol=1e-06).sum()
        # duplicate will be different from 0 if there is a duplicated angle
        frames_logical[idx] = frames_logical[idx] * (
            duplicate == 0)  # set frames_logical to 0 if duplicated angle

    if debugging:
        print("frames_logical after checking duplicated angles:\n",
              frames_logical)

    # find first duplicated angle
    try:
        index_duplicated = np.where(frames_logical == 0)[0][0]
        # change the angle by a negligeable amount
        # to still be able to use it for interpolation
        if cdi_angle[1] - cdi_angle[0] > 0:
            detector_angle[
                index_duplicated] = detector_angle[index_duplicated] - 0.0001
        else:
            detector_angle[
                index_duplicated] = detector_angle[index_duplicated] + 0.0001
        print(
            "RegularGridInterpolator cannot take duplicated values: shifting frame",
            index_duplicated,
            "by 1/10000 degrees for the interpolation",
        )

        frames_logical[index_duplicated] = 1
    except IndexError:  # no duplicated angle
        print("no duplicated angle")

    data = data[np.nonzero(frames_logical)[0], :, :]
    mask = mask[np.nonzero(frames_logical)[0], :, :]
    detector_angle = detector_angle[np.nonzero(frames_logical)]
    return data, mask, detector_angle, frames_logical
                axis=(0, 1, 2),
            )

        support = np.zeros(amp.shape)
        support[amp > isosurface * amp.max()] = 1
        zcom, ycom, xcom = center_of_mass(support)
        zcom, ycom, xcom = [
            int(np.rint(zcom)),
            int(np.rint(ycom)),
            int(np.rint(xcom)),
        ]  # used for the reference phase
        phase, extent_phase = pu.unwrap(amp * np.exp(1j * phase),
                                        support_threshold=0.05,
                                        debugging=debug)
        phase = util.wrap(phase,
                          start_angle=-extent_phase / 2,
                          range_angle=extent_phase)
        phase = phase - phase[zcom, ycom, xcom]
        phase, extent_phase = pu.unwrap(amp * np.exp(1j * phase),
                                        support_threshold=0.05,
                                        debugging=debug)
        phase = util.wrap(
            phase, start_angle=-extent_phase / 2,
            range_angle=extent_phase)  # rewrap after modifying phase
        ref_amp = np.copy(amp)
    else:  # align it with the reference object
        shiftz, shifty, shiftx = reg.getimageregistration(ref_amp,
                                                          amp,
                                                          precision=1000)
        print("Shift of array", index, "with the reference array:", shiftz,
              shifty, shiftx)