Source code for SimExLite.DiffractionData.EMCFormat

import h5py
import numpy as np
import os
from pathlib import Path
from tqdm.autonotebook import tqdm
from scipy.sparse import csr_matrix
from libpyvinyl.BaseFormat import BaseFormat
from . import writeemc, DetectorEMC
from SimExLite.utils.io import parseIndex
from SimExLite.utils.io import UnknownFileTypeError


class EMCFormat(BaseFormat):
    """Defines I/O for the EMC (Dragonfly orientation recovery) format ."""

    def __init__(self) -> None:
        super().__init__()

    @classmethod
    def format_register(self):
        key = "EMC"
        description = "EMC photon format for DiffractionData"
        file_extension = [".h5", ".emc"]
        read_kwargs = ["index", "pattern_shape"]
        write_kwargs = [""]
        return self._create_format_register(
            key, description, file_extension, read_kwargs, write_kwargs
        )

    @staticmethod
    def direct_convert_formats():
        # Assume the format can be converted directly to the formats supported by these classes:
        # AFormat, BFormat
        # Redefine this `direct_convert_formats` for a concrete format class
        return []

    @classmethod
    def read(cls, filename: str, index=None, pattern_shape=None) -> dict:
        """Read diffraction patterns into an array from a file."""
        data_dict = {}

        if pattern_shape is None:
            frame_shape_guessed = guessFrameShape(filename)
            raise ValueError(
                f"read() missing 'pattern_shape' argument, it might be ({frame_shape_guessed}, {frame_shape_guessed})."
            )

        index = parseIndex(index)
        arr_size = len(np.arange(getPatternTotal(filename))[index])
        if isEMCH5(filename):
            ireadPattern = ireadPattern_h5
            # Flush to print it before tqdm
            print("Reading EMC h5 data...", flush=True)
        elif isEMCBinary(filename):
            ireadPattern = ireadPattern_binary
            # Flush to print it before tqdm
            print("Reading EMC binary data...", flush=True)
        else:
            raise UnknownFileTypeError(
                "This is not an EMC file, please provide the correct file type."
            )

        arr = np.zeros((arr_size, pattern_shape[0], pattern_shape[1]))
        # if isinstance(index, (slice, str)):
        with tqdm(total=arr_size) as progress_bar:
            for i, pattern in enumerate(ireadPattern(filename, index, pattern_shape)):
                arr[i] = pattern
                progress_bar.update(1)  # update progress

        data_dict["img_array"] = arr
        # There is no quaternion in EMC pattern (?)
        data_dict["quaternions"] = None

        # TODO: convert EMC geom into DiffractionData/EXtra-geom geom.
        data_dict["geom"] = None

        # There is no distance in EMC pattern (?)
        data_dict["distance"] = None
        # good_pixel = 1, bad_pixel = 0.
        # There is no pixel_mask in EMC pattern (?)
        # data_dict["pixel_mask"] = pixel_mask
        # There is no beam information in EMC pattern (?)
        # data_dict["beam"] = beam

        return data_dict

    @classmethod
    def write(cls, object, filename: str, key: str = None):
        """Save the data with the `filename`."""
        data_dict = object.get_data()
        arr = data_dict["img_array"]
        img_shape = arr[0].shape
        emcwriter = writeemc.EMCWriter(filename, arr[0].shape[0] * arr[0].shape[1])
        for photons in tqdm(arr):
            emcwriter.write_frame(photons.astype(np.int32).ravel())
        emcwriter.finish_write()
        if key is None:
            original_key = object.key
            key = original_key + "_to_EMCFormat"
        return object.from_file(filename, cls, key, pattern_shape=img_shape)


def isEMCH5(fn):
    """If the data is a EMC HDF5 file"""
    try:
        with h5py.File(fn, "r") as h5:
            # If the h5 file has these keys
            if h5.keys() >= {"count_multi", "num_pix", "place_multi", "place_ones"}:
                return True
            else:
                return False
    except OSError:
        return False


# Essential
def ireadPattern_h5(filename, index=None, pattern_shape=None):
    """Iterator for reading diffraction patterns from a file."""
    index = parseIndex(index)
    pattern_total = getPatternTotal(filename)
    indices = np.arange(pattern_total)[index]
    for i in indices:
        yield getFrameArray(filename, i).reshape(pattern_shape)


# Essential
def ireadPattern_binary(filename, index=None, pattern_shape=None):
    """Iterator for reading diffraction patterns from a file."""
    index = parseIndex(index)
    pattern_total = getPatternTotal(filename)
    indices = np.arange(pattern_total)[index]
    for i in indices:
        yield getFrameArrayBinary(filename, i).reshape(pattern_shape)


def getPatternTotal(filename):
    """The total number of diffraction patterns in the EMC photon file"""
    if isEMCH5(filename):
        with h5py.File(filename, "r") as h5:
            npattern = len(h5["count_multi"])
        return npattern
    else:
        with open(filename, "rb") as fptr:
            num_data = np.fromfile(fptr, dtype="i4", count=1)[0]
        return num_data


def isEMCBinary(fn):
    """If the data is a EMC HDF5 file"""
    try:
        pdict = parse_binaryheader(fn)
        if len(pdict["ones_accum"]) > 0 and len(pdict["multi_accum"]) > 0:
            return True
    except OSError:
        return False


class PatternsSOne:
    """A class to store the EMC photon sparse data. `Format introduction
    <https://github.com/duaneloh/Dragonfly/wiki/Data-stream-simulator#make_data>`_

    Args:
        num_pix (int): Number of pixels per pattern
        ones (numpy.1darray): Number of one-photon events in each pattern
        multi (numpy.1darray): Number of multi-photon events in each pattern
        place_ones (numpy.1darray): The locations of the single photon pixels in each pattern
        place_multi (numpy.1darray): The locations of the multiple photon pixels in each pattern
        count_multi (numpy.1darray): Number of photons in each of those multiple photon pixels
    """

    ATTRS = ["ones", "multi", "place_ones", "place_multi", "count_multi"]

    def __init__(
        self,
        num_pix: int,
        ones: np.ndarray,
        multi: np.ndarray,
        place_ones: np.ndarray,
        place_multi: np.ndarray,
        count_multi: np.ndarray,
    ) -> None:
        self.num_pix = num_pix
        self._ones = ones
        self._multi = multi
        self._place_ones = place_ones
        self._place_multi = place_multi
        self._count_multi = count_multi
        self._ones_idx = np.zeros(self.num_data + 1, dtype=self._ones.dtype)
        np.cumsum(self._ones, out=self._ones_idx[1:])
        self._multi_idx = np.zeros(self.num_data + 1, dtype=self._multi.dtype)
        np.cumsum(self._multi, out=self._multi_idx[1:])

    def __len__(self):
        return self.num_data

    @property
    def num_data(self):
        return len(self._ones)

    @property
    def shape(self):
        """Return (number_of_patterns, number_of_pixels_per_pattern)"""
        return self.num_data, self.num_pix

    def write(self, path) -> None:
        with Path(path).open("wb") as fptr:
            header = np.zeros((256), dtype="i4")
            header[:2] = [self.num_data, self.num_pix]
            header.tofile(fptr)
            for g in PatternsSOne.ATTRS:
                self.attrs(g).astype("i4").tofile(fptr)

    def attrs(self, g):
        if g == "ones_idx":
            return self._ones_idx
        if g == "multi_idx":
            return self._multi_idx
        if g == "ones":
            return self._ones
        if g == "multi":
            return self._multi
        if g == "place_ones":
            return self._place_ones
        if g == "place_multi":
            return self._place_multi
        if g == "count_multi":
            return self._count_multi
        raise ValueError(f"What is {g}?")

    def _get_sparse_ones(self) -> csr_matrix:
        _one = np.ones(1, "i4")
        _one = np.lib.stride_tricks.as_strided(  # type: ignore
            _one, shape=(self._place_ones.shape[0],), strides=(0,)
        )
        return csr_matrix((_one, self._place_ones, self._ones_idx), shape=self.shape)

    def _get_sparse_multi(self) -> csr_matrix:
        return csr_matrix(
            (self._count_multi, self._place_multi, self._multi_idx), shape=self.shape
        )

    def todense(self) -> np.ndarray:
        """
        To dense ndarray
        """
        return np.squeeze(
            self._get_sparse_ones().todense() + self._get_sparse_multi().todense()
        )


def dense_to_PatternsSOne(arr: np.ndarray) -> PatternsSOne:
    """Convert diffraction pattern array data to EMC sparse data

    Args:
        arr (np.2darray): A multi-snapshot array with diffraction patterns flattened

    Returns:
        PatternsSOne: EMC photon sparse data
    """
    mask_one = arr == 1
    mask_multi = arr > 1
    place_ones = np.where(mask_one)[1]
    pmask, place_multi = np.where(mask_multi)  # pmask: pattern mask
    return PatternsSOne(
        arr.shape[1],
        np.sum(mask_one, axis=1, dtype=np.int32),
        np.sum(mask_multi, axis=1, dtype=np.int32),
        place_ones.astype(np.int32),
        place_multi.astype(np.int32),
        arr[pmask, place_multi].astype(np.int32),
    )


def parse_bin_PatternsSOne(fn: str):
    """Parse a EMC sparse binary file

    Args:
        fn (str): The name of the sparse binary file

    Returns:
        PatternsSOne: EMC photon sparse data
    """
    path = Path(fn)
    with path.open("rb") as fin:
        num_data = np.fromfile(fin, dtype=np.int32, count=1)[0]
        start, end = 0, num_data
        num_pix = np.fromfile(fin, dtype=np.int32, count=1)[0]
        fin.seek(1024)
        ones = np.fromfile(fin, dtype=np.int32, count=num_data)
        multi = np.fromfile(fin, dtype=np.int32, count=num_data)
        fin.seek(4 * ones[:start].sum(), os.SEEK_CUR)
        place_ones = np.fromfile(fin, dtype=np.int32, count=ones[start:end].sum())
        fin.seek(4 * (ones[end:].sum() + multi[:start].sum()), os.SEEK_CUR)
        sum_multi = multi[start:end].sum()
        place_multi = np.fromfile(fin, dtype=np.int32, count=sum_multi)
        fin.seek(4 * (multi[end:].sum() + multi[:start].sum()), os.SEEK_CUR)
        count_multi = np.fromfile(fin, dtype=np.int32, count=sum_multi)
        fin.seek(4 * multi[end:].sum(), os.SEEK_CUR)
        if fin.read(1):
            raise Exception(f"Error when parsing {fn}")
    ones = ones[start:end]
    multi = multi[start:end]
    return PatternsSOne(
        num_pix,
        ones,
        multi,
        place_ones,
        place_multi,
        count_multi,
    )


def readH5frame(fname, frame_num):
    with h5py.File(fname, "r") as fptr:
        num_pix = fptr["num_pix"][()][0]
        place_ones = fptr["place_ones"][frame_num]
        place_multi = fptr["place_multi"][frame_num]
        count_multi = fptr["count_multi"][frame_num]
        ones = np.array([len(place_ones)])
        multi = np.array([len(place_multi)])
    return num_pix, ones, multi, place_ones, place_multi, count_multi


def readBinaryframe(fname, frame_num):
    pdict = parse_binaryheader(fname)
    num_pix = pdict["num_pix"]
    with open(fname, "rb") as fptr:
        num_data = np.fromfile(fptr, dtype="i4", count=1)[0]

        accum = [pdict["ones_accum"], pdict["multi_accum"]]
        offset = [0, 0]
        size = [0, 0]

        if frame_num == 0:
            size = [accum[0][frame_num], accum[1][frame_num]]
        else:
            offset = [accum[0][frame_num - 1], accum[1][frame_num - 1]]
            size[0] = accum[0][frame_num] - accum[0][frame_num - 1]
            size[1] = accum[1][frame_num] - accum[1][frame_num - 1]

        fptr.seek(1024 + num_data * 8 + offset[0] * 4, 0)
        place_ones = np.fromfile(fptr, dtype="i4", count=size[0])
        fptr.seek(1024 + num_data * 8 + accum[0][-1] * 4 + offset[1] * 4, 0)
        place_multi = np.fromfile(fptr, dtype="i4", count=size[1])
        fptr.seek(
            1024 + num_data * 8 + accum[0][-1] * 4 + accum[1][-1] * 4 + offset[1] * 4, 0
        )
        count_multi = np.fromfile(fptr, dtype="i4", count=size[1])
        ones = np.array([len(place_ones)])
        multi = np.array([len(place_multi)])
    return num_pix, ones, multi, place_ones, place_multi, count_multi


def guessFrameShape(filename: str):
    """Guess the frame shape of a EMC file. It can be a binary or HDF5 file.

    Args:
        filename (str): The filename.
    """
    if isEMCH5(filename):
        num_pix, *_ = readH5frame(filename, 1)
    elif isEMCBinary(filename):
        num_pix, *_ = readBinaryframe(filename, 1)
    else:
        raise UnknownFileTypeError(
            "This is not an EMC file, please provide the correct file type."
        )
    return int(np.sqrt(num_pix))


def getFrameArray(fn, idx=0):
    """Get a flatten diffraction array from a EMC HDF file"""
    sPattern = PatternsSOne(*readH5frame(fn, idx))
    data = sPattern.todense()
    return data


def getFrameArrayBinary(fn, idx=0):
    """Get a flatten diffraction array from a EMC binary file"""
    sPattern = PatternsSOne(*readBinaryframe(fn, idx))
    data = sPattern.todense()
    return data


def parse_binaryheader(fname):
    pdict = {}
    with open(fname, "rb") as fptr:
        num_data = np.fromfile(fptr, dtype="i4", count=1)[0]
        pdict["num_pix"] = np.fromfile(fptr, dtype="i4", count=1)[0]
        fptr.seek(1024, 0)
        ones = np.fromfile(fptr, dtype="i4", count=num_data)
        multi = np.fromfile(fptr, dtype="i4", count=num_data)
    pdict["num_data"] = num_data
    pdict["ones_accum"] = np.cumsum(ones)
    pdict["multi_accum"] = np.cumsum(multi)
    return pdict


def writeEMCGeom(
    out_fn: str,
    det_dist: float,
    dets_x: int,
    dets_y: int,
    pix_size: float,
    in_wavelength: float,
    stoprad: float,
):
    """Get EMC geometry from several parameters.

    Args:
        out_fn (str): Output filename
        det_dist (float): Sample to detector distance (mm)
        dets_x (int): Number of pixels in x direction
        dets_y (int): Number of pixels in y direction
        pix_size (float): Pixel size (mm)
        in_wavelength (float): X-ray wavelength (angstrom)
        stoprad (float): Beamstop radius in pixels
    """
    # Reference: https://github.com/JunCEEE/Dragonfly/blob/8e9075818f00f5d2c45756d2b98803509be67cf0/utils/convert/geomtodet.py#L23
    # Sample to detector distance
    # width number of pixels
    # dets_x = geom["mask"].shape[1]
    # height number of pixels
    # dets_y = geom["mask"].shape[0]
    # pixel size
    # pix_size = geom["pixelSize"] * 1e3  # milimeter
    # wavelength
    # in_wavelength = beam.get_wavelength(unit="angstrom")
    # Radius of curvature of the Ewald sphere in voxels. See:
    # https://github.com/duaneloh/Dragonfly/wiki/Configuration-parameters-for-experimental-data#parameters-
    ewald_rad = det_dist / pix_size

    q_pm = writeemc.compute_q_params(
        det_dist, dets_x, dets_y, pix_size, in_wavelength, ewald_rad
    )
    # q_sep = 2sin(min_angle)/lambda
    y, x = np.indices((dets_y, dets_x))
    center_x = (dets_x - 1) / 2
    center_y = (dets_y - 1) / 2
    y = y - center_y
    x = x - center_x
    z = det_dist / pix_size
    det = DetectorEMC.Detector()
    qscaling = 1.0 / in_wavelength / q_pm["q_sep"]
    norm = np.sqrt(x * x + y * y + z * z)
    det.qx = x * qscaling / norm
    det.qy = y * qscaling / norm
    det.qz = qscaling * (z / norm - 1.0)
    det.corr = det_dist / np.power(norm, 3.0)
    # x polorization
    det.corr *= writeemc.compute_polarization("x", x, y, norm)
    radius = np.sqrt(x * x + y * y)
    rmax = min(np.abs(x.max()), np.abs(x.min()), np.abs(y.max()), np.abs(y.min()))
    det.raw_mask = np.zeros(det.corr.shape, dtype="u1")
    det.raw_mask[radius > rmax] = 1
    det.raw_mask[radius < stoprad] = 2
    det.detd = det_dist
    det.ewald_rad = ewald_rad
    print("Writing detector file to", out_fn)
    det.write(out_fn)


def write_emc_balcklist(fn: str, sel: list, total: int):
    """Write a emc black list selecting patterns to be included in dragonfly reconstruction.

    Args:
        fn (str): The file name of the blacklist.
        sel (list): The selection indices of the patterns to be included in reconstruction (set to 0 in the blacklist).
        total (int): The total number of the dataset.
    """
    black_list = np.ones(total)
    black_list[sel] = 0
    np.savetxt(fn, black_list, fmt="%d")