def check_column_for_alignment(cluster_strategy, rotation_adjustment, channel, snr, sequencing_chip, um_per_pixel, fia,
end_tiles, column, possible_tile_keys, h5_filename):
base_name = os.path.splitext(h5_filename)[0]
with h5py.File(h5_filename) as h5:
grid = GridImages(h5, channel)
# we assume odd numbers of rows, and good enough for now
if grid.height > 2:
center_row = grid.height / 2
rows_to_check = (center_row, center_row + 1, center_row - 1)
else:
# just one or two rows, might as well try them all
rows_to_check = tuple([i for i in range(grid.height)])
for row in rows_to_check:
image = grid.get(row, column)
if image is None:
log.warn("Could not find an image for %s Row %d Column %d" % (base_name, row, column))
return
log.debug("Aligning %s Row %d Column %d against PhiX" % (base_name, row, column))
fia = process_alignment_image(cluster_strategy, rotation_adjustment, snr, sequencing_chip, base_name, um_per_pixel, image, possible_tile_keys, deepcopy(fia))
if fia.hitting_tiles:
log.debug("%s aligned to at least one tile!" % image.index)
# because of the way we iterate through the images, if we find one that aligns,
# we can just stop because that gives us the outermost column of images and the
# outermost FastQ tile
end_tiles[h5_filename] = [tile.key for tile in fia.hitting_tiles], image.column
break
del fia
gc.collect()
def iterate_all_images(h5_filenames, end_tiles, channel, path_info):
# We need an iterator over all images to feed the parallel processes. Since each image is
# processed independently and in no particular order, we need to return information in addition
# to the image itself that allow files to be written in the correct place and such
for h5_filename in h5_filenames:
base_name = os.path.splitext(h5_filename)[0]
with h5py.File(h5_filename) as h5:
grid = GridImages(h5, channel)
min_column, max_column, tile_map = end_tiles[h5_filename]
for column in range(min_column, max_column):
for row in range(grid._height):
image = grid.get(row, column)
if image is None:
continue
stats_path = os.path.join(
path_info.results_directory, base_name,
'{}_stats.txt'.format(image.index))
alignment_path = os.path.join(
path_info.results_directory, base_name,
'{}_all_read_rcs.txt'.format(image.index))
already_aligned = alignment_is_complete(
stats_path) and os.path.exists(alignment_path)
if already_aligned:
log.debug(
"Image already aligned/checkpointed: {}/{}".format(
h5_filename, image.index))
continue
yield row, column, channel, h5_filename, tile_map[
image.column], base_name
def load_image(h5_filename, channel, row, column):
with h5py.File(h5_filename) as h5:
grid = GridImages(h5, channel)
return grid.get(row, column)