Custom Readers

You can define your own PIMS reader with a minimal amount of customized code.


By subclassing FramesSequence, you get PIMS’s lazy-loading and slicing behavior for free. You have to provide:

  • a method for reading a single frame into a numpy array
  • a method for knowing the length of the sequence
  • properties giving the shape and the numpy data type of each frame

Basic Example

from pims import FramesSequence, Frame

class MyReader(FramesSequence):

    def __init__(self, filename):
        self.filename = filename
        self._len =  # however many frames there will be
        self._dtype =  # the numpy datatype of the frames
        self._frame_shape =  # the shape, like (512, 512), of an
                             # individual frame -- maybe get this by
                             # opening the first frame
        # Do whatever setup you need to do to be able to quickly access
        # individual frames later.

    def get_frame(self, i):
        # Access the data you need and get it into a numpy array.
        # Then return a Frame like so:
        return Frame(my_numpy_array, frame_no=i)

     def __len__(self):
         return self._len

     def frame_shape(self):
         return self._frame_shape

     def pixel_type(self):
         return self._dtype

The __init__ method is completely customizable. It can take whatever arguments make sense for your file format.

Optionally, you might also wish to customize the __repr__ and define a close method, which the base class calls when exiting a context manager.

Plugging into PIMS’s open function

The function dispatches to a PIMS reader based on the file extension. To associate your reader with particle file extensions, add a class_exts class method. For example, the following code will invoke MyReader to open files ending in .xyz.

class MyReader(FramesSequence):


    def class_exts(cls):
        return {'xyz'} | super(MyReader, cls).class_exts()

New readers can be defined or imported at any time. They will be detected the next time is called, as it searches all subclasses of FramesSequence for an eligible reader.

To prioritize readers, a field class_priority can optionally be given to the reader. A higher priority will be chosen over a lower priority. Default for all readers is 10.

Example Demonstrating Generality of PIMS Design

Here is a reader that extracts tiles from a tiled image or “sprite sheet.”

class SpriteSheet(FramesSequence):
This is a class for providing an easy interface into
a sprite sheet of uniformly sized images.

sheet : ndarray
    The sprite sheet.  It should consist of N paneled images.  In
    this version all possible positions have an image, this may be changed
    to limit the number of acessable images in the sheet to be less than
    the possible number.
rows : int
cols : int
    The number of rows and columns of sprites.
    The sprite size is computed from these + the shape of the sheet.

def __init__(self, sheet, rows, cols):
    self._sheet = sheet
    sheet_height, sheet_width = sheet.shape
    if sheet_width % cols != 0:
        raise ValueError("Sheet width not evenly divisible by cols")
    if sheet_height % rows != 0:
        raise ValueError("Sheet height not evenly divisible by rows")

    self._sheet_shape = (rows, cols)

    self._im_sz = sheet_height // rows, sheet_width // cols
    self._sprite_height, self._sprite_width = self._im_sz

    self._dtype = sheet.dtype

def pixel_type(self):
    return self._dtype

def frame_shape(self):
    return self._im_sz

def __len__(self):

def get_frame(self, n):
    r, c = np.unravel_index(n, self._sheet_shape)
    slc_r = slice(r*self._sprite_height, (r+1)*self._sprite_height)
    slc_c = slice(c*self._sprite_width, (c+1)*self._sprite_width)
    tmp = self._sheet[slc_r, slc_c]
    return Frame(self.process_func(tmp), frame_no=n)