Base class for Readers

[lauramurgatroyd]

It would be good for all readers in CIL to have the ability to slice and bin as they read (if possible, depending on format).

We could set it up so that a user could do the following steps to achieve this:

reader = <ReaderName>(file_name=...)
reader.set_slice(…)
reader.set_binning(…) 
reader.read()

This links to issue: #871

We could create a base class for all readers: NEXUSDataReader, NikonDataReader, TIFF, TXRMDataReader, and the new NXTomoReader which is being developed.

The base class could have the following methods:

class BaseReader():
    def __init__(self, **kwargs):
        ...

    def set_up(self, file_name=None, ..., **kwargs):
        ...

    def set_slice(self, roi):
        self.slice_roi = roi

    def set_binning(self, roi):
        self.binning_roi = roi

    def read(self):
        # reads the data, applying the slicing and binning as it reads from the file
        # returns AcquisitionData or ImageData

Notes on the current readers:

• NexusDataReader - no option to bin or slice
• NikonDataReader - option to bin or slice in the constructor (kwargs: roi and mode)
  

  CIL/Wrappers/Python/cil/io/NikonDataReader.py

  Lines 90 to 95 in 2235c65

  	def set_up(self,
  	file_name = None,
  	roi = {'angle': -1, 'horizontal': -1, 'vertical': -1},
  	normalise = True,
  	mode = 'bin',
  	fliplr = False,

• TIFFStackReader - option to bin or slice in the constructor (controlled by same kwargs as NikonDataReader)
• TXRMDataReader - no option to bin or slice
• NXTomoReader - currently being developed - would like it to use this proposed base class.

It seems the option in NikonDataReader and TIFFStackReader is to bin or slice, is there any reason not to allow both (allow you to slice, and then bin the data)?

[paskino]

The TXRM reader uses dxchange, which in turn uses olefile. At any rate one could read a TXRM file slice by slice with slice_range:

https://github.com/data-exchange/dxchange/blob/master/dxchange/reader.py#L187-L236

I wonder whether we should contribute a function to dxchange to just read the metadata which are currently returned together with data.
An option could be to read just 1 slice, which wouldn't cost too much.