Explore SlideFlow interface

[Leengit]

SlideFlow's interface is different from HistomicStream's. We should decide whether we want to add functionality to HistomicsStream to support a similar interface.

My notes: The SlideFlow interface for dataset management is different from HistomicsStream, in part, because SlideFlow uses configuration files. The user supplies the name of the configuration files (or a directory with several files) instead of supplying the actual parameters. Without changing the HistomicsStream interface all that much, we could support configuration files. That is, where we are currently specifying

my_study = {"version": "version-1"}
my_slides = my_study["slides"] = {}
my_slide0 = my_slides["Slide_0"] = {}
my_slide0["filename"] = wsi_path
my_slide0["slide_name"] = "TCGA-AN-A0G0-01Z-00-DX1"
my_slide0["slide_group"] = "Group 3"
my_slide0["number_pixel_rows_for_chunk"] = 2048
my_slide0["number_pixel_columns_for_chunk"] = 2048

we could also support something like

my_study = {"version": "version-1"}
my_slides = my_study["slides"] = {}
my_slides["Slide_0"] = {"configuration_file": "my_slide0_configuration.json"}

Or we could also change HistomicsStream more drastically so that one doesn't even realize that there is a hierarchy of dictionaries. For example, with SlideFlow a top-level configuration file can include the names of lower-level configuration files. We could then read the whole configuration -- high-level, study-wide parameters and individual slides' parameters -- with something like

my_study = hs.new_study("my_study_configuration.json")