Data management overhaul

[dpolhamus]

Allow ability to handle multiple datasets in the application. Right now we allow uploading of a single NONMEM run table file, a single source dataset, and VPC data.

• Extend this to manage multiple datasets from disk (any of the readr or haven supported types)
• Allow user to define merge specifications to create new datasets from any combination of these
• Update all plots to allow for specification of which dataset to use per plot

[dpolhamus]

Suggesting we have an overview panel showing all the datasets one has uploaded into the app. What info to show about it?

• number of records, number of columns
• group attribute?
• a way to access column names and character types

This can replace the data summary tabs that are in the app currently

[dpolhamus]

Improve the panel for specifying datasets to read in from disk.

• Perhaps some better mechanism is available now for point and click browsing. The performance hit we took on this last time when sitting on NMStorage was enormous, we should check again.
• Indication of whether file exists
• Allow user to pass column info via a yaml spec
  • labels
  • class
  • significant digits to use when printing
  • range, values (and their mappings)
• Allow user to give this dataset a name for reference in remainder of app

[dpolhamus]

Given how easy it is to get into reactive hell with these apps, move to a feather backend:

• After user specifies where file should be read from disk, move to a satchel like process where all future reads of that dataset will occur from a binary

[dpolhamus]

Code parser is sufficient as is, or do we need or is there anything we can do to improve it?

[dpolhamus]

Estimate 100 hrs