Information flow labels

[shreyashankar]

We want to track information flow throughout a pipeline at a user-defined label granularity. The user will specify labels on some inputs (i.e., the most upstream input), and we must store and propagate them throughout the pipeline.

Steps:

• [x] Create 2 tables: (1) mapping label key to IOPointer, (2) mapping deleted label to time of deletion request Add ability to index by IOPointer
  • [x] Perform db migration
  • [x] Modify existing commit functions to also commit a label or dummy label
• [x] Implement retrieval functions for queries
  • [x] Get all IOPointers for a label
  • [x] Get all labels for an IOPointer
  • [x] Forward trace (get all artifacts derived from a label)
• [x] Implement delete_label function
• [x] Modify component run function to:
  • [x] compute labels for outputs based on earlier retrieval functions
  • [x] check if label is fully deleted and block computation if necessary
• [ ] Write dummy pipelines to test functionality