History: Phase 2

[mikekucera]

• Snapshots are taken automatically. (Can keep the snapshot button).
• Snapshots are automatically pruned.
• Investigate a more efficient way of storing snapshots, that doesn’t involve copying the whole database.
  • For layout changes the x/y attributes of nodes change, edge documents are unaffected.
  • For style changes the _style field in the network document changes, all other node/edge documents are unaffected.
• Add thumbnails to the history panel.
  • non-trivial.
  • It would be more natural to create the thumbnail on the server at the same time as creating the snapshot. However, that requires the ‘cytosnap’ extension which uses Cytoscape.js inside a headless chrome instance, which is very heavyweight.
  • It probably makes more sense to create the thumbnail on the client and send it to the server.
  • Thumbnails can be stored as CouchDB attachments on the network document.

[maxkfranz]

It would be more natural to create the thumbnail on the server at the same time as creating the snapshot. However, that requires the ‘cytosnap’ extension which uses Cytoscape.js inside a headless chrome instance, which is very heavyweight.

Creating and destroying Chrome instances is very heavyweight. Usually this means you need to keep a headless Chrome instance running all the time and you have a job queue that reuses that instance. Biofactoid does this: https://github.com/PathwayCommons/factoid

It probably makes more sense to create the thumbnail on the client and send it to the server

Another possibility is to not store images on the server at all. If we assume that the snapshot data is already stored in the client so that the user can just click to quickly restore that snapshot, then the client could dynamically generate those images from the data using Cytoscape.