Roadmap

[danielballan]

xref: https://forum.image.sc/t/napari-plugin-and-tutorial-for-remote-data-access/70257/2

Broadly, we want to replicate https://tiled-demo.blueskyproject.io/ui/browse/ but in a Napari plugin. Rough plan for order to build this in....

Step 0

From the IPython terminal embedded in Napari, load data into the viewer. Untested suggestion:

from tiled.client import from_uri
c = from_uri("https://tiled-demo.blueskyproject.io/api")
ac = c["generated"]["small_image"]  # tiled ArrayClient
viewer.add_image(ac)

MVP

• Text entry for URL https://tiled-demo.blueskyproject.io.api and a button to "connect". This should stash a client c as state in the plugin.
• List contents of root node as a table. To start, just show their keys. For example, for BlueskyRuns, these will be uids. c.keys()[:]
• Support pagination. c.keys()[0:5], c.keys()[5:10], ...
• Clicking a row should navigate into that sub-"directory". c[key]
• Breadcrumbs should indicate the current location and enable navigation back up the tree.
• If a given item is an array (c["a"]["b"]["c"].item["attributes"]["structure_family"] == "array") show a button the corresponding row.
• The button adds a Layer to the view.
• Customize the columns in the table. Perhaps use the same configuration that the React UI uses, where for a given Tiled "spec" we provide a list of columns:

  specs:
  - spec: CatalogOfBlueskyRuns
    columns:
      - header: Bluesky Plan
        select_metadata: start.plan_name
        field: plan_name
      - header: Scan ID
        select_metadata: start.scan_id
        field: scan_id
      - header: Time
        select_metadata: start.time
        field: start_time
    default_columns:
      - plan_name
      - scan_id

• Solve the design question of how to conveniently get to "The Image" when presented with a composite object like BlueskyRun, without making the user walk the tree to dig the image node out.

Later

• Search queries (prerequisite: needs work on the server to tell more about what the available options are)
• Access tiled "profiles" as a convenience / alternative to entering a URL to connect. (Imagine a drop-down box listing the profiles.) Maybe make it possible to create and manage profiles within the plugin.
• Draw on @michael-sutherland's work on mosaic-like layouts of images.
• Napari emphasizes N-dimensional image viewer, but sometimes it's also useful to see a line plot or some other kind of data visualization next to the image. Some Napari plugins can integrate matplotlib figures and other things. We might use them to display non-array data (e.g. dataframes).