ROI shape loading improvements and strategies

[rgozim]

OMERO currently does not support a scalable strategy for loading a large number of shapes in the forms of regions of interest (ROI).

There are instances were users are creating and using images with large numbers of ROIs, sometimes in the regions of hundreds of thousands.

The current implementation in the JSON API returns ROIs as a composite object, containing shapes, details and other meta data. This results in a 13 second load time for a single HTTP request from the eel OMERO server and subsequently is limited to returning only 500 JSON array items.

Solution

• Reduce server load by only fetching the exact data required to list all ROIs (this will still require a limit on max ROIs)

• Add ability for loading ROIs by a viewport size and an x, y position for images of a high resolution

• Add function for loading ROIs by timepoint

[joshmoore]

Brief discussion today: initial focus will be optimizing the current calls as well as loading more than the first page of data.

[joshmoore]

@chris-allan mentions that in the case of masks the current behavior is "all on by default" which causes issues. There may be some changes to defaults that would be beneficial for non-masks as well.

[jburel]

related https://trello.com/c/tCU2TbCc/18-loading-of-5800-rois-fails