single-cell workbench - get_dataset_info.cgi needs to be faster

[adkinsrs]

When selecting a dataset in the single-cell workbench, the get_dataset_info.cgi script can take several seconds. This includes nearly 15 seconds on one particular dataset I spot-checked. Currently there is no "loading" indicator to show things are happening and so people may think things have stalled.

[adkinsrs]

For comparison, get_dataset_list.cgi which returns multiple datasets (for the expression page) works in less than a second. Both servers create Dataset objects within

[adkinsrs]

The get_dataset_info.cgi also returns the dataset shape (genes x cells) which requires reading the file into an AnnData object and determining the shape. That's the bottleneck.

[adkinsrs]

Going to close this for now. We cannot really speed up the reading of the h5ad file per se, but I moved some code around so that a) the analyses are retrieved independently of the dataset information and b) the loading indicator shows while the dataset is retrieved instead of only for drawing the preliminary plots.

[adkinsrs]

Given my testing in #889 I think it may be wise to reopen this and change the shape retrieval in geardb to use the shadows package