Save as Figure

[will-moore]

To test:

• Open 1 or more images, zoom, pan, change rendering settings and add ROIs.
• File > Save Viewport(s) as Figure
• Enter a name for figure
• Click on link in the success dialog -> Open figure in new tab
• Only the ROIs (Shapes) visible at the time of export (and intersect with the viewport) will be included in the Figure.
• Pixel sizes, timestamps, Dataset names etc should be included in the figure

[pwalczysko]

Bug: The panel in the newly created figure is shifted with respect to the ROIs.

1. Create a "figure from viewport" with couple of precisely drawn freehand polygon ROIs (e.g. "segmentation" of nuclei in a large image) - you can either draw these ROIs manually at the time of opening the image in iviewer, or have them already drawn and saved, or, draw them and save these ROIs. Click the "figure from viewport" menu item, select the name for the new figure and navigate to this figure using the link in the dialog.
2. In the newly created "figure from viewport" - when inspected in OMERO.figure, observe that the ROIs which were in the viewport when the "figure from viewport" was created are positioned as expected.
3. Still in OMERO.figure, select the panel, go to Labels tab, click Edit in the ROI section and add the ROIs which are saved from step 1. above in OMERO (these are, or rather should be, the same ROIs which you just transported onto your figure using the new "figure from viewport" feature in iviewer
4. Observe that the ROIs which you are newly adding now are slightly shifted, always towards bottom right - see screenshot.
5. Save your figure, close it. Reopen it.
6. Repeat the addtion of the ROIs step in OMERO.figure as per Ad 3. above - observe again that the newly added ROIs are shifted one more time towards bottom right.
   

[pwalczysko]

RFE:

Some misformatting of the dialog - note the inverted commas are only on one side of the ID (there probably should not be any inverted commas)

[pwalczysko]

Two problems noted, the shift of the pane with respect to ROI position probably needs a full rethink.

[pwalczysko]

To comment further on the ROI-shifting problem #465 (comment) - this behaviour is present only when working with figures created by the feature implemented in the present PR #465 - the figures which are created within OMERO.figure itself do not have any problems with shifted ROIs, i.e. the ongoing work on ome/omero-figure#477 does not seem to be cause of this, cc @jburel

[will-moore]

@pwalczysko thanks for the testing. Re: #465 (comment) - I think that what is going on here is that OMERO.figure is trying not to add duplicate ROIs on top of one-another. If a Shape already exists when you try to add one, the Shape is automatically offset to the lower-right. In the same way that when you copy and paste a Shape it is offset to the lower-right. If you first delete the Shapes in OMERO.figure before loading and adding them, you should see that they are added in the right place.

[pwalczysko]

@pwalczysko thanks for the testing. Re: #465 (comment) - I think that what is going on here is that OMERO.figure is trying not to add duplicate ROIs on top of one-another. If a Shape already exists when you try to add one, the Shape is automatically offset to the lower-right. In the same way that when you copy and paste a Shape it is offset to the lower-right. If you first delete the Shapes in OMERO.figure before loading and adding them, you should see that they are added in the right place.

Ah, sorry, thanks for explanation. Indeed, this is as you say, the shapes are shifted only because they are already added once. This behaviour is not different from what is in Figure by default and is not introduced by this PR.

[pwalczysko]

I will wait for tomorrow's build for retest of the double-quotes efda997

[imagesc-bot]

This pull request has been mentioned on Image.sc Forum. There might be relevant details there:

https://forum.image.sc/t/omero-iviewer-multiple-images-views-exporting-to-omero-figure/94675/2

[pwalczysko]

Double-quotes are fixed.

The only remaining issue is that points are ignored.
As this is done without warning, I wonder if we should add one ?
If points are not supported in Figure, then this is fine... but still they are supported in iviewer - if someone gets into great lenght and creates a lot of points, then they might be unpleasantly surprised not to find them in figure later...

[pwalczysko]

#465 (comment)

[will-moore]

@pwalczysko Points are now converted to Ellipses in figure...

[pwalczysko]

@pwalczysko Points are now converted to Ellipses in figure...

I am really not sure about this. If somebody has already some Ellipses (which were meant to be ellipses) on an image, and adds points, then, in Figure, they get a surprise.

Maybe we can discuss in a meeting tomorrow - my preference would be to drop the points if there are any and warn the user. Offering some distortion of the ROIs by default is not the best practice imho.

[will-moore]

@jburel Just to clarify, do we need to have Points support in OMERO.figure before this PR is approved?

@pwalczysko suggested at #465 (comment) that a warning might be enough, although I think he's changing his mind?!

If we need to wait for Points in OMERO.figure, that needs the Vite PR to be merged, then a follow-up PR. Then iviewer will depend on Points support in OMERO.figure, so we'll need to only enable "Save as Figure" if we have the most recent version of Figure. This will have to be detected in some manual way (we don't want to add omero-figure as a python dependency of iviewer).
Currently we detect if Figure is installed and don't show the "Save As Figure" option if it's not found, but detecting the version will be a bit harder.

[jburel]

I don't think we are in any rush to have this PR included and release.
My preference, as mentioned earlier today, will be to:

• release Figure with Vite
• add support for point in figure and release
• include this feature and release iviewer. As part of the release, we recommend to upgrade to the version of figure with point support. (I am not expecting a detection of the version). I think we should avoid having a release of iviewer with the feature implemented knowing that the shapes are silently converted. Points are more common than ellipse so I will not ignore them as suggested above