This seems to occur when the user initiates the search while it's still generating autocomplete suggestions.

Right now it doesn't really look good on portrait screens. This can be fixed by using a vertical layout for these screens, where the metadata cards are below the image instead of beside it.

This should replicate all the functionality that we currently have for images.

Zooming in and out with the mouse wheel would be a nice feature to have. See Google Photos for a good implementation of this feature.

This should show the currently logged in user and provide options to log out or change their profile.

I was able to close it while it was still applying the change by clicking on the scrim. This should be a really easy fix.

When an upload fails, I'm thinking that it should just show an "X" icon next to that file in the uploader. Ideally, the user could hover over the icon and see an option to retry the upload.

This is likely because the counter in the Redux state is being updated earlier than it should be.

This doesn't actually have to be complete video support. All that needs to be present is the following:

• Video thumbnails are displayed on the homepage, much like image thumbnails.
• Clicking on a video thumbnail takes you to a details page where videos can be played in the browser or downloaded.

The file uploader should be enhanced with full support for video files.

I think this is because theclearImageView action doesn't actually do anything, since there are initially no uploaded items. The InfiniteScrollingElement update cycle is triggered by a size change in the element, and since nothing gets cleared, the size doesn't change, and this update never triggers. Consequently, the user has to manually reload to see newly-uploaded data after the first upload.

When downloading multiple files, it will produce weird results if some of the images have the same name. The MALLARD frontend should automatically detect this condition and disambiguate the names.

The upload workflow should have an overall progress bar that reflects the percentage of the total number of images that are uploaded. This should be very easy to implement, although some thought needs to go into the UI.

I've seen this off and on forever, but until now, I didn't have a reliable way to reproduce it. Here's how to do that:

• Click on a video thumbnail.
• Press the play button to play the video.
• Pause the video
• Press the back button

The back button will disappear, but it will not actually go back to the home page.

Ideally, this text should be preserved.

This is basically the video version of the "size always gets set the same" bug for images, and can probably be fixed the same way.

This is because it doesn't actually check if an auth token is valid before trying to use it.

It's a minor enhancement, but it annoys me every time I press the arrow keys and nothing happens.

I have some directories that contain large TIFF orthophotos. If you try to upload these to MALLARD, it will just hang. I don't think it's even getting to the preprocessing stage. No errors appear in the console, so I suspect this is failing somewhere in an action creator and the error is getting eaten by Redux.

These types of images don't necessarily need to be supported, but it should at least handle them gracefully, possibly by ignoring them as it does with other non-image files.

This should basically replicate all of the functionality we currently have for images.

For some media, especially wide ones like videos, part of the notes card can be cut off, at least on certain screen sizes. This issue might plausibly be merged with #12, since fixing that should fix this too.

This was observed on Firefox running on Windows 11, on a machine connected to the VPN where I hadn't tried loading the site before. I see the following error in the console:

Cross-Origin Request Blocked: The Same Origin Policy disallows reading the remote resource at https://bsailn2.ad.ufl.edu:8085/.well-known/openid-configuration. (Reason: CORS request did not succeed). Status code: (null).

This probably just means that Fief needs to be added to the list of allowed origins.

It would be useful to be able to select one image, and then hold down shift and click another to select everything in between.

This does not work the way it should. Instead, it dispatches a whole bunch of selectImages actions, each for only a single item. I think something weird is going on here with the selection logic in artifact-thumbnail.

When you click on a thumbnail, enter the details view, and then go back to the main view, it will scroll back to the top of the page. It would be nice to preserve and restore the users' scroll position instead.

This does not need to be fancy. We just need a stop-gap authentication method so that our data isn't completely exposed on the UF network.

Vidstack seems like a good recent project. After fighting for hours with Video.js and still not getting it to work, I appreciate that it supports web components natively. I think that this could significantly improve the experience of watching videos within MALLARD,

It looks like this is caused by a 422 error, so it's probably not correctly separating images and videos for two calls to batch_update on the frontend.

Again, the title says it all. This should theoretically be an easy fix.

Currently, thumbnail generation is bottlenecking uploads with high-resolution images.

I think Pica supports some options to do nearest-neighbor downsampling and stuff like that, but I haven't really investigated it.

This should be relatively easy to implement. It's mostly a matter of reusing the metadata entry form that it currently uses for uploads.

This goes beyond showing a mere video thumbnail on the homepage. The thumbnail will be styled in such a way as to make clear that it's a video (similar to Google Photos). Additionally:

• Hovering over the thumbnail starts playing the video.
• Moving the mouse horizontally while hovering over the thumbnail scrubs through the video.

Things that are deleted shouldn't be permanently gone.

This is less intuitive, but allowing users to request unions of arbitrary queries has some non-negligible performance implications. Furthermore, the only place where this capability is actually used right now is for searching, and this would work just as well (and be much more efficient) if we made this change.

Pretty self-explanatory. It just shows a blank screen initially, which is kind of confusing.

(API Reference)[https://developer.mozilla.org/en-US/docs/Web/API/Intersection_Observer_API]

I suspect this will be simpler and more performant than the existing scroll-event-based implementation.

With really big uploads, batch_upload can take a non-trivial amount of time. If you close the modal before it's complete, the user will see incorrect data. If the user then reloads the page, batch_upload will be cancelled and that incorrect data will become permanent.

So... we should probably wait for batch_upload to complete before closing the modal. This can be implemented similar to the bulk deletion action.

This project is long overdue for some sort of CI, possibly using Github actions. While we're at it, we can build Docker images and push them to Docker Hub.

You should be able to go to the next or previous item without returning to the home screen.

If you do this, it ends up in a weird state where it says the items are selected, but none of them show as selected.

This is especially important as the amount of data stored in MALLARD continues to grow.

• MySQL backups are pretty easy. We can mostly copy the workflow from eLabFTW.
• MinIO backups are harder, but we might be able to cook something up using s3fs.

This makes it easier to navigate really huge sessions, which can have thousands of images and currently require a lot of scrolling.

When you mouse over a video, it should only preview the video and not the audio.

If you search for something, and then upload new data while searching, when you finish the upload and it reloads the image view, it will use the incorrect default query instead of the search query.