cmudig / emblaze Goto Github PK

The following thumbnail specification causes an error when loading the widget:

img_thumbnails = emblaze.ImageThumbnails(images, ids=image_ids)
text_thumbnails = emblaze.TextThumbnails(names, ids=text_ids)
thumbnails = emblaze.CombinedThumbnails([img_thumbnails, text_thumbnails])

The error seems to occur because text_thumbnails doesn't have a descriptions object when it is serialized to JSON. Current workaround is to add a descriptions parameter containing None for every id: emblaze.TextThumbnails(names, [None for _ in text_ids], ids=text_ids).

it looks like you do some clever stuff under the hood to resize images and partition them into spritesheets. my images are currently on disk in PNG format (512x512): since this tool is going to do it's own i/o and preprocessing anyway, it would be nice if I could just pass it filenames for it to work with however rather than loading all the images into memory before I can pass the thumbnails into the widget.

I tried installing your library on python 3.11, but could only install version 0.9, as your latest 0.10 version requires the redis_server package, which only contains wheels for cp35, cp36, cp37 and cp38. It seems this package is not being actively maintained anymore.

It seems this redis_server package is optional and simply allows one to automatically start a redis instance. Could you make this package optional ?
This would allow users to install the package without redis (eg. if you only need the notebook part) by running pip install emblaze.
If you need the standalone part, you could then run pip install emblaze[standalone] (or any other name you decide to use).
You can add other dependencies for the standalone package in here as well.

Seeing as this package does not seem to be actively maintained anymore, you might need to find a more suitable solution for the long term, but this would already solve the problem for most users, as I imagine many people use this package in a jupyter notebook environment.