hookspec for just doing something without providing a dock widget

🚀 Feature

Hi all,

I would like to request for another hookspec that takes functions which are executed when a menu entry is clicked. Examples could be “Reset Brightness & Contrast” or “Copy current layer”.

Motivation

I’d like to implement functions such as for resetting brightness contrast

def reset_brightness_contrast(image: Image):
    image.contrast_limits = (image.data.min(), image.data.max())

At the moment, the user has to confirm the action in a dock widget which appears a bit cumbersome:

If the operation could just be executed on the current layer[s], that would be cooll. You find more examples here

Pitch

We could extend napari with easy-to-use single-click functions.

Alternatives

Additional context

At some point, we may also start collecting single-click functions such as “Reset brightness & contrast” in a common repository.

[tlambert03]

yep, now that #2556 is in, we need to think about how to expose these take-a-layer-and-do-something functions externally. This is all rather tightly connected to the concept of the action manager, which is coming along nicely as well. More generally, we can describe these single-click functions as callables that happen in response to an event (button click, mouse press, etc.) with some degree of contextual awareness (a specific layer type is selected, etc…)

We definitely plan on exposing such a mechanism, but probably need the action manager and context concept to mature just a bit.

Side note: “auto-scaling” definitely should just be easier to access in the gui. It’s the “reset” button in the contrast limits controller that you get when you right click (we agree it’s not discoverable enough)

programmatically, you can use layer.reset_contrast_limits()… so don’t need a new function for that one (but if you want to add percentile to it, that would be a nice PR).

[tlambert03]

If a labels layer is selected, this could just be executed

yep, agreed… but that “if a labels layer is selected” sentence there is exactly the sort of context I was referring to above when I said “there’s a lot of contextual stuff to take into consideration”. In other words: yes, it’s very simple to say “hey, is a label’s layer selected?” and satisfy your use case above. but what if someone wants their function to be called only if two images are selected? what if they want an image and a labels? what if they only want a 3D image, not a 2D image? These are the sorts of contextual things we’re trying to codify.

I definitely suggest taking a look through the concepts introduced in #2556, since I think that’s a promising way forward here. Specifically, have a look at the “available layerlist contexts” defined here, and how those context keys are used in the “enable_when” keys of these actions here. Those layer “actions” are the equivalent of your make_labels_editable command there and the enable_when context allows us to know when it’s a valid function to offer up to the user, and when it’s not. for instance, now if you click on an RGB image, the “split RGB” action is available:

whereas if you have three grayscale images of the same shape selected, you get “merge to stack”

make sense? So, we’re close to being able to open up that sort of a “do this simple action if this simple condition is met”… but I do think it’s a touch more complicated than at first it may appear