improvement suggestion : closing a bypassed transformation

[pierrelabreche]

PWP incorrectly infers that closing a bypassed transformation will change the downstream images, thus triggers the re-computation of all downstream transformations.

[jsachs]

This does not happen when I try it from within the image browser. First click bypass button which causes a recalc of downstream images. Then click close box which does not.

[pierrelabreche]

Screen capture sent by email

[jsachs]

As the warning message indicated, closing this transformation invalidated a mask or input to another transformation and thus a recalc was required.

[pierrelabreche]

I am missing something maybe.

The transformation was already bypassed. Consequently to bypassing, all downstream effects had been computed, presumably.

When closing the bypassed transformation, why would there be other effects ?

[jsachs]

If the output image or a mask of the bypassed transformation was used as a mask or an auxiliary input for some other transformation, deleting it invalidates the mask or input.

[pierrelabreche]

1000021893 (2).jpg (25.42 KiB) Viewed 384 times

In the case being examined, the Equalize transformation is bypassed. Note that Equalize has no associated mask.

Bypassing the Equalize transformation triggers a complete recalculation downstream.

From thereon, the output of Equalize is not being used downstream, and no downstream mask depends on the output of Equalize ( it is bypassed ).

Moreover, the masks in the Multipass sharpen and Precise Gaussian transformations were created by the Mask paint tool and the mask Blur tool. Both masks are image-independent.

Just before clicking on the close button for the bypassed Equalize transformation, all downstream transformations output images are in their intended final state.

Why would the downstream transformations need to be recalculated?

[jsachs]

I was able to reproduce this by adding Precise Gaussian with an initial Amount mask copied from Multipass Sharpen to the workspace you emailed me. Since the initial mask in Precise Gaussian depends on the mask in Multipass Sharpen and Multipass Sharpen depends on Equalize, PWP thinks closing Equalize can affect Precise Gaussian, even though in this case it does not since Equalize is bypassed. Consequently, the close triggers a recalc.

I think I have fixed this but need to do more testing.