Regression on image caching

[Galadirith]

@leipert I like your changes moving as much logic as possible to the pathwatcher listener in 7c0949f but there is a regression on deleted images (which I should have added a spec for, my fault). Steps to reproduce:

1. Preview an image
2. Modify the image bumping query to ?v=2
3. Delete the image setting query to ?v=deleted
4. Restore image however it sets the query to ?v=1

Part of the problem is that iv > 0 evaluates as false if iv is deleted or renamed. But simply conditioning on that is likely to cause similar regressions after further delete restore cycles.

A possible design to fix would be to push the version to imgVersion[src][1] if the event is not change and restore that version when the image is restored, although I don't know which of the three events is triggered when an image is restored, I would guess change by the steps experienced above.

Thanks @leipert :D

[leipert]

Actually that is really what I wanted 💃
What about using unix timestamp instead of an integer

[Galadirith]

Oh of course my bad, there is nothing else that condition could be checking for, ok :D.

So the only problem then is that resetting the query to ?v=1 is that then uses the ?v=1 cache from before the image was deleted.

I like the unix timestamp idea a lot, solves the problem without adding any extra logic to resolveImagePaths(), great :D

[Galadirith]

Surprisingly the specs didn't pass for me @ 7c0949f. The issue seemed to be that on Windows when fs.renameSync() is invoked a rename event is fired followed immediately by a change event; I don't know weather its a different order on OSX or simply that only rename fires or some other reason. Again my fault for not having a spec to test renaming in the first place.

I think this can be fixed easily enough by decoupling the debounce from the listener and applying just to triggering a rerender.

[Galadirith]

@leipert let me know if you wanted me to handle these changes and if you wanted me to integrate the minor changes sitting on a486279 :D

[leipert]

#59 is merged, just needs #60 to be working really great.

If you want, do those changes!

[Galadirith]

Great, will handle that shortly :D