It doesn't feel half bad, so, RFC ;).

It's slightly awkward on large screens because you can do a fairly large snap to pinch, but the inverse is not necessarily true because there's more screen space, so even a fully extended spread (which is awkward) won't necessarily correct for it.

(What I had in mind would need knowing the distance between fingers at start - and the one at end - which I guess is/was not available and why I couldn't do much initially when I made ImageViewer.)

We do know that (it's not exported, but we have the info). But I'm not quite getting what you meant, though.
(What I'm getting sounds like the sum of the travel distance of both fingers, which is already what distance is set to).

Or did you mean applying a zoom factor that would match the ratio between the span between the starting position of the fingers and the span between the ending position of the fingers? (because, again, that would probably very much match what we do with distance already).

Or did you mean applying a zoom factor that would match the ratio between the span between the starting position of the fingers and the span between the ending position of the fingers? (because, again, that would probably very much match what we do with distance already).

Yes. I think that's it.
If at start my fingers are separated by 3cm, and at end by 5cm (so, spread), the image should be scaled by 5/3: these 3 cm should become 5cm. the parts of the image I had my fingers on at start should move to where I had my fingers at end/lift time.

Yes. I think that's it.

Eh, I'll try. It feels like it should behave relatively closely to what we do with distance, but who knows ;o).

(But that's my naive expectation ;) i'm happy with what we have since some days ago - and it's quite possible doing it according to this naive expectation will feel more bothering.)

Yeah, tried it, it didn't work out ;p. (It lead to large jumps, which makes sense, pinched fingers make for a very small rectangle, so the ratio is bound to be high).

Made it configurable to see if it actually makes that much difference ;o).

(It would appear that: not so much on the H2O ^^).

I'll do more tests tomorrow, but, yeah, probably not worth bothering about, after all ^^.

(Was a fun experiment, though).

I'll probably do a different PR w/ just the code cleanups, and archive the code in a branch.

Tested this PR with its state as this time, and may be I messed my patching, but it feels quite strange and buggy:

• sometimes, consecutive pinches zoom in then out then in then out
• other strange zooms that feels like my gesture wasn't recognized (and also, possibly a lot more toggling of the bottom menu, like more touch/releases recognized as a tap)

That was with the default new setting checked and untouched (didn't try unchecked).

• sometimes, consecutive pinches zoom in then out then in then out

I got this with some SVG image (the first on in my SVG test suite EPUB) - so it's possible this affects only SVG, which get their scale_factor handled a bit specifically (image_wg always getting scale_factor=1).
I also couldn't reproduce it with other SVG images :/
(Just had a quick test this morning, didn't have the time to update and apply this PR on a clean/recent nightly.)

Yeah, agree that it's usually close enough to the current behavior when it works, and when it doesn't, it's actively irritating ^^.

Superseded by #9544