whoa. almost forgot about this place. what's new?
If I had to pick one thing, I would recommend you
Howard Day's recent thread about realtime 3d-rendering 'pseudo-pixelart'. Very interesting and potentially very useful -- both short term and educationally -- to pixellers, if he chooses to release it.
For the purpose of an pixel image preview tool, scaling up is useful. Yet comparing true scale also can be useful. However, I'd say the first one is probably more useful in most cases. So either that, or make an option for both, with auto-scaling the default. Otherwise it depends what you want to do with the auto-tiling feature, which might shift the balance of importance back again. Good work anyway, looking good.
Thanks for the feedback.
I hadn't really considered having a toggleable option.. that might be good, since there are many different strategies to try out.
For example, you can try to preserve relative scale by considering the whole page of thumbs at once, picking an auto-zoom level accordingly. This would preserve their relative scale at the cost of not zooming the smaller ones as much as you can. And you can execute that particular strategy in a few different ways (absolutely prevent all distortion by using the maximum size as basis, or prevent most distortion by taking median or average size as basis) too.
Actually, that helps.. I just realized that only the current page of thumbs has guaranteed known dimensions, so probably I must limit the checking to that anyway.
The auto-tiling feature does not interact with the thumb autozooming, as 'image mode' runs off different code than thumbnail mode. Do you think it should?
I'm finding that the auto-tiling in combination with the silhouette modes (renders the image as a single color) is a nice design feature, it shows the flow of the image very clearly and can also help spot readability problems.
I also have a modification partly completed that intentionally distorts the image in different proportions -- 1:1, 1:2, 2:1, 1:3, 3:1 etc.. (to help see if the proportions you are using are actually optimal.). I've found it useful so far but it's definitely still buggy:)
EDIT: I've now implemented a compromise strategy, which can be summarized like this:
"Pick a 'global zoom' level based on the average of the base[unzoomed] dimensions of all currently visible thumbs. When drawing each thumb, calculate how much zoom it could have.
If that factor > (global_zoom + 1), then clip it to global_zoom+1."
This curtails some of the more gross excesses (really tiny 8x8 icons like the skull would previously get huge magnifications that would make them feel way too big.)
I feel like it puts each page in good context with the other members of the page, although some mental adjustment is required to the fact of thumb sizes changing..
Overall it feels like one of those graphical docks where the icons get bigger as you approach them..
Bugs shown here:
* Thumbnail outline is using the wrong size cause it's not using the adjusted metrics yet.
Features not shown here:
* Images often 'bounce' (pop out towards you) cutely when you scroll. Totally unintended, an artefact of how the display is rerendered during scroll.
A tunable option, I guess, would substitute that value 1 with 2, 4, 8, or 64 (or 0, or maybe also -1024), which would let you choose relative importance of 'correct relative scale' vs 'easily see pixels'
