Jeff Laing

Since the user has no idea whatsoever what the "strength" of the selection is, it is pointless to offer precision in selection process. Displaying exact pixel numbers in the user interface is stupidly misleading, particularly when you allow them to be explicitly typed in. This is all somewhat moot, for this task I have switched to Acorn which has no problem respecting requests for precise selection sizes. When I ask it to crop to a specific selection size, that is the exact size that I get, not something nearby that it thinks is indicated by the transparency of the pixels. Anyone creating images for use as sprites will understand why transparent edges around the selection and exact sizes are critical.

What resize? I didn't ask for any resize. I swept out a visible set of pixels, then used the transform panel to explicitly snap the size of that selection as per the previous suggestion in the thread. At no point did I attempt to resize any content? If resizing the content was somehow implied by the transform panel then it does not meet the original requirement. I could see where someone might argue that the New From Clipboard resampled because it didn't propagate on the 600DPI - which I believe happens, dropping the new file to 96DPI - but downsizing shouldn't make things fuzzier and it shouldn't infer data that wasn't inside the selection set. Frankly, I think the New From Clipboard should have maintained the source data resolution, not throw away information. In this case, "New From Clipboard" has apparently done "New something that's a lot like but not the same as the clipboard" Again, I can understand the pixel-level issues involved, but those are not communicated out through the user interface which currently implies exact values being input will be turned into exact output.

Note, I'm really not interested in response of the form "its by design, we grow/shrink the selection to include fuzzy content" - within a document, I can see why that's appropriate, especially if you have masks/blend modes/transparency etc. involved. But I had *none* of that. Single layer, Normal Blend, No Mask, Rectangular selection. I'm talking about the specific case where I told Photo an explicit size to copy and to then create a new document using what I assumed to be the size I copied. I wanted to include those new documents into another application as part of its texture map, and randomly injecting/omitting a few extra pixels in that context screwed me up... If you are going to use the information in the Transform panel as "advice" rather than "specification", then you need to make the "effective size" of the clipboard visible somewhere.

Create a simple document by opening a 600DPI PNG with Affinity Photo Sweep a rectangular selection, then use the transform panel to precisely size the selection to 1500x2100 (those are the numbers I was using, smaller ranges show similar issues) Copy Merged, then New From Clipboard and look at the dimensions of the resulting image. For me, it is nearly always off by 2 or more pixels in each dimension. Over on I described the problem initially, but have been redirected here. Uploading the actual files is prohibitively expensive (>100MB) but the problem seems relatively easy to reproduce. After some back and forward, Walt managed to create the same symptoms using Photo on Windows (I believe). (As an aside, the "newbie post limit" is a bit frustrating - I can appreciate that it relates to keeping spamsters and trolls out but when a persons first contact with the forums is to look for support on an issue, it is quite frustrating to discover that you can't provide a response to a question you are asked. Even switching to personal messaging only allowed a single reply - you should make very visible in the interface "as a newbie, you have (n) posts left today" - I'm going to assume it isn't there because I can't see it on this page)

I don't think Walt is saying its not a problem, just that he didn't immediately see it when I described the symptoms and it took a while to work out what was different about us; I messaged him privately (because my "newly post limit" ran out quite quickly) I'm not surprised that people haven't noticed it before, I did some experimenting with lower-resolution PNGs and at times it was hard to reproduce - it was only in the images that I was actually working with for real that it manifested easily, and it could be worked around with a little care. I think that 600DPI might aggravate it, or perhaps using a single pure pixel layer but still using "Copy Merged/Flattened" - I'm going to leave it to the experts who can look inside the software rather than try to guess. I'm not sure if posting here counts as an official bug report or there is some other place that you go to report these sorts of issues.

Maybe this is a Windows / MacOS thing?

I see it now - being in the Document menu, I took that as an Edit, not as a View Option. Something like "Make Background Transparent" - I expected the menu options that control visibility to be over in View and labelled "Show <something" Anyway, I'll keep that selected from now on

Wow, I wouldn't ever have gotten there by myself. Thanks, I'll give this a try when I get a chance.

How do I do that? I can't see any option to mark the background as transparent? I appreciate that it was my lack of experience with Affinity that had me messing with "no layer at all" - mildly surprised that "New Document" didn't give me something but that's understandable. However, as soon as I added a Pixel Layer, I expected it to be fully populated until I added a mask or something to indicate that it was in fact sparse.

Start from this Copy Merged / New from Clipboard results in this

While I appreciate all the technical detail, that really is just an excuse. I explicitly entered the size I wanted selected - not a general idea of what I would like it to try for. Remember that it was Affinity that suggested that this was a method that could be used to get a specifically sized selection back in Oct 2018. (My example was actually made from a new document which I assumed was filled with white, since I had done nothing to it other than "new document, make a selection, try to copy, damn error, add a layer, try to copy, damn errors, paint tool to slap some colour over the selection but it constrained the painting to the selected area, not to a few fractional pixels outside the area like copy apparently does. I am mildly annoyed that it thinks I wanted the selection snapped to content, rather than the area I explicitly chose - perhaps I'd like the blank background copied as well - or is that a copy merged vs copy problem?) In the "expanding" case, I was *NEVER* selecting from an area that had no pixels. It was always a subset from the middle of the image. I was probably zoomed out because I needed to see the entire image segment I was copying, and because you can't easily/accurately grow the selection larger than screen size if you are zoomed in. So, what is the solution if I have a hi-res page of (say) 20 playing cards and I want to accurately select each card using a rectangle (that I need explicitly sized to ensure they are all the same), then drag around? It sounds like you are saying I can never reliably have both size and position because Affinity will fuzz the selection if I start zoomed out.

There ya go. This time the selection shrinks by 5 pixels when you copy merged/new from demo.afphoto

Tried it, didn't seem to make a difference. I tried to upload a sample but it >150MB and it was taking forever. I suspect that it might be because the document is a 600DPI PNG (that I am copying from) - I can imagine where Copy Merged is perhaps downscaling the pixels for some reason and going fractional, in which case "please don't do that either"

According to the data entry form, I was definitely on integer coordinates. Perhaps the user interface is hiding fractional pixel positions for me. It was so predictable that I just used 1498 x 2098 as the size and regardless of where I dragged it from, I got 1500x2100.

Late to the party here, but what I've noticed (in the version I'm running) is that I set the transform size explicitly to 1500x2100, copy, new from clipboard and the resulting image is 1501x2102. That isn't too helpful...