Creating Compound Shape Changes Shape Geometry

[deeland]

I was wondering why a fairly simple task: creating a [fairly] simple geometric icon, was giving me alignment issues between compound shapes. Unfortunately, it looks like I’ve found another bug. Ive recreated the issue in a new document.

Circle over circle
Remove small from large (creates donut-esque shape)

Pen tool > shape intersects over donut

Intersect as compound vs intersect as join, moves the geometry points when using compound.

See video screen capture …

I’m off to back to Adobe Illustrator, regretfully. But CS6 keeps my compounds aligned and true. I need my accuracy.

I hope I’m doing something stupid, but the issue is easy to recreate. File is available, on demand.

[loukash]

2 hours ago, deeland said:

I hope I’m doing something stupid

Your words…

Why not simply nest the "Curve" object inside the existing Compound container to keep the boolean operations nondestructive, and adjust until you're satisfied with the result?

 

[deeland]

@loukash Hi and thanks for replying and indulging. I get what you mean, but your example deals with building a single instance of a compound. Much like my example, but I was demonstrating with a simple example.

Let’s say the compounds are multi-levels deep. It’s time to deliver the artwork, and so you release the compounds to a single shape (you know, as one does.) Does your technique promise to not move any of the geometry? My example was very simple, but with something complex, which is where I found the issue, you find shifts in geometry, where overlapping compounds no longer align.

Both Compound and Join should result in the exact same geometry.

[loukash]

5 minutes ago, deeland said:

Much like my example, but I was demonstrating with a simple example.

Frankly, I didn't really see from your screencast what the exact issue is.
You seem to be at a very high zoom level.
Is it about "microscopic" rounding errors?
There are known issues with Affinity's boolean operations, some of which can get seriously annoying.

One workaround is the non-destructive compound where you can adjust each object until it fits, even though it seems "mathematically off".
If you then "bake" it destructively, you may still have to clean up nodes manually. It depends on the actual shapes used.

Another workaround is to use an external app for any "mission critical" boolean operations, then copy and paste paths back and forth. Been there done that…

13 minutes ago, deeland said:

Does your technique promise to not move any of the geometry?

I'm not promising anything. You'll have to try case by case.

13 minutes ago, deeland said:

It’s time to deliver the artwork

If you literally have to deliver on tight schedule, use the tool that you know will do the job safely.
Also: been there done that. That's why it took me 6 years (!) until I was ready for a transition from Illustrator CS5 to Designer 1.8.x, despite being an "early adopter" buying Designer back in 2014 when it was at v1.1. It just wasn't ready yet for what I needed, not at all. Nowadays, after intensively "studying" and experimenting with almost all Affinity features for the past 6 months, I found my ways either to work around most issues/bugs, or learned to find and use different workflows that will get me to the desired result.

3 hours ago, deeland said:

File is available, on demand.

That would be a good idea.

[lepr]

4 minutes ago, deeland said:

Both Compound and Join should result in the exact same geometry.

The results should be the same in theory, but Affinity is currently using different code for non-destructive versus destructive Boolean operations. The older non-destructive code is generally more robust than the newer destructive code, but, as you realise, the destructive code has greater precision in some situations.

[deeland]

@loukash Well, I have my workaround: don’t use Designer and go back to Illustrator, for these particular tasks.

I’m reporting what I think I’ve identified as a bug. I’m guessing your solution to use a workaround suggests what you see as a known issue, yes?

If it’s a known bug for 1.9.3, then I guess it will be filed away for later action by the developers.

[deeland]

@anon2 So, a known bug, in your opinion? Or class it as a limitation? Either way, it would be nice if Designer behaved reliably, when releasing artwork as a vector shape, independent of the end users software. I can hardly release live compounds, Affinity Designer-based. Would be great, if we didn’t live in an Adobe world.

[loukash]

11 minutes ago, deeland said:

Either way, it would be nice if Designer behaved reliably

100% agreed!

15 minutes ago, deeland said:

I’m reporting what I think I’ve identified as a bug.

A Serif staff member will eventually likely chime in, confirm and add an issue tag to this thread to log it with the others.

18 minutes ago, anon2 said:

Affinity is currently using different code for non-destructive versus destructive Boolean operations.

Ah, I didn't know that.

[lepr]

30 minutes ago, deeland said:

@anon2 So, a known bug, in your opinion? Or class it as a limitation?

I expect the imprecision of the code employed for Compounds is a limitation rather than a bug.

Serif proclaimed improved precision in Boolean operations when 1.8 was released. However, along with that came bugs that made Boolean operations wildly unreliable and basically unusable for many people. Several minor releases gradually removed some, but certainly not all, of these bugs. That is what we have now for the destructive operations. We still have the older, sometimes less precise, but significantly more reliable code being used for Compounds.

[Gabe]

Hi all,

Sorry for the delayed reply. 

@deeland,

Can you attach the file in question? 

[deeland]

3 hours ago, Gabe said:

Can you attach the file in question? 

@Gabe Here you go, thanks. This issue may explain the horrible "punked" results with the Profile Tool, previously reported by others. Could be a connection? Save that for another day, one step at a time, etc. 

Creating Compound Shape Changes Shape Geometry.afdesign

[Gabe]

That's a 0.000239 px (or 0.00002 mm) move. Far from anything that would be ever get noticed. 

[loukash]

44 minutes ago, Gabe said:

Far from anything that would be ever get noticed.

And we likely only know because we can zoom in that far in the first place… [*while watching Illustrator CS5 zooming in to whopping… uh, 6400%!*]

That said…
@Gabe, aren't these imperfections at the root of many of those known boolean operations issues, like the "myriads of ghost nodes" bug etc.?

[lepr]

2 hours ago, Gabe said:

That's a 0.000239 px (or 0.00002 mm) move. Far from anything that would be ever get noticed. 

I don't know about getting noticed, but the bottom node's move is 200 times worse than you say. It is 0.05 px.

[Gabe]

0.05px is still reasonably accurate.