Possible bug in dashedStroke

[jeffreyrosenbluth]

In the following image the straight line coming out of the top does not belong there. It seems like a numerical instability since it does not occur if I change the image size.

http://i.imgur.com/3OniKGh.png

If you want I can give you the diagrams code that generated it, but I'm not sure how to isolate the problem

[Twinside]

Yes please, if you could dump the Rasterific geometry for the whole circle, it would be even better.

Is the "big" dashing on the middle right of the circle normal?

[jeffreyrosenbluth]

I'm not sure how to dump the geometry of the whole circle. But if you have a way let me know and i'll do it. The "big" dashing on the middle right of the circle is a diagrams problem - not rasterifice. Here is a link to the code:

https://github.com/diagrams/diagrams-share/blob/master/Diagrams/gears.hs

The offending fucntion is galaxy.

[Twinside]

I think I'm going to add a dumping function around Drawing, it should help reproduce bad cases like that

[jeffreyrosenbluth]

That would be great. Are you able to run the diagrams program? The latest version of the rasterific backend needs to be built with the HEAD branches of diagrams-core and diagrams-lib. I made a small change diagrams-lib that is only in HEAD, but used in diagrams-rasteriric.

On Thu, Mar 20, 2014 at 9:17 AM, Vincent notifications@github.com wrote:

I think I'm going to add a dumping function around Drawing, it should help reproduce bad cases like that

Reply to this email directly or view it on GitHubhttps://github.com/Twinside/Rasterific/issues/4#issuecomment-38164931 .

[jeffreyrosenbluth]

Sorry to bombard you, but this code also produced a funny result. The circle is deformed.

{-# LANGUAGE NoMonomorphismRestriction #-} import Diagrams.Prelude import Diagrams.Backend.Rasterific.CmdLine

example :: Diagram B R2 example = hcat [ unitSquare # fc blue , unitCircle # fc green , unitSquare # fc teal ]

main :: IO () main = defaultMain example

On Thu, Mar 20, 2014 at 9:19 AM, Jeffrey Rosenbluth < jeffrey.rosenbluth@gmail.com> wrote:

That would be great. Are you able to run the diagrams program? The latest version of the rasterific backend needs to be built with the HEAD branches of diagrams-core and diagrams-lib. I made a small change diagrams-lib that is only in HEAD, but used in diagrams-rasteriric.

On Thu, Mar 20, 2014 at 9:17 AM, Vincent notifications@github.com wrote:

I think I'm going to add a dumping function around Drawing, it should help reproduce bad cases like that

Reply to this email directly or view it on GitHubhttps://github.com/Twinside/Rasterific/issues/4#issuecomment-38164931 .

[Twinside]

I don't have diagrams hackage nor HEAD installed yet, and I'm at work so I cannot check right now :]

[jeffreyrosenbluth]

No worries, maybe we can talk on irc later

[Twinside]

Hmm I don't see the problem with the circle, but I'm approximating with bezier curve, would it be a rendering error.

[jeffreyrosenbluth]

take a look at:

http://i.imgur.com/zBtXgVO.png

this is what i'm getting

On Thu, Mar 20, 2014 at 5:59 PM, Vincent notifications@github.com wrote:

Hmm I don't see the problem with the circle, but I'm approximating with bezier curve, would it be a rendering error.

Reply to this email directly or view it on GitHubhttps://github.com/Twinside/Rasterific/issues/4#issuecomment-38226193 .

[Twinside]

Hmm the image border are not the same, I'll have to try when reducing the image size, there may be a problem with clipping

[Twinside]

Could you retry the circle? I can't reproduce the problem.

[jeffreyrosenbluth]

I don't get the funny line anymore but some of the dashes are too short or missing.

http://i.imgur.com/jmEuHIh.png

On Fri, Mar 21, 2014 at 3:37 PM, Vincent notifications@github.com wrote:

Could you retry the circle? I can't reproduce the problem.

Reply to this email directly or view it on GitHubhttps://github.com/Twinside/Rasterific/issues/4#issuecomment-38314697 .

[Twinside]

Hmm I need to tweak some constants

[Twinside]

It should be OK with the new constants, is there still a problem with http://i.imgur.com/zBtXgVO.png ? (I cannot reproduce it :-/)

[jeffreyrosenbluth]

The dashing works now with the new constraints. But I still get get a deformed circle like http://i.imgur.com/zBtXgVO.png and I'm running a fresh pull of the code on a different machine - still OSX 10.9.1 though.

On Fri, Mar 21, 2014 at 4:48 PM, Vincent notifications@github.com wrote:

It should be OK with the new constants, is there still a problem with http://i.imgur.com/zBtXgVO.png ? (I cannot reproduce it :-/)

Reply to this email directly or view it on GitHubhttps://github.com/Twinside/Rasterific/issues/4#issuecomment-38322471 .

[jeffreyrosenbluth]

It's very strange when I add 1 percent padding around the diagram the problme goes away. It looks like a diagrams bug, but it doesn't occur in other backends.

main = defaultMain $ pad 1.01 example

On Fri, Mar 21, 2014 at 5:09 PM, Jeffrey Rosenbluth < jeffrey.rosenbluth@gmail.com> wrote:

The dashing works now with the new constraints. But I still get get a deformed circle like http://i.imgur.com/zBtXgVO.png and I'm running a fresh pull of the code on a different machine - still OSX 10.9.1 though.

On Fri, Mar 21, 2014 at 4:48 PM, Vincent notifications@github.com wrote:

It should be OK with the new constants, is there still a problem with http://i.imgur.com/zBtXgVO.png ? (I cannot reproduce it :-/)

Reply to this email directly or view it on GitHubhttps://github.com/Twinside/Rasterific/issues/4#issuecomment-38322471 .

[Twinside]

Are you on GHC 32 or 64 bits? If you're on 64 bits I got an idea (and the implications would not be pleasant for me)

[jeffreyrosenbluth]

Sorry to say but I'm on 64 bits - hope that isn't really bad news

On Fri, Mar 21, 2014 at 5:52 PM, Vincent notifications@github.com wrote:

Are you on GHC 32 or 64 bits? If you're on 64 bits I got an idea (and the implications would not be pleasant for me)

Reply to this email directly or view it on GitHubhttps://github.com/Twinside/Rasterific/issues/4#issuecomment-38328274 .