13 of 17 Wallpaper groups at symmetry tab of clones behaves erratically

[nevit]

13 of 17 Wallpaper groups at symmetry tab of clones behave erratically.

I attached 17 clones of a basic shape. 13 of 17 symmetries are not the way they expected to be.

P1, CM, PG, PM behave normally.
Others are broken.

What is expected from symmetries?
http://en.wikipedia.org/wiki/Wallpaper_group
http://www.clarku.edu/~djoyce/wallpaper/seventeen.html
http://escher.epfl.ch/escher/

A wallpaper group (or plane symmetry group or plane crystallographic group) is a mathematical classification of a two-dimensional repetitive pattern, based on the symmetries in the pattern. The pattern produced should have equal distance between tiling blocks. Some symmetries are hexagonal and I am not sure how parameters in other tabs apply to a hexagonal symmetry.

I attached tiles produced by Inkscape clone tool. Please compare them to tiles in reference sites.

Best Regards,
Nevit Dilmen

[su_v (suv-lp)]

I haven't checked each of your SVG files but those that I peeked into all have a translated layer group which indicates that your issues are due to the same underlying cause as reported in Bug #168651 “"Tiled Clones" inside transformed groups broken”.

Can you test if your wallpaper groups also fail in new default documents that have not been resized or changed in format or page orientation?

[nevit (nevilo)]

Hi, I redid the experiment with symmetry groups. In previous files I used default A4 Landscape as opening document. I did not rotate the base image but I do not remember if I did some scaling or moving of object before cloning.

In this file series default A4 portrait was used as opening document. I did not move, rotate or resize before cloning.

I did a 4 row, 4 columns cloning.

More groups are Ok now. (cm, cmm, p1, p2, p4, pg, pgg, pm, pmg, pmm)

Some still do not pack together. Some of them being the hexagonal tilings (p3x and p6x), which I am not sure if any correct implementation can be possible with row and column logic.

P4g and p4m are rectangular and do not pack as Ok groups above.

Best regards,
Nevit Dilmen

[su_v (suv-lp)]

Please read <http://tavmjong.free.fr/INKSCAPE/MANUAL/html/Tiles.html> and the following pages of the manual detailing all available 17 tiling symmetries and the other options of the 'Tiled Clones' dialog.

nevit wrote:
> Some of them being the hexagonal tilings (p3x and p6x),
> which I am not sure if any correct implementation can be
> possible with row and column logic.

The manual page explicitly mentions "The terms Rows and Columns are only really appropriate for tiling of rectangular tiles (…)."

> P4g and p4m are rectangular and do not pack as Ok groups above.

see <http://tavmjong.free.fr/INKSCAPE/MANUAL/html/Tiles-Symmetries.html>

[su_v (suv-lp)]

> In previous files I used default A4 Landscape as opening
> document. I did not rotate the base image (…)

The default A4_landscape template that ships with Inkscape 0.47 does not have a transform attribute on the top layer - did you use a custom template maybe?

> I did a 4 row, 4 columns cloning. (…) Some still do not pack
> together.

IMHO you are too strict in using the same square tile and rows&columns settings for the different tiling symmetries. AFAICT all 17 can be created with the 'Tilde Clones' dialog, although the parameters as well as the original tile have to be adjusted to the specific chosen symmetry as depicted in the manual page.

I don't think the tilings behave erratically (besides the known issue with incorrect transformations if inside transformed groups) nor that the resulting tilings are incorrect (but I'm not a mathematician ;) ) - your report seems more a RFE to improve the tiling dialog insofar as the labels for each option change when selecting different symmetries, or maybe even leading to a complete redesign of the dialog, no longer based on the row&column UI logic?

[nevit (nevilo)]

I used a shift in x of 50% and a shift in y of −25% per row as indicated in your manual link for a hexagon, with "P1" which was a nice tip.
But there are no other specific directions for p4G or rotations who rotate the sample 60 or 120 degrees in the manual.

I was using square as a shape since rectangles might behave odd and leave gaps or overlapping areas when rotated. (Such as in those with rotational symmetries.)

The attached p4g file was created with 4 row x 4 col but it produced a 2 column 8 row tile. Still 2 x 8 = 16 tiles but different rows and columns counts from settings.

Did you try Escher web sketch I sent in my previous post? It has a very simple interface and no confusion at all.
http://escher.epfl.ch/escher/

Best regards...
Nevit Dilmen

[su_v (suv-lp)]

On 13/5/10 11:05, nevit wrote:
> I used a shift in x of 50% and a shift in y of −25% per row as
> indicated in your manual link for a hexagon, with "P1" which was a
> nice tip. But there are no other specific directions for p4G or
> rotations who rotate the sample 60 or 120 degrees in the manual.

attaching samples, using different original tiles and adjusted values
for rows and columns, but no shift or rotate settings, just the tiling
symmetry as listed in the front tab of the dialog (ok, I sometimes used
alternating opacity values for better visibility, and I added a stroke
to the base shape (triangle or square) after creating the tiled clones
because I didn't want it to be included in the tile dimension)

> Did you try Escher web sketch I sent in my previous post? It has a
> very simple interface and no confusion at all.
> http://escher.epfl.ch/escher/

not yet.

[nevit (nevilo)]

Thank you, I guess I am doing something wrong.

Nevit Dilmen

[prkos (prkos)]

I can reproduce the difference between 0.47 and 9500dev for P3, P31M, P3M1, P6 and P6M symmetries.

Attached is the example of P3M1 with all the other settings default.

[prkos (prkos)]

Forgot to add that the screenshot on the left is 9500 and on the right is 0.47 (expected result) so this is a regression.

If I copy the object from dev version to 0.47 I get the same bug in 0.47. The bug appears even when I remove the transform attribute and create clones from that position.

[su_v (suv-lp)]

@prkos - can you attach both SVG files used for the screenshot, from 0.47 and r9500?

[su_v (suv-lp)]

nevermindm - I see the same when testing P3M1 with a polygon (3 corners) in r9495.

Odd - must be a recent change, the example in comment #7 have been created with 0.47+devel r9405.

[su_v (suv-lp)]

Removing the 'inkscape:transform-center-y' attribute of the star/polygon (which seems to be added automatically in 0.47+devel, unlike in 0.47) restores the correct transformations for e.g. P3M1.

(tested with 0.47 and r9495 and default preferences on OS X 10.5.8)

Regression possibly caused by the fix (r9204) for bug #462157 “Centroid for odd stars is wrong”:
<https://bugs.launchpad.net/inkscape/+bug/462157>

[su_v (suv-lp)]

> must be a recent change, the example in comment #7 have
> been created with 0.47+devel r9405.

Because I constructed the triangles for those examples with the pen tool (using grid and guides) and not the polygon tool, this type of error did not occur. AFAICT the regression only concerns shapes created with the polygon/star tool (with an odd number of corners).

Workaround: group the triangle and use the group for 'Cloned tiles'.

[su_v (suv-lp)]

> Workaround:
> 1) group the triangle and use the group for 'Cloned tiles'.
2) reset the rotation center by <Shift>-clicking it, before using 'Tiled Clones' with a (3-sided) polygon/star.

[Alexandre Prokoudine (alexandre-prokoudine)]

P1 seems to be broken too. I can't reproduce http://istarlome.deviantart.com/art/halftone-effect-in-inkscape-85648573 anymore.

[su_v (suv-lp)]

> P1 seems to be broken too.

Issue was a transformed parent layer (see bug #799797) - P1 is not broken in general.

[su_v (suv-lp)]

Closing as 'Invalid' since all wallpaper groups can be recreated based on the details described in the manual (and working around the known issues with clones inside groups/layers with transforms). None of symmetry modes behaves 'erratically' as originally reported. Please add a comment and reopen if you don't agree and think closing this report was unjustified.

The request for a better UI for tiled clones should be discussed separately IMHO - preliminary work has been done on the mailing list:
<http://thread.gmane.org/gmane.comp.graphics.inkscape.devel/34785>
<http://wiki.inkscape.org/wiki/index.php/Tiled-Clones>

Already filed related feature requests had been listed in this mail:
<http://article.gmane.org/gmane.comp.graphics.inkscape.devel/34832>

No blueprint has been filed yet, based on the specification in the wiki.