On Mon, 2010-08-30 at 13:15 -0300, bulia byak wrote:
> On Mon, Aug 30, 2010 at 8:38 AM, Thorsten Wilms <t_w_@...> wrote:
> > Hi!
> > I had another look at the dialog.
> > Inkscape defaults to shifting the clones by 100% width of the selected
> > object for columns and 100% height for rows. So the parameters on the
> > Shift page are actually about the deviation from that, but the interface
> > doesn’t make that clear at all.
> Well, it only makes sense to say that base shift is 100% for P1. If
> the shift is accompanied by various rotations and flips, it's not
> clear how to count percentages and whether they make sense at all. The
> best formulation is: these values move clones from the position they
> would have after the symmetry transform, in the units of the tile's
> width/height. This is why I'm against displaying 105% instead of 5%
> there, because that 105 would be a sum of unsummable values: the 100%
> move is different from the 5% move (except when we're using P1).
Let's see if I get things right:
The P1 scheme creates a clone and translates it and will treat the clone
as the new original for the possible next operation (all links refer to
the true original, though, I guess).
The operation will be repeated until the number of clones +1 reaches the
specified number of steps for one direction. Afterwards the entire
result is used as selected object for the other direction.
But all other wallpaper groups have a cell size > 1. To get to see one
complete P4G pattern, you need to specify 2 rows, 8 columns, even though
the resulting arrangement is 4 x 4. I wasn't able to deduce those
numbers, I had to try.
This makes me think that for all but P1, using the original object as
unit for rows and columns is the wrong approach. What is really of
interest are repetitions of the full pattern (though depending on the
wallpaper group, 1/2, 1/3, 1/4, 1/6 or 1/12 could be the resolution).
Then I guess for all wallpaper groups (except P1), once one pattern is
full, additional clones will be used to create translated duplicates of
that pattern. So an implicit P1 as follow-up, no recursion.
> > "Width, height": it could be made clearer that this option will fill the
> > specified area and in what directions (original defines top left).
> Feel free to edit the tooltip or label!
Fill width and height, starting from the top left: [ ]x[ ] [px]
Fill area, starting from the top left:
Width: [ ] Height: [ ] [px]
> > Use saved size and position of the tile checkbox: what is the use case
> > for this option?
> You made a tiling, then scaled the tile, and want to repeat or
> slightly change the tiling, but without scaling it - leaving it the
> same frequency (even if with overlapping or gaps). Feel free to edit
> the tooltip to make it clearer.
Sounds to me like a user having such needs could well just copy the tile
and scale it back, making the cost of such an odd option taking space
and diverting user attention too high. I mean, it's like an undo that
doesn't happen, except for the result of the next operation, so you work
with one state, while you see another.
> > The Exponent parameter on the Shift and Scale pages depends on the
> > tool-tip to explain that it defines whether rows will be spaced evenly
> > (1), converge (1).
> You can suggest a way to make this clear without a tooltip?
* Add an info text
* Use sliders with 3 marks for converging, even and diverging
> > Trace page: Well, non of my tests produced anything sensible or useful.
> So what were you tracing?
Tried with a photo and various shapes in a new drawing without using
layers. I can get some transparency and size variations, but no color
changes (I did check "Color" in "3. Apply ...".
> > The Symmetry page only contains a pop-up list (what GTK+ erroneously
> > calls a combo box) full of mysterious items:
> The empty space below is reserved for explanatory illustrations, which
> would make this a lot less mysterious.
I think a list box with integrated illustrations could work.
> > Here’s an attempt at creating the most simple schematics, leaving out
> > points of rotation and mirror axes to just depict orientation. The place
> > taken by the selected object is darkened:
> > http://thorwil.files.wordpress.com/2010/08/wallpaper_groups.png?w=176&h=1023
> Nice start, but I think they can be made even more clear. Why the
> skew? How will this fill a larger space, not just the first bunch of
You mean the first 2, left side? just to mark the ones that work with
parallelograms. That should be left out anyway, as long as there is no
parameter to assume a skewed bounding box.
I think showing more than 1 cell/repetition only makes it harder to
understand the patterns. Maybe faded out to the right and bottom, but
even that takes precious space.
Oh, and should there be an option to rotate the wallpaper patterns by
90° in either direction, original tile staying top left? Especially
looking at PM, PG, CMM.
> > These could be added to the descriptions given now.
> Definitely yes.
I'm very much willing to bring the SVG into shape so it can be dropped
in, just tell me how you want it.
thorwil's design for free software: