I've found rendering the outline as a stroked line is not always enough. See for example this elephant shape: http://i.imgur.com/vEv9k5H.png Although the outline is right, it does not produce the desired result as a significant part of the original shape (pink) has been clipped.
Miters can partly alleviate the problem but due to being linear they offer poor approximation: http://i.imgur.com/8xSgdnr.png
Having found this issue now I'm using fill instead of stroke. http://i.imgur.com/7ntxkqV.png
Regards, Juan Luis.
On 25/08/15 21:07, Juan Luis Boya García wrote:
Hi.
I'm Juan Luis Boya. As a Semester of Code project (http://semesterofcode.com/) I've have been for some time trying to add alternative stroke alignment options to Inkscape.
There is an in-progress W3C draft for SVG2 that attempts to introduce a property named `stroke-alignment` for this purpose. This is an attempt to create an implementation of it. http://www.w3.org/TR/svg-strokes/
This property would control the placement of path strokes. This is better explained with an example. See this image: http://i.imgur.com/GIqKrwj.png
The image shows the same rectangle with four different stroking modes. The black line represents the shape path, as that you would see with the 'Edit paths by nodes' tool in Inkscape. Shown in orange is the stroke of the path as would be seen by the user.
A issue with the previous proposal was there was no definition for open paths. I propose to amend this in two ways:
a) Adding two new values: `left` and `right`. `left` would place the stroke at the left of each path segment. The direction of the path is determined by the order of the nodes, in the same manner arrow lines work. The following image shows an example of how this would work: http://i.imgur.com/otULloP.png
b) `inset` and `outset` would work as if the shapes were closed with a straight line, except that line would not be rendered.
The property would be set as an style property (in CSS code). Add a few toggle buttons would be added to the Inkscape UI in the Stroke Style tab in order to control the value of the property.
As for the implementation:
- The offsetting would be done only to the stroke line, which will
still be rendered as an stroked line (instead of being converted to a polygon). 2. The property would be added to NRStyle and SPStyle classes. 3. Add buttons to the Stroke Style UI. 4. As far as I know, paths are rendered in DrawingShape::_renderItem(). This method reads style properties from a class member named _nrstyle. (Therefore we need to add properties to both NRStyle and SPStyle). Now, this method would read the property and stroke appropriately. At the time of writing Cairo does not have support for a style property like this, so we will do the offsetting in Inkscape.
a) If the value is center, the same path would be used for both stroke and fill. b) In other case, first fill will be applied using the original path, then the offsetted path would be calculated and stroked.
For the stroke-alignment values 'right' and 'left' applying the offset is as simple as offsetting by stroke-width/2, positive or negative respectively.
'inset' and 'outset' are a bit more problematic, especially for self-intersecting shapes. This figure shows how a self-intersecting pentagram would approximately look with different values of stroke-alignment if 'inset' and 'outset' are considered valid for self-intersecting paths. http://i.imgur.com/s8sLumk.png
In this document I explain a procedure to stroke them, but there are some required algorithms I don't know, remarked in "Issue" boxes: https://inkscape.org/en/gallery/item/6356/stroke-alignment.pdf
As for the state of the implementation:
- After quite a bit of time I've got to understand the classes that
make Inkscape render shapes. I have been able to add the required SVG property to Inkscape. I'll commit it soon. The controls in the UI will be hidden under a compile time flag.
I've written code to outline the stroke, but also...
I've found several important bugs in half_outline() that cause it to
fail even for simple shapes. For instance, this is the result of passing a triangle to half_outline() when it has been drawn counterclockwise: http://i.imgur.com/3Tq9SvZ.jpg For that specific bug I have a provisional fix, but it should be examined more closely.
Regards, Juan Luis.