velara3 / svgweb Goto Github PK

See improvement in r157 from James.

The trunk currently determines when rendering is finished by tracking
render start and end using certain events. Proposed improvements in r150
and r151 from James change the way the start of rendering is tracked.

See proposed improvement in r157 from James.

Simplify implementation by folding SVGRoot into SVGSVGNode. Proposed
improvements in r135, r136, and r137 by James.

The goal would be to remove the nodes that are used as place holders for
holding clipping paths. These are used when guassian blurs are used because
flash inexplicably blurs the clipping path if it is on the same node as the
blur filter. A good test case is gallardo.svg. There is a blurred path next
to the rear view mirror.
Proposed removal in r141, r146, and r149 by James.

The current implementation checks the attribute and recursively checks
parents to determine whether to draw. Proposed changes in r143 remove this
check. 

Separate the retrieval of gradient values into another function. Proposed
improvements are in r147 and r157 from James.

Create SVGEvent class and improve event handling. See improvements in r154
and r155 from James.

The current implementation of parseTransform(), which is mainly used for
parsing gradient transforms right now, is missing several features that
exist in transformNode(). Proposed improvements are in r116 from James.

Integrate the console.log code I have for Internet Explorer and change
Rick's branch to use this instead of svgviewer.info.

Separate SVGViewer base class and change the Makefile to build the
different flavors. Proposed improvements in r152, r153, and r159 from James.

The current mouse targeting implementation which attempts to simulate SVG 
behavior is complex and the goal is to replace the multitude of function
closures which handle target dispatching with a single handler which uses
flash targeting variables. See SVGViewerWeb handleAction and related code
in r159 from James.
To keep the photos.svg demo working, event.currentTarget is preferred over
event.target in handleAction.

The current implementation creates unique ids unnecessarily for the cloned
node. There should be no need to register an id at all for USE nodes.
Proposed improvements are in r115, r134, and r157 from James.

r185

Add basic support for Pattern element using Graphics.bitmapFill().

The current implementation copies the xml of referenced nodes
(recursively). The proposed improvement is to retrieve the values on
demand. See r141 and r143 from James.

See testing/svg/shapes-polyline-01-t.svg

The top right and bottom center shape render with a closed border which
should remain open for the last segment. In other words, flash is
automatically doing a lineTo from the last point of the polyline to the
first point of the polyline.

Possible Cause:
Unfilled polylines on this test are OK. Also, when endFill is called right
after beginFill on the filled polylines, then the border line does not
close, which is correct. This leads to the speculation that calling endfill
after drawing a polyline causes flash to automatically close the polyline
potentially as a side effect of doing the fill.

Potential Solution:
A potential workaround to explore would be to draw the polyline twice, once
with a fill and no border, and once with a border and no fill.

Improve animation implementation. 
Proposed improvements in r132, r140, and r166 by James.

Proposed change is to parse XML children recursively from the constructor
of the nodes. The current implementation parses children on the frame
render  event. Proposed improvement is in r141 and r143 from James.

r186

Change rgb % calculation to use 255 instead of 256.  
256 falls outside the possible value range of 0-255.

See test: color-prop-03-t

svg_inline.html does not work; get it working.

Integrate the browser detection code I have, which was adapted from Dojo.
Browser detection is trickier than just looking at the browser string,
since some browser's like Opera can masquerade as IE.

Proposed improvement in r157 from James.

I've created a license file that details any third party code that I've
used; I've mostly copied over my own details from my fork. Most of the
things I refer to I haven't yet brought into Rick's svg.js fork.

There is a workaround in the code for a rendering bug when a negative
transform is used on a node that has a gradient. The gradient was not doing
the right matrix calculation. A workaround was found which was to separate
the transform from the node although the code was rather ugly. A proposed
simplification in r141 from James removes this workaround. A reproducible
case of this problem will need to be found to confirm whether this
workaround is still needed.

The goal is to simplify the code that handles setting and retrieving styles
and attributes of nodes. Proposed improvements are in r117, r118, r144,
r155 and r157 from James.

This issue is to track my own local build system's integration into James'
Makefile.

r191

Fix for polygon and polyline point parsing.
Provides support for points sets delineated by a mix of commas and spaces.
Also removes automatic path closing for polyline.

See tests: shapes-polygon-01-t, shapes-polyline-01-t

The goal is to remove the convoluted undo calculations currently in
transformNode that are used to reverse the effects on x and y. Proposed
improvements in r143 and r145 from James.

r187

In beginNodeFill(), when setting capsStyle, an 'else' is missing.

            if (capsStyle == 'round'){
                capsStyle = CapsStyle.ROUND;
            }
            /*** The following should be else if **/
            if (capsStyle == 'square'){
                capsStyle = CapsStyle.SQUARE;
            }
            else {
                capsStyle = CapsStyle.NONE;
            }

Currently the gradient calculations assume userSpaceOnUse. Supporting
objectBoundingBox requires tracking the object width and height. Proposed
improvements in r156 and r157 from James. 

What steps will reproduce the problem?
1. Look at w3c test full-text-align-01-b.html

What is the expected output? What do you see instead?
  The text-anchor:end text should end exactly on the middle of the black
dot but it currently displays about 2 pixels off to the right. If the -2
pixel adjustment is made without accounting for size or scale, then the
blocks_game.svg used in demo.html does not display correctly because it
uses a 1px high font which is then scaled bigger with a transform. A -2px
adjustment ends up throwing that text off the screen.

Proposed Solution:
Create a targeted test image with various sizes of text and various
transforms and find the scales where the adjustment helps and where it does
not. Modify the end of SVGTextNode.setAttributes() to adjust
this._textField.x by an amount appropriate for the font size. If that does
not work, try a combination of increasing the font size, adjusting x, and
applying an inverse transform to produce the correct solution.

r193

Support transformation using spaces to separate arguments

Change getHeight and getWidth to retrieve height and width from SVGViewer
when needed and to support % better. Proposed improvement in r142 from James.

What steps will reproduce the problem?
1. Load the demo.html
2. The first image "scimitar" should be selected
3. Press the > button until the image moves far to the right

The white highlight of the blade does not render correctly. It appears to
have a mask that is not transformed the same as the rest of the image. 

Pressing the > button adjusts the viewBox attribute of the top SVG element.
This works correctly in revisions prior to r174.

Move transform handling to when image is loaded. Proposed improvement in
r157 from James. See related Issue 25.

Script nodes require special handling to handle new lines. This code
belongs in the Viewer, not in SVGNode. Proposed improvement in r141 from James.

Nodes within a DEF node are not drawn. They are available to be used by
other nodes. The current implementation has a recursive check of ancestors
to avoid drawing when an ancestor is DEF node. A proposed change to DEF in
r143 and r157 from James removes this check.

r188

Add implied line support for SVGPathNode.

See test: paths-data-09-t

[deleted issue]

This is a general placeholder issue as I work to integrate the bulk of my
JavaScript into what Rick has done. I'll record any revisions that affect
this integration here; some of the checkins might be in flux for awhile as
I figure out the best way to do things. I'll indicate when its safe to
merge these back into the main branch.

There are several places where numbers need to be trimmed or stripped
better. See proposed improvements in r129 and r157 from James.

Bringing in photos.svg sample and the Tetris-clone blocks_game.svg after
getting permission from the authors. Modifying demo.html to be able to use
these.

r184
Enabled cacheAsBitmap for both Masks and Clip-Paths.  
This is needed to support using device fonts in TextField to mask a node.  

See test: masking-path-04-b

Grouping of functions and comments could use improvement in SVGNode.as.
Proposed improvements are in r158 from James.

The goal is to simplify the support for when to update nodes that are
referring to another node via href that has changed. The current
implementation keeps track of references. The proposed change of href
support to on demand removes invalidateReferers(). See Issue 19 and r143
from James.

See proposed improvement in r157 from James.

The goal is to only set a few attributes in setAttributes() which
simplifies it greatly. Proposed improvement is in r128 from James.

r196

Improve gradient bounds calculations.
Support % calculations for x1, x2, y1, y2
Support value scope for "objectBoundingBox" unit space


See tests: pservers-grad-10-b, pservers-grad-09-b, pservers-grad-07-b,
pservers-grad-06-b

Create core and remove data directory as proposed by James in r130, r131,
and r133.

Integrate improvements: separating SVGViewer flavors, getting and setting
attributes, code structure of gradients, transforms, cloning, parsing, etc.