Improve the SVG parser

There are many weird quirks of the current custom SVG parser. For example, the colors white and black are swapped if the SVG element is a rect. This is some really odd behavior, and I think the best way to solve these issues is to use a full SVG parsing library. Either way, SVG support is pretty limited at the moment. (Also, I think I can add SkewX and SkewY transform support right now)

[xy-23]

Hi, I was trying to improve the svg parser sometime ago, but during my attempt, I found if we want to improve it (more stable and powerful), we need to deal with some problems first.

1. We need to let SVGMobject support <text>, <image> and so on. But currently SVGMobject is a subclass of VMobject. I think we should let SVGMobject become a subclass of Group, which I think is more intuitive. But then you will find that Group doesn’t support many existingAnimations e.g. Write, and you will also notice that there is a subclass of VMobject called VGroup, so far, I have made a new class called Image which is a subclass of VMobject/.../Rectangle and is able to be animated in those Animations, but I think it is a compromised way. Then we need to dig into the deepest process of manim, if you are interested, please tell me, because I think this is a little offtopic…

2. We need to PARSE svg in a better way in order to let it support attributes like color, stroke and so on. I found a python package called cairosvg, which does this job almost perfectly, but it can’t be used directly in manim, so I think we can probably look into the source code of it and use(copy) some of it such as its class called Tree

Because I’m busy with other things now and don’t have much time to do it. If you want to improve it, I hope my experience can help a little…

[xy-23]

Keep in mind that Image is also a part of PIL, so to minimise confusion, a better name should be chosen imo.

I just use the name Image to distinguish the current ImageMobject, but I think using the name Image, Text, Tex, SVG, is simpler and more convince to use from user‘s perspective than ImageMobject, TextMobject, TexMobject, SVGMobject (though they are not the same thing currently).

The confusion of the PIL Image can probably be solved by using PIL.Image.

Maybe we can do something like this:

class ImageMobject():
    def __init__(self):
        pass

and create a abbreviation version

class Image(ImageMobject):
    pass

But this is not very important…