Split the library

[tomek-he-him]

Right now the library does two things implicitly:

1. parse an XML tree looking for parametric attributes, returning the so-called VirtualTree;
2. update the XML tree with new calculations.

I’d like to split the two steps into two separate libraries – a parser and a patcher. They’d communicate using a separately-versioned AST format. Many other low-level libs have gone this way and succeeded (http://npm.im/commonmark, http://npm.im/virtual-dom, http://mdast.js.org/).

We get lots of benefits:

👍 The parser can take any form of XML (for example, from a saved SVG file), generate an AST and pass it over to any other patcher (for example, a virtual-dom-based one).

👍 The AST can be precompiled and loaded separately as a JS file. That would radically boost performance for complicated stuff.

👍 The patcher can take an AST generated by another parser (for example, by some non-JS plugin for Inkscape).

👍 The API will be explicit, without overloading – easier to document and understand.

I see one tradeoff there however:

👎 The learning curve will be higher. Non-programmers (designers, makers, etc) probably won’t like the API.

Perhaps I’ll leave the API as it for now – and work on the parser and generator separately. When the low-level stuff is ready, I’ll wire it up with this lib. The API here will keep its implicit, jQuery-like spirit.

I’m still open to other ideas though!

[tomek-he-him]

Update: there is some work going on already: https://github.com/parametric-svg/ast.

[tomek-he-him]

Moved to projectshaped/parametric-svg#21.