Skip to content

Latest commit

 

History

History
64 lines (38 loc) · 2.19 KB

README.md

File metadata and controls

64 lines (38 loc) · 2.19 KB
Raw

Dziban

WIP

Prototype Status

Current prototype: MKIV.

Tutorial & Demo

Check out the Notebook.

Here's a quick peek.

img

Installation

  1. Clone the multi-vis branch of draco, found here. And follow the installation instructions.

  2. Install dependencies: pip install -r requirements.txt.

  3. Install Dziban: pip install -e ..

  4. Start up a notebook: jupyter notebook.

  5. Import Dziban and perhaps a dataset from vega_datasets.

from dziban.mkiv import Chart
from vega_datasets import data

Follow the demo for usage examples.

The important stuff

base.py <>

base definition for a chart, including initialization

chart.py <>

programming layer, exposed to user

encoding.py <>

AST node for encoding objects

field.py <>

field methods for a chart, exposed to user

graphscape2asp.js <>

script for translating Graphscape definitions and weights into ASP (to be placed into Draco). See below.

In addition, Dziban uses a modified version of Draco, which supports multi-view and Graphscape's transition reasoning. This can be found on this branch. There are changes to the original Draco scattered about. Noteably:

compare.lp <>

transition definitions for Graphscape

compare_weights.lp <>

weights for these transitions

The whole system

has been modified to allow for multiview reasoning (many other files, e.g. define.lp, soft.lp, optimize.lp, run.py, asp2vl, optimize_draco.lp, optimize_graphscape.lp, and more and more).