ROADMAP

This document sets the official future development goals of IndieK v0.1.0
Date: December 22, 2017
Author: Adrian Ernesto Radillo

Vision

If you wish to read about my global vision for IndieK, check out the VISION.md document and my blog.

Development

I am the only core developer at the moment.

Disclaimer: Because IndieK is a FOSS, we greatly encourage public debate and discussions about IndieK’s development and design questions. However, as head developer of the project, I (the author of this document) make all the final decisions about the IndieK’s official directions of development. Of course, I shall always justify my decisions publicly, as best as I can.

Needed collaborators

If you are a developer and would like to get involved with IndieK, please get in touch!

Current architecture of IndieK

Arch

Immediate development goals

1. Port existing code to Python

Why python?

We want a GUI
We will use machine learning algorithms in the future, and computations on graphs

One aspect of this migration involves reproducing the existing CLI with pure Python / SQL code. A good source of inspiration is this tutorial.

Enhance architecture
Make use of Object Oriented (OO) capabilities of Python. Below is a sketch of some objects and methods.

Classes:

Class name: Item

Attributes: (The list below is merely reporting the info contained in the SQL DB)

ItemID
Created
LastModifDate
Author
Content
Latex
Parents
Children
InTopics

Methods:

query_db — fills Python object with data fetched from SQL DB
save_db — writes Python item to SQL DB
delete_db — removes item from DB
plot — sends item to GUI for graphical interaction
edit — edits item properties in Python workspace (without writing to SQL DB)
delete — destructor, may not be needed

Class name: Topic

Attributes:

TopicID
TopicName
TopicDescription
Created
LastModifDate
Author
ItemList
SupraTopics
SubTopics

Methods:

Class name: Graph
Contains (and fetches if needed) required info to produce GraphViz output

Class name: Query

Attributes:

Author

Methods:
All the possible queries that the user might make with the application. Also, internal queries corresponding to graph computations, such as:

GetTopicAncestors
GetTopicDescendents
GetItemAncestors
GetItemDescendents

Class name: Workspace

The workspace is the place where most of the interaction with the user will take place. This interaction will happen via the freehand drawing module of the GUI. Whenever the user will work on some project, he/she will use the workspace as a whiteboard. Also, the GUI will allow for dragging-dropping objects from the graphical results windows into the workspace window.

Workspace should allow import of external images. Should also provide a scanning option, to run OCR on images.

Adopt coding guidelines
Mainly, use a specific template for class and function definitions . In general, use PEP 8.

Hunt for bugs
We will start to systematically document the application’s bugs.

2. Start developing a GUI

Note that developing a GUI will not render the CLI obsolete. We will keep the CLI development ongoing. The GUI will have two aspects.

All URLs should be clickable

Standard button-fields-display GUI

For the graphs, keep using GraphViz

Window 1: Queries

Window 2: CLI output

Window 3: Graphical results (to display GraphViz graphs). Will offer minimal interaction (mainly translation, zoom in and zoom out).

Window 4: Workspace

Touch-screen freehand drawing GUI
This part of the interface will receive freehand drawings and freehand writings as input from the user. The input device will most probably be a drawing tablet or a touch screen. This part of the GUI will need to send its produced images to the OCR module of IndieK for further processing.

A GUI development tool must be chosen. PyQt5?

Important question: Do we go for vector graphics or raster images?

OCR
The main goal is to convert:

handwritten letters to their UTF-8 counterpart
handwritten mathematical symbols to their LaTeX counterpart

A pre-processing of the image, before the actual OCR operation, will be needed. Ideally we seek automatic image segmentation, where the different parts of the handwriting area are classified into one of the following:

word sequence → UTF-8
mathematical expression → LaTeX
freehand drawing → Possible image file OR vector graphics file
noise / unclassified

Convnets are the best for handwritten digits classification. Maybe start with one of these trained CNN.

3. Database

Extending SQL support
We want to develop the support for IndieK to the following widely used SQL frameworks:

PostgreSQL
SQLite

Recall that the FOSS nature of IndieK restricts our choice of databases to pure FOSS (AGPL 3 compatible).

A promising module for SQL is dataset (I started using it with PostgreSQL for school).

Start experimenting with NoSQL
We also want to start experimenting with NoSQL and graph databases, to comparatively weigh the benefits and drawbacks with SQL, as far as IndieK’s goals are concerned.

4. Automatic search

The search capabilities of IndieK are crucial to the usefulness of the software. There is no point in using IndieK, if finding back stored information is a pain. A powerful Open Source engine is PyLucene.

5. Markdown export

IndieK should also facilitate the publication process. A quickly attainable goal is to write a function that allows the export of a topic to a MarkDown document.

6. Write new functions for IndieK

To-do module
We want to incorporate into IndieK a functionality for handling TO-DO lists.