iceclassic

Toolkit for exploring data.

Long-term plan: public PyPI package with tools for manipulating data easily.

Context

Primary use:

• used in class settings as a demonstration tool
• used by students to work on assignments
• used in interactive TeachBooks pages
• open source community for expanding applications

Related:

• iceclassic documentation: Sphinx pages (numpy documentation style) that illustrates the package features only (TeachBook for application)
• TeachBook that introduces the Ice Classic, the package and explores various science, engineering, programming, modelling concepts
• Contributors should somehow be able to set up more complex analyses and share them with the community

Features

Highest level:

• summarizes breakup record (and includes data in package) via ability to export data in many formats: as output (formatted tables, figures, etc) as well as in several ready-to-go data types (ndarray, dataframe, etc)
• includes a short list of extremely simple models (e.g., linear regression, univariate probability distributions, etc)
• facilitates exploration of modelling concepts (calibration, verification, validation)
• standardizes the way a prediction is defined, presented, etc
• let's one choose/explore assumption of "start date" for the year or "reference date"
• let's one explore the concept of prediction, extrapolation, etc (i.e., which info do you include, prediction variables, etc)

Lower levels:

• students will be asked to read documentation and read code explicitly to teach good practices for programming
• if possible, consider usage of various data types as well as OOP versus functional paradigms
• names of objects are chosen very carefully
• Consider how decorators, BMI, etc, could be used to incorporate contributions and modularity
• advanced visualization is supported by the package via the design and implementation, but not accomplished internally
• figures must be handled carefully to ensure that best practices are illustrated; must be easy to produce and modify them

Implementation

• pure Python package
• minimize dependencies (limit to common tools like Numpy, Scipy, etc)
• avoid large packages (import in webassembly should be fast)
• consider adopting BMI, or related frameworks
• object oriented, but not exclusively so
• should come packaged with a set of essential data (e.g., breakup record, discharge, stage, temperature, precipitation, snowfall, etc)

Can create a second package for "heavy" stuff, if needed. For example, setting up and running models.