DESIGN.md

In decisionpatterns/coercion: Utilities for Coercing (and Parsing) Values Between Types, Modes and Classes

TODO.md - coercion package TODO list

Mixed formats, Conditional Transformation

One often encountered problem in reading in data is handling data in which there was little or no validation of the input. This occurs often with webforms that allow free-text entry for a field that has a natural structure. A typical example is date form-field that does not specify a date format. The data entered may be entered in a number of formats: mm/dd/yyyy or yyyy-mm-dd. Some packages such as lubridate or anytime attempt to read these values, but envariable they often do not catch all of the values. Some manual intervention is necessary.

The manual approach is rather tedious. Here we provide the facilities for trying multiple conversions/transformations choosing the correct one. This is not trivial as the results:

• may not be correct,
• may not be non-NA values
• may be ambigous (i.e. multiple coercisions producing valid values)

Coersion should handle this.

For example, p_excel_dt.character should first try mdy_hms, mdy and p_excel_dt.numeric in that order.

A CASCADE PROCEDURE

This is like a similar to a vectorized na_replaces

Have a number of coercisions

• Initialize NA vector to length of input
• Try the first using na_replace replace na values with the

Validations

Transformations are all about capturing data in its rawest form an capturing its data using semantics, context, etc.

It is expected that there is a validation for the informations. For example, if you are expecting the height of a person, you know that a person's height should be a numeric, continuous value and can not be more than 8'11.1".

These validattions should be included.

Units

decisionpatterns/coercion documentation built on Nov. 4, 2019, 10:23 a.m.