Overview

CodeInspector is a linter and static analysis tool for Wolfram Language code.

Stages of rules

There are multiple stages of rules that are applied.

Token rules

Certain symbols have typos or do not exist.

Example: Boolean is not a System` symbol.

Concrete rules

Where whitespace and newlines matter.

Example: implicit Times across lines.

Aggregate rules

Various issues with operators and precedences.

Abstract rules

Function calls themselves.

Severity levels

So this brings up the interesting problem of how to categorize these problems.

I have settled on 2 dimensions (for now): Severity and ConfidenceLevel

If something is indeed a problem, then Severity should denote how severe it is. Will it ever impact users? Will it accidentally launch nuclear warheads?

ConfidenceLevel is supposed to denote the confidence that this is actually a problem. ConfidenceLevel -> 0.0 means no confidence at all, while ConfidenceLevel -> 1.0 means something like mismatched brackets in a function.

Note that even “obvious” problems don’t necessarily have ConfidenceLevel -> 1.0.

Remark

Issues such as character encoding assumptions, formatting issues.

Example: A character was saved as extended ASCII, and cannot be treated as UTF8.

Example: Token is not on same line as operand.

Warning

Most likely unintended uses of functions.

Error

Recoverable errors in syntax.

Code that is never intentional.

"\[Alpa]"

f[a,]

Fatal

Unrecoverable errors in syntax.

1 + {}}