Guidelines for rules that are included in the standard distribution
Note:
These guidelines are new and most rules don’t follow these guidelines yet.
The goal is, that eventually all rules are updated.
Why do we need these guidelines?
- To prevent low quality contributions
- To reduce time reviewing rules
They just apply to rules included in the standard distribution.
Requirements for standard rules
To be included in stock PMD, a rule needs
- Broad applicability. It may be specific to a framework, but then, this framework should be widely used
- Solid documentation. See below
- If it’s a performance rule: solid benchmarks. No micro-optimization rules
- No overlap with other rules
Dos/Don’ts (rule rules)
- Rule naming
- Don’t put the implementation of the rule in the name, because it will be awkward
if the scope of the rule changes
- Eg. SwitchStmtShouldHaveDefault -> since enums are a thing they don’t necessarily need to have a default anymore, they should be exhaustive. So the rule name lies now…
- Eg. MissingBreakInSwitch -> it’s obvious that this is supposed to find fall-through switches. Counting breaks is not a clever way to do it, but since it’s in the name we can’t change it without renaming the rule.
- Do use rule names that name the underlying problem that violations exhibit
- Eg. instead of SwitchStmtShouldHaveDefault, use NonExhaustiveSwitchStatement -> this is the problem, the description of the rule will clarify why it is a problem and how to fix it (add a default, or add branches, or something else in the future)
- Eg. instead of MissingBreakInSwitch, use SwitchCaseFallsThrough
- Don’t create several rules for instances of the same problem
- EmptyIfStmt and EmptyWhileStmt are actually the same problem, namely, that there’s useless syntax in the tree.
- Don’t limit the rule name to strictly what the rule can do today
- Eg. UnusedPrivateField is a bad name. The problem is that there is an unused field, not that it is private as well. If we had the ability to find unused package-private fields, we would report them too. So if one day we get that ability, using a name like UnusedField would allow us to keep the name.
- Don’t put the implementation of the rule in the name, because it will be awkward
if the scope of the rule changes
- Rule messages
- Do write rule messages that neutrally point out a problem or construct that should be reviewed (“Unnecessary parentheses”)
- Don’t write rule messages that give an order (“Avoid unnecessary parentheses”) especially without explaining why, like here
- Don’t write rule messages that are tautological (“Unnecessary parentheses should be removed”). The answer to this would be an annoyed “yes I know, so what?”.
- Do use Markdown in rule descriptions and break lines at a reasonable 80 chars
- Do thoroughly comment rule examples. It must be obvious where to look
- Do comment your xpath expressions too
Rule description template
- What the rule reports (1 summary line)
- Why the rule exists and where it might be useful (including, since which language version, etc)
- Blank line
- Explain all assumptions that the rule makes and keywords used in the previous paragraph. (“overridden methods are ignored”, “for the purposes of this rule, a ‘visible’ field is non-private”).
- Describe known limitations if any
- Blank line
- For each property, explain how it modifies the assumptions and why you would want to use it. If you can’t explain why it’s there then it shouldn’t be there!