diff --git a/rules/S7172/cfamily/metadata.json b/rules/S7172/cfamily/metadata.json index 8d6d7d4efac..801695cea93 100644 --- a/rules/S7172/cfamily/metadata.json +++ b/rules/S7172/cfamily/metadata.json @@ -1,12 +1,12 @@ { - "title": "FIXME", + "title": "Named methods should be used to avoid confusion between testing an optional or an expected and testing the wrapped value", "type": "CODE_SMELL", "status": "ready", "remediation": { "func": "Constant\/Issue", - "constantCost": "5min" + "constantCost": "10min" }, - "tags": [ + "tags": ["confusing" ], "defaultSeverity": "Major", "ruleSpecification": "RSPEC-7172", @@ -16,10 +16,8 @@ "quickfix": "unknown", "code": { "impacts": { - "MAINTAINABILITY": "HIGH", - "RELIABILITY": "MEDIUM", - "SECURITY": "LOW" + "MAINTAINABILITY": "HIGH" }, - "attribute": "CONVENTIONAL" + "attribute": "CLEAR" } } diff --git a/rules/S7172/cfamily/rule.adoc b/rules/S7172/cfamily/rule.adoc index 69c2a03a830..7129aef68d6 100644 --- a/rules/S7172/cfamily/rule.adoc +++ b/rules/S7172/cfamily/rule.adoc @@ -5,40 +5,124 @@ FIXME: add a description == Why is this an issue? -FIXME: remove the unused optional headers (that are commented out) +`std::optional`, `boost::optional`, and `std::expected` are types that allow to represent a contained value of a certain type, and additional data that denote the absence of a value (`nullopt` for `optional`, and error values for `expected`). To check if the wrapper contains a value, it is possible to call the member function `has_value()`. It is equivalent to converting the wrapper to `bool`, which can be more concise, especially when used in a place that allows _contextual conversion to bool_ (for instance, the condition of an `if`). It means that the following syntax is valid: -//=== What is the potential impact? +[source,cpp] +---- +std::optional flag; +if(flag) { ... } +// Equivalent to +if(flag.has_value()) { ... } +---- + +When the wrapped type is also convertible to `bool`, using this concise syntax can be confusing. What is tested: The wrapper, or the wrapped value? +This rule raises an issue when `std::optional`, `boost::optional`, or `std::expected` wrap a basic type, and the presence of the value is tested with the concise syntax. However, if in a single expression, the presence of the value is tested and the value is accessed, the risk of confusion is reduced, and no violation is raised. + +=== What is the potential impact? + +There are two possibilities, depending on the initial intent of the autor of the code: + +- If the intent was actually to test the presence of the value, the code is correct, but it is not clear. When reading it, people might think that the wrapped value is tested, and not the presence of the value. + +- If the intent was to test the wrapped value, the code is incorrect. This situation can especially happen when evolving code that worked directly with values to work with optional or expected values, and forgetting to update the test. This will lead to code that does not behave in the intended way, but still works in a plausible way, and the lack of clear clue on twhat is tested makes finding the issue difficult. == How to fix it + +The most direct way to fix this issue is to replace the concise syntax with a more verbose call to `has_value()`. However, both syntaxes leads to a code that is full of `if`/`else` and becomes difficult to read, especially if the program works with many `optional` or `expected` values. + +In many circumstances, it is possible to write code in a more direct way, using member functions such as `value_or()` or, starting with {cpp}23, the monadic interface of `optional` and `expected` (`and_then()`, `transform()` and `or_else()`), which are especially convenient when chaining operations on those types. + + //== How to fix it in FRAMEWORK NAME === Code examples +[source,cpp,diff-id=2,diff-type=compliant] + ==== Noncompliant code example -[source,cpp,diff-id=1,diff-type=noncompliant] +[source,cpp] ---- -FIXME +void perform(Node node, optional recursive, Action action) { + action(node); + bool recurse; + if (recursive) { + recurse = *recursive; + } else { + recurse = getDefaultRecurseMode(); + } + if (recurse) { + // perform recursion... + } +} ---- ==== Compliant solution -[source,cpp,diff-id=1,diff-type=compliant] +The most direct solution is to replace the concise syntax by the more explicit one. + +[source,cpp] +---- +void perform(Node node, optional recursive, Action action) { + action(node); + bool recurse; + if (recursive.has_value()) { + recurse = recursive.value(); + } else { + recurse = getDefaultRecurseMode(); + } + if (recurse) { + // perform recursion... + } +} +---- + +Since the rule will not trigger if the value is accessed in the same expression where its presence is tested, another possibility if the following: + +[source,cpp] +---- +void perform(Node node, optional recursive, Action action) { + action(node); + bool recurse = recursive ? *recursive : getDefaultRecurseMode(); + if (recurse) { + // perform recursion... + } +} +---- + +In this simple case, the use of both `recursive` and `*recursive` hints that recursive is more than a simple `bool`. However, it is possible to write the code with a higher level semantic: + +[source,cpp] +---- +void perform(Node node, optional recursive, Action action) { + action(node); + bool recurse = recursive.value_or(getDefaultRecurseMode()); + if (recurse) { + // perform recursion... + } +} +---- +The downside of this version is that `getDefaultRecurseMode()` is called even if `recursive` contains a value. If `getDefaultRecurseMode()` is a complex function, it can be a performance issue. In this case, it is possible to use the monadic interface of `optional`, at the cost of a more complex syntax: + +[source,cpp] ---- -FIXME +void perform(optional recursive) { + bool recurse = recursive.or_else([]() -> optional { return getDefaultRecurseMode();}).value(); + if (recurse) { + // perform recursion... + } +} ---- -//=== How does this work? +This syntax is more complex, but it is also more flexible, especially when chaining operations on `optional` or `expected` values. -//=== Pitfalls +== Resources +=== Documentation -//=== Going the extra mile +* {cpp} reference - https://en.cppreference.com/w/cpp/utility/optional[`std::optional`] +* {cpp} reference - https://en.cppreference.com/w/cpp/utility/expected[`std::expected`] +* {cpp} reference - https://en.cppreference.com/w/cpp/utility/optional[`std::optional`] +=== Articles & blog posts -//== Resources -//=== Documentation -//=== Articles & blog posts -//=== Conference presentations -//=== Standards -//=== External coding guidelines -//=== Benchmarks +* Microsoft Developper Blog - * https://devblogs.microsoft.com/oldnewthing/20211004-00/?p=105754[Some lesser-known powers of std::optional]