Add MethodInvocationCounter Metric to Track Method Invocations within Classes

[BrenoFariasdaSilva]

Overview

This PR introduces the MethodInvocationCounter metric, a new ClassLevelMetric designed to analyze and report on method invocation patterns within Java classes. The metric captures the frequency of invocations for each method within a class, providing insights into method dependencies and utilization.

Key Changes

1. CKClassResult.java (REFACTOR):

   • New Data Member: Added a private string field methodInvocations to store the formatted string representing method invocation statistics for each class.
   • Accessor and Mutator Methods: Added setMethodInvocation(String methodInvocations) and getMethodInvocations() to handle this new data. This allows the stored invocation data to be set during metric calculation and later retrieved for output.

2. MethodCounter.java (FEAT):

   • New Utility Class: Introduced to assist with counting and formatting method invocations. This class includes nested classes and methods that:
     • Store method invocation data (MethodInformation class).
     • Sort and format the collected data into a readable format (formatResult and toFormattedString methods).
   • This utility class provides a centralized way to manage method invocation data, ensuring consistency and reusability across different parts of the metric implementation.

3. MethodInvocationCounter.java (FEAT, FIX, REFACTOR):

   • Metric Implementation: This file defines the MethodInvocationCounter class which implements CKASTVisitor and ClassLevelMetric. It contains the logic to visit method declarations and invocations, and it populates the MethodCounter.MethodInformation data structure with this information.
   • Visitor Methods: Implements the visit(MethodDeclaration node) to set the current method context and visit(MethodInvocation node) to record invocations within the context of the current method.
   • Set Results: Overrides setResult(CKClassResult result) to format and set the method invocation data on the CKClassResult.

4. ResultWriter.java (REFACTOR):

   • Output Addition: Modified to include the methodInvocations in the output of the CK tool. This ensures that the collected method invocation data is written out alongside other metrics.

Implementation Details

The MethodInvocationCounter leverages the AST (Abstract Syntax Tree) to visit and analyze MethodDeclaration and MethodInvocation nodes within Java source files. For each class, it maintains a mapping of methods and the methods they invoke, along with the invocation counts. This data is then formatted into a human-readable string that summarizes the invocation patterns, making it easy to identify which methods are central to the functionality of the class.

Output

Here is an example of the Output generated when using CK to generate the class.csv for the ZooKeeper repository:

Usefulness

The MethodInvocationCounter metric is particularly useful for:

• Identifying Key Methods: By highlighting methods with high invocation counts, developers can quickly identify the core components of a class.
• Refactoring Opportunities: Methods that are heavily dependent on others or have high complexity might be candidates for refactoring to improve modularity and maintainability.
• Optimizing Performance: Understanding method invocation dynamics can help in optimizing performance-critical parts of the application.
• Documentation and Understanding: This metric provides a quick overview of class behavior, which can be beneficial for new developers or in documenting the system.

Conclusion

The addition of the MethodInvocationCounter provides valuable insights into method usage patterns, making it a useful tool for developers looking to maintain and improve the quality of their Java applications.

[mauricioaniche]

This MR is great, thanks for the feature!

However, this one makes the output way larger. Therefore, this feature should be off by default, and enabled only optionally!

[BrenoFariasdaSilva]

Hi again, @mauricioaniche.
So, the changes were made. Basically, the CK and Runner classes were modified to handle a new boolean execution argument (verbose).
In addition, the ClassLevelMetric interface was modified to include a new method named isVerbose, whose default return is false. With that in mind, any new class metric can overload this method and set the return value to true, which was done in the MethodInvocationCounter class.
If the verbose execution argument is set to true, every metric will be processed. Otherwise, it checks if the metric is not verbose, so it can be processed.

I acknowledge that this PR is massive. If you prefer, i can split it into two PRs: One for the verbose execution argument and another for the new class metric itself (MethodInvocationCounter).
Also, keep in mind that the build is failing. That's due to the fact that my older PR that fixes the build was not approved yet (#115). It's a simple PR, so it's fast to analyze.

[BrenoFariasdaSilva]

Hi @mauricioaniche. The readme's conflict is solved :).

[BrenoFariasdaSilva]

All unnecessary changes were undone

[BrenoFariasdaSilva]

The unnecessary changes were undone.

[mauricioaniche]

Thanks for the PR!

My main concern is about the new metric being a very long string. I think we should:

• Make this feature disabled by default, so we don't produce such large extra data for people that are already using CK
• Output it to a different file, so the CSV file stays short and sweet

[mauricioaniche]

There's a typo here, Additionally

[mauricioaniche]

verbose flag, flag shouldn't be highlighted

[mauricioaniche]

The string produced by this class can be really long. I'm not sure I'd like to mix the CSV, which is mostly numbers with this super large string. Also, given that this string can be quite long, just concatenating it to the CSV might make the CSV invalid. We'd need to use proper escaping here. An alternative would be to output this to a separate file. And make this feature disabled by default, so we don't start producing large data files to people that have been using CK so far (and do not expect it).