JaSoMe: Java Source Metrics

Jasome (JAH-suhm, rhymes with awesome) is a source code analyzer that mines internal quality metrics from projects based on source code alone. This distinguishes Jasome from similar tools by not requiring the project first be compiled, or even compilable.

Most analyzers only work on projects that successfully compile, which in the case of Java projects means all dependencies must be satisfied, any external libraries are properly on the classpath, any code generation utilities must properly run, and so on. However, when all you need is the quality metrics for Java source code, this compilation requirement is unnecessary since most metrics can be derived from the source code alone.

Jasome scans a given directory for .java files and performs Best Guess Metrics Analysis to derive various quality metrics, and then outputs those metrics on a a per-file basis as XML. Each XML output will contain the name/path of the file analyzed, and the values of its per-class and per-method metrics. It will also list directory structures and their per-package metrics.

Best Guess Metrics Analysis means Jasome will continue collecting metrics even when individual files make it difficult. When code references methods or objects defined in dependencies that may or may not be on the class path, Jasome will assume those dependencies are present and properly compilable. If Jasome cannot make a clear determination for analysis, it will make a best guess and output metrics whose values are as close as possible to the actual metrics values one would achieve if analyzing the data with a metrics engine that relies on proper compilation. When a java class is syntactically invalid and would not compile, Jasome will either skip the offending methods or classes, or skip the file entirely but continue processing the remaining files. In other words, there are definitely ways someone can manipulate their code to intentionally change the measurements for metrics, but by and large on a normal project Jasome will be accurate.

Getting Started

Download the latest distribution and unzip, change into directory, then run:

bin/jasome <directory to analyze>

JaSoMe will gather metrics and output them to the console. You can save the XML to a file using the --output <file> option.

JaSoMe runs in Linux and MacOS - Windows is not supported and there are reports that it will not run properly in Windows.

Metrics

Jasome is either currently tracking or planning to track the following metrics:

• Raw Total Lines of Code (RTLOC) - The actual number of lines of code in a class, using the line numbers of the file itself. Comments, whitespace, and everything else is counted. (class)
• Total Lines of Code (TLOC) - The total number of lines of code, ignoring comments, whitespace, and formatting differences (project, package, class, method)
• Number of Attributes (NF) - The number of fields/attributes (class)
• Number of Static Attributes (NSF) - The number of static attributes (class)
• Number of Public Attributes (NPF) - The number of public attributes (class)
• Number of Methods (NM) - The number of methods (class)
• Number of Static Methods (NSM) - The number of static methods (class)
• Number of Public Methods (NPM) - The number of public methods (class)
• Number of Classes (NOC) - The number of classes within a package (package)
• Number of Parameters (NOP) - The number of parameters a method takes (method)
• Depth of Inheritance Tree (DIT) - The maximum depth of the inheritance hierarchy for a class. (class)
• Number of Overridden Methods (NORM) - The number of methods a class overrides or implements from a parent class (class)
• Number of Inherited Methods (NMI) - The number of methods a class inherits from parent classes (class)
• Number of Methods Added to Inheritance (NMA) - The number of methods a class inherits adds to the inheritance hierarchy; methods defined on the class that it did not override or inherit (class)
• Specialization Index (SIX) - How specialized a class is, defined as (DIT * NORM) / NOM (class)
• Number of Methods Inherited Total (Mit) - Methods inherited overall (class)
• Number of Methods Inherited (Mi) - Methods inherited but not overridden (class)
• Number of Methods Defined (Md) - Methods defined within class (same as NMA) (class)
• Number of Methods Overidden (Mo) - Methods in class that override an otherwise-inherited method (class)
• Number of Methods (All) (Ma) - Methods that can be invoked on a class (inherited, overridden, defined). Ma = Md + Mi Same as NORM. (class)
• Method Inheritance Factor (MIF) - Mi / Ma (class)
• Number of Public Methods Defined (PMd) - Number of defined methods that are public (class)
• Number of Public Methods Inherited (PMi) - Number of inherited (but not overridden) methods that are public (class)
• Public Methods Ratio (PMR) (PMd+PMi)/(Md+Mi) (class)
• Number of Hidden Methods Defined (HMd) - Number of defined methods that are non-public (class)
• Number of Hidden Methods Inherited (HMi) - Number of inherited (but not overridden) methods that are non-public (class)
• Method Hiding Factor (MHF) PMd / Md (class)
• Number of Methods Inherited Ratio (NMIR) (Mi / Mit) * 100 (from Lorenz and Kidd, renamed here for distinction from NMI) (class)
• Coupling Factor (CF) - see http://www.cs.kent.edu/~jmaletic/cs63901/lectures/SoftwareMetrics.pdf (note: defining this differntly than the text. text divides by TC^2-TC for the total number of relationships. that is a system level metric, but this metric is as the class level, which means the total number of relationships is 2(TC-1). if a class uses every other class and is used by every other class, that is its value for a CF of 1)(class)
• Polymorphism Factor (PF) - Mo / (Md * NOD) (class)
• Number of Attributes Inherited Total (Ait) - Attrobites inherited overall (class)
• Number of Attributes Inherited (Ai) - Attributes inherited but not overridden (class)
• Number of Attributes Defined (Ad) - Attributes defined within class (same as NMA) (class)
• Number of Attributes Overidden (Ao) - Attributes in class that override an otherwise-inherited attributes (class)
• Number of Attributes (All) (Aa) - Attributes that can be referenced in a class (inherited, overridden, defined). Aa = Ad + Ai (class)
• Attribute Inheritance Factor (AIF) - Ai / Aa (class)
• Number of Public Attributes Defined (Av) - Number of defined attributes that are public (class)
• Number of Hidden Attributes Defined (Ah) - Number of defined attributes that are non-public (class)
• Method Hiding Factor (AHF) Ah / Ad (class)
• McCabe Cyclomatic Complexity (VG) - The number of unique possible paths through code (method)
• Weighed Methods per Class (WMC) - The summation of all of the cyclomatic complexities of all methods on a class (class)
• Lack of Cohesion Methods (LCOM) - A measure for the Cohesiveness of a class. Calculated with the Henderson-Sellers method, based on the number of disjoint sets formed by comparing methods with the attributes they use (class)
• Number of Interfaces (NOI) - The number of abstract classes (and interfaces) in a package (package)
• Afferent Coupling (Ca) - Number of classes outside a package that depend on it (package)
• Efferent Coupling (Ca) - Number of classes inside a package that depend on classes outside of it (package)
• Instability (I) - Effectively the riskiness of a package, how often it has a reason to change, Ce/(Ce+Ca) (package)
• Abstractness (A) - The number of abstract classes (and interfaces) divided by the total number of types in a package, NOI / NOC (package)
• Normalized Distance from Main Sequence (DMS) - Robert Martin's metric for a packages distance from ideal, | A + I - 1 | (package)
• Nested Block Depth (NBD) - The maximum depth of the deepest level of nesting within a method (method)
• Number of Children (NOCh) - Number of classes that directly extend this class (class)
• Number of Parents (NOPa) - Number of classes that this class directly extends (class)
• Number of Descendants (NOD) - Total number of classes that have this class as an ancestor (class)
• Number of Ancestors (NOA) - Total number of classes that have this class as a descendant (class)
• Number of Links (NOL) - Number of links (associations, generalizations, use links) between a class and all others (class)
• Number of Dependants (NODa) - Total number of classes that depend on this class (class)
• Number of Dependencies (NODe) - Total number of classes that a class depends on (same as NOL) (class)
• Class Category Relational Cohesion (CCRC) - The rate of cohesion between a package's classes. Sum(NOL) / NOC (package)
• Number of Comparisons (NCOMP) - Number of comparisons in a method (method)
• Number of Control Variables (NVAR) - Number of control variables referenced in a method (method)
• McClure’s Complexity Metric (MCLC) - NCOP + NVAR (method)
• Fan-out (Fout) - The number of methods immediately subordinate to a method (method)
• Fan-in (Fin) - The number of methods that invoke a method (method)
• Structural Complexity (Si) - Fout^2 (method)
• Input/Output Variables (IOVars) - NOP + 1 (0 if void return type) (method)"
• Data Complexity (Di) - (IOVars)/(Fout+1) (method)
• System Complexity (Ci) - Si + Di (method)
• Class Total System Complexity (ClTCi) - sum(Ci) over all methods in class (class)
• Class Relative System Complexity (ClRCi) - avg(Ci) over all methods in class (class)
• Package Total System Complexity (PkgTCi) - sum(Ci) over all methods in all classes in a package (package)
• Package Relative System Complexity (PkgRCi) - avg(Ci) over all methods in all classes in a package (package)
• Number of Collaborations (CRC) - The number of collaborations between a class and all others (class)
• Data Abstraction Coupling (DAC) - Number of attributes in a class that represent other class types. See Li and Henry
• Size2 (SIZE2) - Number of attributes and the number of local methods defined in a class. Li and Henry
• Data Access Metric (DAM) - Bansiya and Davis
• Direct Class Coupling (DCC) - Bansiya and Davis
• Measure of Aggregation (MOA) - Bansiya and Davis

More metrics coming soon, I plan to gather every metric outlined in the following sources:

• http://www.aivosto.com/project/help/pm-syscomplexity.html system complexity metrics, sums and avgs of Ci per class? package?
• http://support.objecteering.com/objecteering6.1/help/us/metrics/toc.htm
• https://link.springer.com/chapter/10.1007%2F978-3-642-15228-3_7 (debbabi2010metrics)
• http://metrics.sourceforge.net/
• http://www.cs.kent.edu/~jmaletic/cs63901/lectures/SoftwareMetrics.pdf
• http://www.alibris.com/Object-Oriented-Metrics-Measures-of-Complexity-Brian-Henderson-Sellers/book/29695100
• http://www.objectmentor.com/resources/articles/oodmetrc.pdf
• http://www.win.tue.nl/~aserebre/2IS55/2009-2010/10.pdf
• Halstead measures (1977)

And a lot more

MOOD metrics note: MOOD are system level metrics. where possible these are calculated for lower entities such as classes