escomplex

escomplex

Software complexity analysis of JavaScript-family abstract syntax trees. The back-end for complexity-report.

• Metrics
• Links to research
• Installation
• Usage
  • Arguments
    • ast
    • walker
    • options
  • Result
    • For a single module
    • For multiple modules
• Related projects
• Development
• License

Metrics

Currently the library reports on:

• Lines of code: Both physical (the number of lines in a module or function) and logical (a count of the imperative statements). A crude measure.
• Number of parameters: Analysed statically from the function signature, so no accounting is made for functions that rely on the arguments object. Lower is better.
• Cyclomatic complexity: Defined by Thomas J. McCabe in 1976, this is a count of the number of cycles in the program flow control graph. Effectively the number of distinct paths through a block of code. Lower is better.
• Cyclomatic complexity density: Proposed as a modification to cyclomatic complexity by Geoffrey K. Gill and Chris F. Kemerer in 1991, this metric simply re-expresses it as a percentage of the logical lines of code. Lower is better.
• Halstead metrics: Defined by Maurice Halstead in 1977, these metrics are calculated from the numbers of operators and operands in each function. Lower is better.
• Maintainability index: Defined by Paul Oman & Jack Hagemeister in 1991, this is a logarithmic scale from negative infinity to 171, calculated from the logical lines of code, the cyclomatix complexity and the Halstead effort. Higher is better.
• Dependencies: A count of the calls to CommonJS and AMD require. Analysed statically from the function signature, so no accounting is made for dynamic calls where a variable or function is obscuring the nature of the dependency. Lower is better.
• First-order density: The percentage of all possible internal dependencies that are actually realised in the project. Lower is better.
• Change cost: The percentage of modules affected, on average, when one module in the project is changed. Lower is better.
• Core size: The percentage of modules that are both widely depended on and themselves depend on other modules. Lower is better.

It is important to note that none of these metrics can compete with the insight of a competent developer. At best, they are an automatable warning system, which can help to identify areas of code that warrant closer inspection by a human being.

Links to research

• A Complexity Measure, by Thomas J McCabe.
• Cyclomatic Complexity Density and Software Maintenance Productivity, by Geoffrey K. Gill and Chris F. Kemerer.
• Resolving the Mysteries of the Halstead Measures, by Horst Zuse.
• Exploring the Structure of Complex Software Designs: An Empirical Study of Open Source and Proprietary Code, by Alan MacCormack, John Rusnak and Carliss Baldwin.
• The Impact of Software Design Structure on Product Maintenance Costs and Measurement of Economic Benefits of Product Redesign, by Andrei Akaikine.
• A Systematic Review of Software Maintainability Prediction and Metrics, by Mehwish Riaz, Emilia Mendes and Ewan Tempero.

Installation

The library is published on npm under the name escomplex. To install, you can add it to the dependencies in your package.json file or simply run:

npm i escomplex --save

Usage

You can load escomplex in your own code by calling require:

const escomplex = require('escomplex');

escomplex exports two primary functions, analyze and processResults

analyse

const result = escomplex.analyse(source, options);

Arguments

ast

The first argument, source, must be either a string or an array of objects. If it is an array, each object should include a path property that is either a relative or full path to the equivalent module on disk and a code with the contents of the module. As well as identifying each of the result objects, the path property is also used during dependency analysis.

options

The third argument, options, is an optional object containing properties that modify some of the complexity calculations:

• options.logicalor: Boolean indicating whether operator || should be considered a source of cyclomatic complexity, defaults to true.
• options.switchcase: Boolean indicating whether switch statements should be considered a source of cyclomatic complexity, defaults to true.
• options.forin: Boolean indicating whether for...in loops should be considered a source of cyclomatic complexity, defaults to false.
• options.trycatch: Boolean indicating whether catch clauses should be considered a source of cyclomatic complexity, defaults to false.
• options.newmi: Boolean indicating whether the maintainability index should be rebased on a scale from 0 to 100, defaults to false.
• options.skipCalculation: only valid for when source is an array of files Boolean indicating if we should skip processing of certain values, such as the adjacency and visibility matrixes, core sizes, and average values loc, etc.
• options.noCoreSize: Skips creating the visibility matrix and calculating the coreSize, which can be very expensive for large projects

processResults

const noCoreSize = false;
escomplex.processResults(result, noCoreSize);

This function takes a report object and computes aggregate scores for all individual files and also adjacency and visibility matrices. This is useful for combining together multiple report objects and recomputing aggregate scores.

Arguments

Result

A result object of the form:

var result = {
  reports: [
    {
      // same format as module return
    }
  ]
}

noCoreSize

a boolean indicating not to calculate the visibilityMatrix or core size

Result Format

Both analyze and processResults return a report of the following format, with some variation depending on the given options.

For a single module

If a single abstract syntax tree object is passed in the ast argument, the result will be a report object that looks like the following:

{
    maintainability: 171,
    dependencies: [],
    aggregate: {
        sloc: {
            logical: 0,
            physical: 0
        },
        params: 0,
        cyclomatic: 1,
        cyclomaticDensity: 1,
        halstead: {
            vocabulary: 0,
            difficulty: 0,
            volume: 0,
            effort: 0,
            bugs: 0,
            time: 0
        }
    },
    functions: [
        {
            name: '',
            line: 0,
            sloc: {
                logical: 0,
                physical: 0
            },
            params: 0,
            cyclomatic: 1,
            cyclomaticDensity: 1,
            halstead: {
                vocabulary: 0,
                difficulty: 0,
                volume: 0,
                effort: 0,
                bugs: 0,
                time: 0
            }
        },
        ...
    ]
}

The meaning of those values, briefly, is as follows (see metrics for more information on each one):

• report.maintainability: The maintainability index for the module.
• report.dependencies: The array of CommonJS/AMD dependencies for the module.
• report.aggregate.sloc.physical: Physical lines of code for the module. Will be undefined if the syntax tree is not annotated with line number data.
• report.aggregate.sloc.logical: Logical lines of code for the module.
• report.aggregate.params: Parameter count for the module.
• report.aggregate.cyclomatic: Cyclomatic complexity for the module.
• report.aggregate.cyclomaticDensity: Cyclomatic complexity density for the module.
• report.aggregate.halstead.vocabulary: Halstead vocabulary size for the module.
• report.aggregate.halstead.difficulty: Halstead difficulty for the module.
• report.aggregate.halstead.volume: Halstead volume for the module.
• report.aggregate.halstead.effort: Halstead effort for the module.
• report.aggregate.halstead.bugs: Halstead bugs for the module.
• report.aggregate.halstead.time: Halstead time for the module.
• report.functions[n].name: Function name.
• report.functions[n].line: Line number that the function starts on. Will be undefined if the syntax tree is not annotated with line number data.
• report.functions[n].sloc.physical: Physical lines of code for the function. Will be undefined if the syntax tree is not annotated with line number data.
• report.functions[n].sloc.logical: Logical lines of code for the function.
• report.functions[n].params: Parameter count for the function.
• report.functions[n].cyclomatic: Cyclomatic complexity for the function.
• report.functions[n].cyclomaticDensity: Cyclomatic complexity density for the function.
• report.functions[n].halstead.vocabulary: Halstead vocabulary size for the function.
• report.functions[n].halstead.difficulty: Halstead difficulty for the function.
• report.functions[n].halstead.volume: Halstead volume for the function.
• report.functions[n].halstead.effort: Halstead effort for the function.
• report.functions[n].halstead.bugs: Halstead bugs for the function.
• report.functions[n].halstead.time: Halstead time for the function.

For multiple modules

If an array of sources is passed in the source argument, the result will be an object that looks like the following:

{
    reports: [
        ...
    ],
    adjacencyMatrix: [
        [ 0 ]
    ],
    firstOrderDensity: 0,
    visibilityMatrix: [
        [ 0 ]
    ],
    changeCost: 100,
    coreSize: 100,
    loc: 0,
    cyclomatic: 1,
    effort: 0,
    params: 0,
    maintainability: 171
}

Those properties are defined as follows:

• result.reports: An array of report objects, each one in the same format described above but with an extra property path that matches the path property from its corresponding syntax tree. This path property is required because the reports array gets sorted during dependency analysis.
• result.adjacencyMatrix: The adjacency design structure matrix (DSM) for the project. This is a two-dimensional array, each dimension with the same order and length as the reports array. Each row and column represents its equivalent indexed module from the reports array, with values along the horizontal being 1 when that module directly depends on another and values along the vertical being 1 when that module is directly depended on by another. All other values are 0.
• result.firstOrderDensity: The first-order density for the project.
• result.visibilityMatrix: The visibility DSM for the project. Like the adjacency matrix, but expanded to incorporate indirect dependencies. Will be missing if noCoreSize is passed as an option.
• result.changeCost: The change cost for the project. Will be missing if noCoreSize is passed as an option.
• result.coreSize: The core size for the project.
• result.loc: The average per-function count of logical lines of code.
• result.cyclomatic: The average per-function cyclomatic complexity.
• result.effort: The average per-function Halstead effort.
• result.params: The average per-function parameter count.
• result.maintainability: The average per-module maintainability index.

Related projects

• plato: JavaScript source code visualization, static analysis, and complexity tool.
• jsc: JavaScript source code complexity tool.
• bob: Minimalist-omakase build tool for node.js projects.
• cardio: A web application health tool.
• grunt-complexity: A JavaScript complexity analysis grunt task.
• brackets-crjs: Brackets extension.
• jscomplexity: JS cyclomatic complexity report generator.
• karma-complexity-processor: A preprocessor for karma runner to give some metrics about code complexity.
• crlint: JS linter based on complexity report results.

Development

Refer to the contribution guidelines before submitting a pull request.

Source code is in /src. Unit tests are in /test. You can run the tests with npm test. You can run the linter with npm run lint. Make sure you've installed all the dependencies with npm install first.

License

MIT