Awesome
escomplex
:exclamation: :anguished: :exclamation: :anguished: :exclamation: :anguished: :exclamation: :anguished: :exclamation: :anguished: :exclamation: :anguished: :exclamation: :anguished: :exclamation: :anguished: :exclamation: :anguished: :exclamation: :anguished: :exclamation:
NOTE: This fork is no longer maintained. Use Jared's fork instead.
:exclamation: :anguished: :exclamation: :anguished: :exclamation: :anguished: :exclamation: :anguished: :exclamation: :anguished: :exclamation: :anguished: :exclamation: :anguished: :exclamation: :anguished: :exclamation: :anguished: :exclamation: :anguished: :exclamation:
Software complexity analysis of JavaScript-family abstract syntax trees. The back-end for complexity-report.
- Abstract syntax trees
- Syntax tree walkers
- Metrics
- Links to research
- Installation
- Usage
- Related projects
- Development
- License
Abstract syntax trees
This library deliberately excludes logic for parsing source code and for navigating parse trees. Both the syntax tree and a matching syntax tree walker are inputs to escomplex, meaning it is not tied to any particular language, parser or data format.
Syntax tree walkers
- escomplex-ast-moz: Walks syntax trees that conform to the format defined in Mozilla's Parser API. This format is returned by Esprima and Acorn, two popular JavaScript parsers.
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 install escomplex
Usage
You can load escomplex
in your own code
by calling require
:
var escomplex = require('escomplex');
escomplex exports two primary functions,
analyze
and processResults
analyze
var result = escomplex.analyse(ast, walker, options);
Arguments
ast
The first argument, ast
,
must be either
an abstract syntax tree
or an array of syntax trees.
If it is an array,
each tree should include
an extra property, path
,
that is either a relative or full path
to the equivalent module on disk.
As well as identifying
each of the result objects,
that path is also used
during dependency analysis.
walker
The second argument, walker
,
must be a syntax tree walker.
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 totrue
.options.switchcase
: Boolean indicating whetherswitch
statements should be considered a source of cyclomatic complexity, defaults totrue
.options.forin
: Boolean indicating whetherfor
...in
loops should be considered a source of cyclomatic complexity, defaults tofalse
.options.trycatch
: Boolean indicating whethercatch
clauses should be considered a source of cyclomatic complexity, defaults tofalse
.options.newmi
: Boolean indicating whether the maintainability index should be rebased on a scale from 0 to 100, defaults tofalse
.options.skipCalculation
: only valid for when ast 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
escomplex.processResults(result, false);
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 (say from different languages) 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 beundefined
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 beundefined
if the syntax tree is not annotated with line number data.report.functions[n].sloc.physical
: Physical lines of code for the function. Will beundefined
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 syntax trees
is passed in the ast
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 propertypath
that matches thepath
property from its corresponding syntax tree. Thispath
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 thereports
array. Each row and column represents its equivalent indexed module from thereports
array, with values along the horizontal being1
when that module directly depends on another and values along the vertical being1
when that module is directly depended on by another. All other values are0
.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 ifnoCoreSize
is passed as an option.result.changeCost
: The change cost for the project. Will be missing ifnoCoreSize
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 contrubution 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.