Code Quality Report (Complexity & Duplicate Code)

About the Code Quality Report (Complexity & Duplicate Code)

Choose a project folder, drop a ZIP or source files, or paste code, and get a report on the code in seconds: how many lines of code, comment and blank lines it has in each language and folder, how complex each function is, where code is copied from one place to another, and which files are the hotspots to look at first. Everything is read and measured in a background worker in your browser; nothing is uploaded.

Lines are counted the way cloc counts them. For JavaScript, TypeScript, Python, Java, C#, Go, PHP, Ruby, C, C++ and Rust every function and method is measured: cyclomatic complexity (counted as lizard counts it), Cognitive Complexity (as G. Ann Campbell’s white paper defines it: breaks in the flow, plus nesting), code lines, parameters and the deepest nesting. Copied blocks of 50 tokens or more are found across all files, as jscpd finds them, and each file gets a maintainability index from 0 to 100. Files that Git ignores (.gitignore at every level), dependency and build folders, lock files, minified files and generated code are left out, and listed.

Without a pass the report is a free preview: the summary counts and the totals by language in full, and the first rows of the hotspot table. With a Pro pass every function, every duplicate block, the folders and the files show, and the HTML and PDF report, the CSV tables, the Code Quality JSON for merge requests and the Markdown summary unlock.

Common uses

• Getting to know a codebase you inherit: its languages, its size and its most complex functions.
• Checking a pull request’s files before review, or finding the functions to refactor first.
• Adding a code quality report to a merge request from a CI job’s files, without a server-side scanner.
• Tracking lines of code, duplication and complexity of a project over time with the CSV tables.

How the measures are counted

• Lines: a line with only white space is blank (inside a comment too), a line with any code is code — a trailing comment does not change that — and the rest are comment lines. Strings are code, so "http://example.com" is not a comment. Python docstrings count as comments, as in cloc.
• Cyclomatic complexity: 1 for the function, plus 1 for each if, for, while, case, catch, &&, || and ternary ?, with each language’s own words (elif, except, and, or in Python; elsif, unless, until, when, rescue in Ruby; each match arm and the ? operator in Rust).
• Cognitive Complexity: +1 for each if, else if, else, ternary, switch, loop and catch, for each sequence of like boolean operators, for a jump to a label and for recursion, and +1 more for each level of nesting an if, ternary, switch, loop or catch sits at. Lambdas and nested functions add a level of nesting and count in the function that holds them; try, finally and null-coalescing operators add nothing.
• Nesting: the deepest nest of if, loop, switch and try blocks in a function. Maintainability index: max(0, (171 − 5.2 ln V − 0.23 G − 16.2 ln L) × 100 / 171), with V the Halstead volume, G the cyclomatic complexity and L the lines of code of the file; 20 and over is good, 10–19 moderate, under 10 low.

What is left out

• Files Git ignores: every .gitignore in the project, at every level, as Git reads them.
• Vendored paths: dependency folders (node_modules, vendor, bower_components), build output (dist, cache), third-party libraries (jQuery, Bootstrap …), tool folders and wrappers (.vscode, .github, gradlew), and declaration files (.d.ts).
• Generated files: lock files (package-lock.json, yarn.lock, Cargo.lock, poetry.lock …), minified JavaScript and CSS (lines over 110 characters on average), source maps, and files that say they are generated (Code generated … DO NOT EDIT, protocol buffers, designer files).
• Binary files, files over 2 MB, and files that are not code (images, documents).
• Untick Leave out vendored and generated files or Follow .gitignore files to count them anyway.

The exports

• HTML report: one file with every table, readable in light and dark, to share or keep with a release.
• PDF: the summary, the hotspots, the most complex functions and the largest duplicates, on A4 landscape pages.
• CSV tables: files, functions, duplicates and languages, one row each, for a spreadsheet.
• Code Quality JSON (GitLab): the findings in the Code Climate format a merge request’s code quality widget reads: a JSON array of issues with a description, check name, severity, a stable fingerprint and the relative path and line, without a byte-order mark. Save it as gl-code-quality-report.json in a CI job and declare it as the job’s code quality report.
• Copy summary: the totals and the top hotspots as Markdown, to paste into a pull request or a ticket.

Limitations

• Functions are found from the code’s tokens, not compiled: unusual macros, code generated at run time and very unusual syntax can be measured a little differently from a compiler-based tool.
• Up to 50,000 files and 300 MB of code; a file over 2 MB is left out. Duplicates are searched in the first 3 million tokens of a very large project.
• Recursion counts when a function calls itself directly; calls through other functions are not followed.
• A ZIP must use the usual compression (deflate); a dropped folder skips .git and node_modules without reading them.

Privacy

On this device: nothing is sent

Everything happens in your browser. What you enter or open here is not uploaded or stored by MySmartCoPilot.