scc v3.7.0
A high-performance source-code counter that reports code, comments, blanks, complexity, COCOMO estimates, unique lines and several report formats.
What scc measures
scc (Sloc Cloc and Code) reports code, comments, blanks and physical lines by language. It can also expose complexity, COCOMO estimates and unique lines, with JSON, CSV, SQL, HTML and other structured outputs for repository inventories and trend reports.
There is no single counting convention
Language detection, comment boundaries, duplicate files and generated content rules can differ between scc, cloc and tokei. Fix the version, target commit, exclusions and parameters before comparing totals; never combine results from different scopes without explaining the difference.
Lines are engineering signals, not quality scores
Generated code, tests, documentation, architecture and refactoring all change line counts. Complexity and cost models are supporting signals only. A useful report states its scope, exceptions and business context while protecting internal paths and project-size details.
Archive note
This page covers the v3.7.0 Windows x64 ZIP package. Confirm the archive name, size and checksum shown by the current catalog before using it.
Save to your cloud drive
Open the cloud drive to get the file directly, or save it for convenient access on another device.
Quark Cloud Drive
RecommendedSave scc v3.7.0 to this cloud drive
Baidu Netdisk
Save scc v3.7.0 to this cloud drive
scc Windows code statistics workflow
Start with a small test repository, define exclusions for dependencies and generated output, then save the version and parameters with every report.
Before you start
- Choose a repository with a known layout and list dependency, cache, generated and third-party directories.
- Run as a regular user and avoid starting from a disk root, home directory or unrestricted network share.
- Define whether the goal is inventory, trend analysis or a CI report; do not treat line count as individual productivity.
Installation steps
- 01
Extract the Windows binary
Unpack the archive into a fixed tools directory, verify the version and run scc --help through its full path.
- 02
Define the scope
Select a test repository and confirm that dependency, mount, binary and secret-containing directories are outside the scan.
- 03
Record the baseline
Save the tool version, target path, exclusions and output format so future reports can be compared on the same basis.
Quick start
- 01
Run the default report
Run scc at the repository root and check languages, files, code, comments and blanks against the project structure.
- 02
Exclude noise deliberately
Use the current help to exclude dependencies, build products, compressed content and generated files, then record the reason for each rule.
- 03
Export structured data
Choose a supported JSON or CSV output, inspect fields and encoding, and send the new file to an analysis or CI step.
Usage tips
- Compare the same scc version, configuration and repository state when measuring trends.
- Report generated code, copied files and third-party dependencies separately when they matter.
- HTML, JSON and CSV reports may expose paths and project scale; review them before sharing.
Troubleshooting and uninstall
Why is the scan slow or unusually large?
Check for dependencies, build directories, binaries, network volumes and symbolic links; test exclusions on a smaller path before expanding the scope.
Why was a file assigned to the wrong language?
Check the extension, shebang and current language definitions, then use a documented mapping or pattern override and record the custom rule.
- Remove the tool and sensitive reportsRemove its directory from PATH, delete the binary when appropriate and archive or clean reports according to the project's confidentiality policy.
Frequently asked questions
Why can scc and cloc show different totals?
They may differ in language detection, comment boundaries, blank-line handling, duplicate-file rules and exclusions; compare identical scope and parameters.
Can complexity or COCOMO be used as a performance score?
They are engineering signals influenced by language, generated code and architecture, so they should not be used alone to judge quality or individual contribution.
Does scc execute the source it scans?
Normal counting reads and classifies files rather than executing project code, but the scan should still be limited to trusted paths and safe report destinations.
Which output formats are useful for automation?
Use a structured format supported by the installed version, such as JSON or CSV, and save the version, exclusions and commit together with the report.