David Wheeler wrote the following about sloccount. This is excerpted from a much longer TODO with many issues that have been solved by the Go port.

TODO items specific to the Go implementation start here:

Line-counter feature comparison

Surveyed 2026-11-04.

Implemented since this survey: --no-duplicates provides optional file content deduplication.

Loccount’s largest feature gaps are richer measurements and reports, comparisons between revisions, duplicate suppression, and accounting for embedded languages.

This survey covers the seven counters integrated into tablegen.py, plus Pygount from earlier discussion and SLOCCount. It compares their upstream documentation and source with loccount’s current implementation; it is not a performance benchmark. SCC explicitly warns that its documentation follows master, so some features below may require a newer or unreleased build.

Loccount already has parallel processing, generated-file filtering, ignore-file handling, content-based language recognition, per-file text output, summary JSON, LLOC, and COCOMO estimates. Its statistics record and reporting code retain SLOC and LLOC but no separate comment, blank, or total-line counts. Reporting also lacks per-file JSON; the HTML options produce language documentation.

Counter Capabilities loccount lacks

SCC

Branch-based and cognitive complexity estimates; unique-line and textual repetition statistics; duplicate-file suppression; byte counts and maximum/mean line lengths. CSV, SQL, OpenMetrics, and HTML measurement reports, including a self-contained graphical report. Runtime language remapping and project configuration files. Configurable COCOMO parameters and an experimental LLM recreation-cost estimator. Git-history analysis covers complexity/churn hotspots, files that change together, author attribution, bus-factor estimates, and timelines of activity and language composition. Documentation

cloc

Detailed comparisons between directories, archives, or Git revisions, reporting added, removed, modified, and unchanged lines. Counts historical Git revisions without checking them out. Reads compressed archives, source from stdin, and lists of input files. Supports runtime language definitions and forced language assignments, content deduplication, report aggregation, and CSV/Markdown/XML/YAML/SQL exports. Can write source copies with comment-only and blank lines removed. Documentation

Tokei

Attributes embedded code to its own language — for example, JavaScript/CSS inside HTML and fenced code inside Markdown. Supports YAML and CBOR exports, and merges previously saved statistics into another report. Has configuration files, language filters, selectable sorting, controls for hidden files and ignore rules, and a reusable Rust library. Documentation

Gocloc

Duplicate-file suppression; language inclusion filters; separate filename and directory matching rules; selectable sorting. Produces Markdown, cloc-compatible XML, SLOCCount-compatible output, and per-file JSON. Also exposes its counting machinery as a Go library. CLI source

Ocloc

Git comparisons between revisions, against a merge base, for staged changes, or for unstaged changes, with rename detection. Can fail a CI job when added-code, changed-line, or changed-file limits are exceeded, including per-language limits. Provides CSV and Markdown comparison reports, a progress display, configurable worker count, file-size limits, and optional symlink following. Documentation

Ohcount

Counts multiple languages within one file. Can emit source annotated line by line with language and code/comment/blank classification, or emit classified entity spans with offsets. Detects software-license notices in source files. Provides a C library and Ruby bindings. CLI source, project documentation

Polyglot

Produces HTML measurement reports, with optional CSS, and supplies colored terminal output and a Vim integration. Its additional feature set is comparatively small. Manual, project documentation

Pygount

Detects and decodes source encodings using BOMs, encoding declarations, fallbacks, and optional encoding detection. Can clone and count remote Git repositories, including selected revisions. Suppresses duplicate files; accepts configurable generated-file patterns; reports why files were skipped, including binary, duplicate, generated, empty, unknown, and error classifications. Offers cloc-compatible XML, detailed JSON, and a Python API. Usage, API

SLOCCount

Saves analysis results for later reporting without recounting. Groups results by component and supports multiple-project estimation, calculating each project’s effort separately. Suppresses duplicate files, with a separate policy for duplicates across projects. Allows adjustment of salary, overhead, effort, and schedule parameters, inclusion of generated files, and following symlinks. User’s guide

There are several distinctions worth preserving when turning this into a development list:

  • Comment and blank-line reporting is the most widespread omission. SCC, cloc, Tokei, Gocloc, Ocloc, Ohcount, and Polyglot expose those counts separately. Adding them would also enable comment percentages and more useful machine-readable reports.

  • Version-control file selection and revision analysis are separate capabilities. Loccount uses version-control metadata to select working-tree files. cloc and Ocloc compare historical contents; SCC additionally analyzes activity across history.

  • Duplicate suppression and repetition statistics measure different things. Content deduplication prevents identical files from inflating totals. SCC’s unique-line statistics measure textual repetition within the counted material; they are not semantic clone detection.

  • Embedded-language attribution requires counter changes. Loccount generally returns one language record per file. Matching Tokei or Ohcount here would require splitting the measurements among languages, beyond adding YAML entries.

  • Pygount deliberately uses a different counting policy. It can exclude punctuation-only and certain no-operation lines, and separately records pure-string lines. That is an alternative metric rather than an accuracy improvement to ordinary SLOC. Counting policy

The recommended priorities are comment/blank/total counts with per-file JSON and CSV, followed by runtime language overrides and inclusion controls, optional content deduplication, and revision comparisons. Embedded-language accounting is the larger parser project. Complexity reporting has a useful starting point already: loccount’s generated language records contain Co token lists, but the counter currently does not consume them.