By the numbers

Data collected on 2026-04-30 from apple/swift main at commit 8123d28cd2.

Size

The compiler and runtime are predominantly C++ and Swift. The standard library and the test suite are overwhelmingly Swift, while the toolchain core is C++ with sizable Python tooling.

xychart-beta horizontal
    title "Lines of code by language"
    x-axis ["C++", "C/C++ headers", "Swift (stdlib + tests)", "SIL test fixtures", "Python (build/test)"]
    y-axis "Lines (thousands)" 0 --> 1300
    bar [1186, 543, 72, 338, 46]

Metric	Count
Tracked files	30,923
.cpp files	1,373 (~1.19M lines)
.h files	2,565 (~543K lines)
.swift files	22,031 (~72K lines, dwarfed by tests in raw count)
.sil test fixtures	1,303 (~338K lines)
.py files	263 (~46K lines)
.gyb templates	282
Markdown / RST docs	336 + 58
Tests under test/	~16,489 files
Tests under validation-test/	~5,600 files
Benchmarks	263 files in benchmark/

The single largest source files are interesting -- they hint at where the compiler's complexity is concentrated:

File	Lines
stdlib/public/stubs/Unicode/Common/ScalarPropsData.h (generated)	26,335
stdlib/public/stubs/Unicode/Apple/ScalarPropsData.h (generated)	26,305
lib/Sema/CSSimplify.cpp	16,668
lib/AST/Decl.cpp	13,602
include/swift/SIL/SILInstruction.h	12,423
lib/ClangImporter/ImportDecl.cpp	11,222
lib/Parse/ParseDecl.cpp	10,976
include/swift/AST/Decl.h	10,399
lib/Sema/CSApply.cpp	9,895
lib/Sema/CSDiagnostics.cpp	9,800

Sema (lib/Sema/) dominates because the constraint-based type checker handles overload resolution, generics, conversions, and diagnostics in one place. AST also shows up because Decl.cpp and Decl.h host the bulk of declaration-node logic.

Activity

The repo has been continuously active since 2010-07-17 and has crossed 199,000 commits. It averages well over 10,000 commits per year:

Year	Commits
2014	10,286
2016	14,989
2018	18,998
2020	15,696
2022	15,865
2024	17,774
2025	15,339

In the last 90 days, 2,780 commits landed.

Hotspots (last 12 months)

The 12 directories with the most file-modifying commits in the past year:

Directory	Commits
include/swift/	3,233
lib/Sema/	2,322
stdlib/public/	2,013
test/Interop/	1,690
lib/AST/	1,410
test/SILOptimizer/	1,103
test/Concurrency/	1,081
SwiftCompilerSources/Sources/	959
lib/SILOptimizer/	925
lib/ClangImporter/	775
lib/IRGen/	731
lib/SILGen/	599

Sema, AST, and the SIL optimizer dominate -- Swift's type system and SIL pipeline are still evolving features (concurrency model, ownership, regex, embedded). Heavy test/Interop/ activity reflects the ongoing C++ interop and typed-throws work. The growing SwiftCompilerSources/Sources/ count tracks the ongoing port of optimizer passes from C++ to Swift.

Bot-attributed commits

Searching the Co-authored-by: trailer for the substring [bot] returns 0 commits in the entire history. The Swift project does not currently use commit-co-authoring bots; CI (@swift_ci) operates by commenting on PRs rather than pushing co-authored commits. AI-assisted work (via Copilot or similar inline tools) is invisible to git history; this number is therefore a firm 0 lower bound but not an upper bound.

Complexity

The compiler is a "fat" codebase: a few directories contribute the bulk of the lines. C++ implementation lives in lib/ (1,413 files), declarations in include/swift/ (991 files), the standard library in stdlib/ (1,091 files), and the test suite in test/ (16,489 files).

Three numbers worth remembering:

• The test/ directory has more files (16,489) than the rest of the repository's source-bearing directories combined.
• lib/SILOptimizer/ (288 files) is now the single largest lib/ subdirectory, narrowly ahead of lib/IRGen/ (172) and lib/AST/ (172).
• The SwiftCompilerSources/ directory has 185 source files spread across AST, Basic, SIL, and Optimizer -- a relatively new component but already substantial.

TODO inventory

Across lib/, include/, and stdlib/ there are roughly 1,041 files containing a TODO, FIXME, HACK, or XXX comment. That's a long tail of bookmarked-but-not-yet-fixed work consistent with a 15+ year-old compiler. See Lore for selected long-running TODOs.