Measured · not claimed

How much of it actually runs.

Every card below comes from a measurement against that language's own authority — an official compiler test suite run unmodified where one exists, and a curated ledger arbitrated by an outside implementation where it does not. No number on this page is written by hand.

official suite · unit: file
Go
21.5%434 of 2,016 in-scope files
curated ledger · unit: ledger row
QBasic
96.6%476 of 493 in-scope ledger rows · 6 wrong
official suite · unit: assert
Lua
56.8%1,422 of 2,505 in-scope asserts · 1 wrong

passing
not yet supported
wrong — a real defect
out of scope / no arbiter

These cards sit side by side, but they are not the same scale. The unit changes from language to language — Go: files, QBasic: ledger rows, Lua: asserts. Counting a file and counting every assert inside that file are not the same question: a file unit rounds everything that works in a test that stops early down to zero, while an assert unit is finer but reads higher. The authority changes too: an official suite is written by the language's own team, while a curated ledger is a set of tests we chose — its percentage says how much of the chosen surface still stands, not how much of the language is covered. Read a percentage as one language's movement over time, not as a ranking between languages.

not measured

Folders with no measurement yet

These are listed so the picture does not mislead: a language folder may carry code, but with no corpus measurement its coverage is simply unknown.

has code, no corpus measurement 2

Dart
Kotlin

surface matrix only 25

AWK
Ada
C
CPlusPlus
Clojure
Erlang
Forth
Haskell
Java
Julia
OCaml
ObjectiveC
PHP
Perl
Prolog
Python
R
Ruby
Rust
SQL
Scheme
Shell
Smalltalk
Tcl
Zig