Why pure Go¶
go-ruby-scanf/scanf reimplements Ruby's scanf in pure Go, with cgo disabled. The
slice of Ruby it covers is deterministic and interpreter-independent: scanning
an input string against a scanf format grammar — the directives, widths,
character-class sets, suppression flag, and literal/whitespace matching — is a
pure function of its inputs, with no live binding and no evaluation of arbitrary
Ruby. That is exactly the part that can — and should — live as a standalone Go
library, separate from the interpreter.
Extracted from rbgo, reusable by anyone¶
This library began life inside go-embedded-ruby's
rbgo. It has been made a reusable standalone library so that:
- any Go program can import
github.com/go-ruby-scanf/scanfdirectly, with no Ruby runtime; - the dependency runs the other way —
rbgobinds this module as a native module (the same pattern as go-ruby-regexp and go-ruby-erb), rather than this module depending on the interpreter; - the behaviour is pinned by a differential oracle against the system
ruby(String#scanf), independent of any one consumer.
It hands back a small, explicit value model (int / *big.Int / float64 /
string, the natural mapping of Ruby's Integer / Float / String) that the
host binds to its own objects.
Why pure Go matters here¶
Because the library is CGO-free and dependency-free, it:
- cross-compiles to every Go target with no C toolchain, and links into a single static binary;
- has no dependency on the Ruby runtime — the dependency runs the other way;
- can be differentially tested against the
rubybinary wherever one is onPATH, while the cross-arch lanes (whererubyis absent) still validate the library itself.
See Usage & API for the surface and Roadmap for what is in scope.