Package 'tccquickr'

Format a tccq type

Description

Format a tccq type

Usage

## S3 method for class 'tccq_type'
format(x, ...)

Arguments

x	A tccq_type object.
...	Unused.

Value

A compact character representation.

---

Print a tccq module summary

Description

Print a tccq module summary

Usage

## S3 method for class 'tccq_module'
print(x, ...)

Arguments

x	A tccq_module object.
...	Unused.

Value

x, invisibly.

---

Print a tccq type

Description

Print a tccq type

Usage

## S3 method for class 'tccq_type'
print(x, ...)

Arguments

x	A tccq_type object.
...	Unused.

Value

x, invisibly.

---

Analyze an R function for typed-compilation reconnaissance

Description

tccq_analyze() collects high-level AST facts and optional compiler-package observations for a function. This is intentionally a side-band analysis to improve compiler diagnostics and help with boundary-reachability, while keeping the typed-IR path as the optimization frontier.

Usage

tccq_analyze(fn)

Arguments

fn	R function.

Details

The output is intentionally conservative and mostly advisory.

Value

A list of class tccq_analysis with fields:

• formals: formals metadata

• ast: AST observations (calls, symbols, assignments, loops)

• compiler: compiler package metadata when available

• recommendations: conservative suggestions for lowering success

See Also

tccq_compile(), tccq_jit()

Examples

f <- function(x, y = 1) {
  declare(type(x = double(NA), y = double(NA)))
  sum((x + y) * y)
}
tccq_analyze(f)

---

tccq backend factories

Description

Backend objects control how emitted C is handled by tccq_compile().

Usage

tccq_backend_source()

tccq_backend_tinycc()

tccq_backend_shlib()

Details

• tccq_backend_source() returns generated C source without compiling it.

• tccq_backend_tinycc() compiles and loads the emitted C in memory through Rtinycc.

• tccq_backend_shlib() compiles the emitted C as a shared library through ⁠R CMD SHLIB⁠ and loads it with dyn.load(). This lives in the same general deployment space as callme.

Backends declare capabilities internally, and tccq_compile() validates the selected backend against the current target, compile context, and explicit boundary APIs before compiling.

Value

A tccq_backend object suitable for the ⁠backend =⁠ argument of tccq_compile().

Examples

tccq_backend_source()$name
tccq_backend_tinycc()$name
tccq_backend_shlib()$name

---

Experimental tccq R-to-C compiler entry point

Description

This is the package's current compiler path. The current milestone supports declared scalar and vector arithmetic, generic fold-style reducers, local bindings, indexed reads, contiguous slices, and local indexed/range writes.

Usage

tccq_compile(
  fn,
  mode = c("compile", "code", "ir"),
  fallback = c("hard", "auto"),
  backend = tccq_backend_tinycc(),
  target = tccq_target_c_rapi(),
  extlibs = list(),
  debug = FALSE
)

Arguments

fn	R function with a leading declare(type(...)) annotation.
mode	One of "compile", "code", or "ir".
fallback	One of "hard" or "auto". In "hard" mode unsupported calls are rejected. In "auto" mode unsupported calls may lower to explicit r_eval boundary nodes.
backend	Backend object. Use tccq_backend_source(), tccq_backend_tinycc(), or tccq_backend_shlib(). Defaults to TinyCC via Rtinycc.
target	Target object. Defaults to C + R C API emission.
extlibs	Optional list of tccq_external_library descriptors.
debug	Print generated C source before compiling.

Details

Current statement semantics are intentionally strict: rebinding a local name is rejected, direct mutation of formal arguments is rejected, and indexed or range assignment currently requires a scalar right-hand side.

Value

A compiled callable, C source string, or module IR depending on mode.

---

Runtime-specialized JIT dispatcher for declared-typed tccq functions

Description

tccq_jit() wraps a declared tccq function with a per-call dispatch cache keyed by observed argument signatures. Each distinct signature is compiled at most once (optionally with exact shape guards) and then reused on subsequent calls.

Usage

tccq_jit(
  fn,
  fallback = c("hard", "auto"),
  backend = tccq_backend_tinycc(),
  target = tccq_target_c_rapi(),
  extlibs = list(),
  debug = FALSE,
  exact = TRUE
)

Arguments

fn	R function with a leading declare(type(...)) annotation.
fallback	One of "hard" or "auto". In "hard" mode unsupported calls are rejected. In "auto" mode unsupported calls may lower to explicit r_eval boundary nodes.
backend	Backend object. Use tccq_backend_source(), tccq_backend_tinycc(), or tccq_backend_shlib().
target	Target object. Defaults to C + R C API emission.
extlibs	Optional list of tccq_external_library() descriptors.
debug	Print generated C source before each (first-time) signature compilation.
exact	If TRUE, each signature is specialized on exact observed shape (length for rank-1 and nrow/ncol for rank-2) and corresponding argument checks are emitted into C entry setup.

Details

This is intentionally a thin layer over tccq_compile() and is aimed at low-latency interactive workflows where many calls share the same runtime shape, e.g. fixed-size vectors/matrices.

Value

A callable R function with signature-cache semantics. Repeated calls with the same observed signature reuse the compiled callable.

See Also

tccq_compile(), tccq_backend_tinycc(), tccq_backend_shlib()

Examples

f <- function(x) {
  declare(type(x = double(NA)))
  x + 1
}
jf <- tccq_jit(f, exact = TRUE)
jf(c(1, 2, 3))

---