Locate a Python Interpreter That Can Import a Module

Search order (first verified hit wins):

getOption("templecbe.python")
Sys.getenv("TEMPLECBE_PYTHON")
Sys.getenv("RETICULATE_PYTHON") and, only when reticulate has already started Python in this session, reticulate::py_exe()
Sys.which("python3"), Sys.which("python")
Platform-specific well-known locations

Usage

find_python(verify = TRUE, module = "pdf2docx")

Arguments

verify: If TRUE, require import <module> to succeed.
module: Name of the Python module an interpreter must be able to import when verify = TRUE: a plain, possibly dotted, module name. Defaults to "pdf2docx".

Value

Path to a usable interpreter, or NULL.

Details

Each candidate is verified by actually running it, so Windows App Execution Alias stubs (which resolve on PATH but do nothing) are rejected. A probe that takes longer than 60 seconds (running --version, or importing module) counts as a failure, so one hung interpreter cannot block the search. When verify = TRUE the candidate must additionally be able to import module. Candidates are probed in a separate process: this function never starts Python inside R, and never starts reticulate's Python, so it is safe to use as a guard before code that would (for example eval = !is.null(find_python(module = "numpy")) on a vignette chunk). Install the pinned Python requirements for PDF conversion with pip install -r on system.file("python", "requirements.txt", package = "TempleCBE").

The tests that run a real PDF to DOCX conversion (Python, LibreOffice or Word) are opt-in: set the environment variable TEMPLECBE_RUN_PDF_TESTS=true to run them. See also find_sas() for TEMPLECBE_RUN_SAS_TESTS.