Getting started

forterp is usable as a library: parse and run FORTRAN from Python, inspect the result, swap the value model or dialect, embed the engine in a host program, or instrument it. The package root exposes a focused surface; deeper machinery lives in explicit namespaces.

The public surface

import forterp gives exactly these names (everything else is in a namespace, below):

Name	What
run_source(text, …)	parse + run a source string; returns the Engine
parse_source(text, …)	parse → {name: ProgramUnit} (raises ParseError)
f66, fortran10, f77	prebuilt, ready-to-run interpreters
Interpreter	roll your own (target + dialect + runtime)
F66, FORTRAN10, F77, Dialect	the front-end dialect axis
NATIVE, PDP10, LP64LE, VAX, Target	the machine value-model axis
ParseError, SourceOptions	the parse error; source-recovery options

Run and inspect

run_source returns the Engine, so you read results out of it after the run. By convention, programs write results into a COMMON block, which the engine exposes as a flat list of words:

import forterp

src = """      PROGRAM T
      COMMON /OUT/ V(40)
      V(1) = 2 ** 10
      END
"""
eng = forterp.run_source(src, dialect=forterp.FORTRAN10)
print(eng.commons["OUT"][0])        # 1024

run_source(text, program=None, dialect=F66, options=None, include_dir=".", **kwargs): program picks the main unit (default: the first); dialect is F77 (default), FORTRAN10, or F77; include_dir is where INCLUDE resolves; extra **kwargs pass through to the Engine (see Embedding).

parse_source(text, dialect=F66, on_error=None, options=None, include_dir=".") parses without running and returns {name: ProgramUnit}. It raises ParseError (with every diagnostic in the message) unless you pass on_error(statement, message), which receives each diagnostic and keeps the partial result.