Export and document lark.tools.standalone.gen_standalone

[fyellin]

Please export and document:
lark.tools.standalone.gen_standalone

While I love the standalone parser, I don't like the fact that the grammar has to live in its own file. My experience is that the grammar and the Transformer evolve in parallel, and having them next to each other makes that task easier.

I've been using the following code to let me quickly switch between using the standalone parser and the standard parser. I've had to reach into lark.tools.standalone and use methods that I'm not sure are public. But the grammar is just a string in the file.

def generate_standalone(filename: Path | str, compress=False, verbose=True):
    from lark import lark
    from lark.tools.standalone import gen_standalone
    parser = lark.Lark(GRAMMAR, parser='lalr', start='statement')
    with open(filename, "w") as out:
        gen_standalone(parser, out=out, compress=compress)
        out.write(f"\nGRAMMAR_HASH = '{get_grammar_hash()}'\n")
    assert filename.exists() and filename.stat().st_size > 1_000, f"Failed to generate {filename}"
    if verbose:
        print(f"Generated {filename} size: {filename.stat().st_size:,} bytes")

def get_grammar_hash() -> str:
    return hashlib.sha256(GRAMMAR.encode('utf-8')).hexdigest()

At use time I then have:

USE_PREBUILT = True

# Note that "lark" is a module if USE_PREBUILT is True and a package if it is False.
# This makes "from XX import YY" difficult. However, qualified names, Lark.XX, work the
# same for both, so we use those.

if USE_PREBUILT:
    import solver.equation_parser_prebuilt as lark
else:
    import lark

.....
def MyParser:

    @staticmethod
    def get_parser(transformer: lark.Transformer | None) -> lark.Lark:
        if not USE_PREBUILT:
            parser = lark.Lark(GRAMMAR, parser='lalr', cache=True, start='statement',
                               transformer=transformer)
        else:
            parser = lark.Lark_StandAlone(transformer=transformer)
            if lark.GRAMMAR_HASH != EquationParser.get_grammar_hash():
                raise Exception("Prebuilt parser is out of date")
        return parser

[erezsh]

Yes, gen_standalone() isn't meant to be an official API, but I think it should be pretty easy to make it into one.

Maybe the API can be something like lark_instance.generate_standalone(...), that will return a string. Then the user can either exec it, or run and then import it.

How does that sound?

Btw, I like having the grammar and transformer in two separate files. That way I get syntax highlighting for both. If I need to go back and forth between them, I can use a split-screen. Just my two cents.

[fyellin]

My preference is that generate_standalone stay exactly as it is. And be documented to do so. The `out` argument means that if I want to write the results to a file, I can do so (as in my code). If I want the results as a String, I can set the `out` argument to be a StringIO. Two files vs one file? Interesting thought. For now, my grammar is reasonably small and one file is a lot more convenient. Thank you for your help and your quick responses.

…

On Mon, Mar 23, 2026 at 2:11 PM Erez Shinan ***@***.***> wrote: *erezsh* left a comment (lark-parser/lark#1583) <#1583 (comment)> Yes, gen_standalone() isn't meant to be an official API, but I think it should be pretty easy to make it into one. Maybe the API can be something like lark_instance.generate_standalone(...), that will return a string. Then the user can either exec it, or run and then import it. How does that sound? Btw, I like having the grammar and transformer in two separate files. That way I get syntax highlighting for both. If I need to go back and forth between them, I can use a split-screen. Just my two cents. — Reply to this email directly, view it on GitHub <#1583?email_source=notifications&email_token=ABC2LIMCXAN63JPF5IXJGG34SGSBXA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTIMJRGM3TGNBSG44KM4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2KYZTPN52GK4S7MNWGSY3L#issuecomment-4113734278>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABC2LIK3JQ7GWVPCIAAO5V34SGSBXAVCNFSM6AAAAACW4T2IEKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHM2DCMJTG4ZTIMRXHA> . You are receiving this because you authored the thread.Message ID: ***@***.***>