def test_dag_cached() -> None: """Test that DAG caching works.""" serv = MockServer() with Fun(serv, defaults=options(distributed=False)): dat = put(b"bla bla") step1 = morph(lambda x: x.decode().upper().encode(), dat) step2b = shell("echo 'not'", inp=dict(file1=step1)) merge = shell("cat file1 file2", inp=dict(file1=step1, file2=step2b.stdout), out=["file2"]) execute(merge) with Fun(serv, defaults=options(distributed=False, evaluate=False)): # Same as above, should run through with no evaluation dat = put(b"bla bla") step1 = morph(lambda x: x.decode().upper().encode(), dat) step2b = shell("echo 'not'", inp=dict(file1=step1)) merge = shell("cat file1 file2", inp=dict(file1=step1, file2=step2b.stdout), out=["file2"]) execute(merge) with Fun(serv, defaults=options(distributed=False, evaluate=False)): dat = put(b"bla bla") step1 = morph(lambda x: x.decode().upper().encode(), dat) # DIFFERENT HERE: Trigger re-evaluation and raise step2b = shell("echo 'knot'", inp=dict(file1=step1)) merge = shell("cat file1 file2", inp=dict(file1=step1, file2=step2b.stdout), out=["file2"]) with pytest.raises(RuntimeError): execute(merge)
def test_dag_dump() -> None: """Test simple DAG dump to file.""" with Fun(MockServer(), options(distributed=False)) as db: dat = put(b"bla bla") dat2 = put(b"blaXbla") errorstep = morph(raises, dat2) step1 = morph(upper, dat) step2 = shell("cat file1 file2", inp=dict(file1=step1, file2=dat)) step2b = utils.concat(step2.stdout, errorstep, strict=False) step3 = shell("cat file1", inp=dict(file1=step2b)) step4 = shell("cat file1", inp=dict(file1=step1)) step4b = shell("cat file2", inp=dict(file2=step4.stdout)) out = utils.concat(step1, dat, step2.stdout, step3.stdout) _dag.build_dag(db, out.hash) execute(step2b) execute(step4b) wait_for(step4b, 1.0) reset(step4) nodes, artefacts, labels, links = _graphviz.export( db, [out.hash, step4b.hash]) dot = _graphviz.format_dot(nodes, artefacts, labels, links, [out.hash, step4b.hash]) assert len(dot) > 0 assert len(nodes) == 8 assert len(labels) == 8 # TODO pass through dot for testing? with open("g.dot", "w") as f: f.write(dot)
def test_concat() -> None: """Test concatenation.""" with Fun(MockServer()): db, store = get_connection() dat1 = put(b"bla") dat2 = put(b"bla") cat = utils.concat(dat1, dat2) run_op(db, store, cat.parent) assert take(cat) == b"blabla" cat = utils.concat(dat1, dat1, dat1, join=b" ") run_op(db, store, cat.parent) assert take(cat) == b"bla bla bla"
def test_subgraph() -> None: """Test that we can isolate the required operators for parametrization.""" with Fun(MockServer(), options(distributed=False)) as db: dat = put(b"bla bla") step1 = morph(capitalize, dat) step2 = shell("cat file1 file2", inp=dict(file1=step1, file2=dat)) # random not included ops stepA = shell("echo 'bla'") _ = concat(dat, dat) _ = morph(capitalize, b"another word") final = shell("cat file1 file2", inp={ "file1": stepA.stdout, "file2": step2.stdout }) ops = _p._parametrize_subgraph(db, {"input": dat}, {"output": final.stdout}) assert len(ops) == 3 assert step1.parent in ops assert step2.hash in ops assert final.hash in ops # get edges edges = _p._subgraph_edges(db, ops) print(edges)
def test_double_execution(nworkers: int) -> None: """Test multiple executions of the same task.""" # This test will fail if a job is re-executed multiple times. # external from rq.job import get_current_job def track_runs(inp: bytes) -> bytes: job = get_current_job() db: Redis[bytes] = job.connection val = db.incrby("sentinel", 1) time.sleep(0.5) return str(val).encode() with f.ManagedFun(nworkers=nworkers): # wait_for_workers(db, nworkers) dat = f.put(b"bla bla") step1 = f.morph(track_runs, dat) step1a = f.shell( "cat file1", inp=dict(file1=step1), ) step1b = f.shell( "cat file2", inp=dict(file2=step1), ) f.execute(step1a) f.execute(step1b) f.wait_for(step1a, timeout=10.0) f.wait_for(step1b, timeout=10.0) assert f.take(step1a.stdout) == b"1"
def test_subdag() -> None: """Test that subdags execute properly.""" def cap(inp: bytes) -> bytes: return inp.upper() def map_reduce( inputs: dict[str, bytes]) -> dict[str, _graph.Artefact[bytes]]: """Basic map reduce.""" inp_data = inputs["inp"].split(b" ") out: list[_graph.Artefact[bytes]] = [] for el in inp_data: out += [morph(cap, el, opt=options(distributed=False))] return {"out": concat(*out, join="-")} with Fun(MockServer(), defaults=options(distributed=False)) as db: dat = put(b"bla bla lol what") inp = {"inp": dat} cmd = _subdag.subdag_funsie(map_reduce, {"inp": Encoding.blob}, {"out": Encoding.blob}) operation = _graph.make_op(db, cmd, inp, options()) out = _graph.Artefact[bytes].grab(db, operation.out["out"]) final = shell( "cat file1 file2", inp=dict(file1=out, file2=b"something"), ) execute(final) data = take(final.stdout) assert data == b"BLA-BLA-LOL-WHATsomething"
def test_toposort() -> None: """Test that we can topologically sort the subset.""" with Fun(MockServer(), options(distributed=False)) as db: dat = put(b"bla bla") step1 = morph(capitalize, dat) step2 = shell("cat file1 file2", inp=dict(file1=step1, file2=dat)) # random not included ops stepA = shell("echo 'bla'") _ = concat(dat, dat) _ = morph(capitalize, b"another word") final = shell("cat file1 file2", inp={ "file1": stepA.stdout, "file2": step2.stdout }) ops = _p._parametrize_subgraph(db, {"input": dat}, {"output": final.stdout}) edges = _p._subgraph_edges(db, ops) sorted_ops = _p._subgraph_toposort(ops, edges) assert sorted_ops[0] == step1.parent assert sorted_ops[1] == step2.hash assert sorted_ops[2] == final.hash
def test_dynamic_dump() -> None: """Test whether a dynamic DAG gets graphed properly.""" def split(a: bytes, b: bytes) -> list[dict[str, int]]: a = a.split() b = b.split() out = [] for ia, ib in zip(a, b): out += [{ "sum": int(ia.decode()) + int(ib.decode()), "product": int(ia.decode()) * int(ib.decode()), }] return out def apply(inp: Artefact[dict[str, Any]]) -> Artefact[str]: out = funsies.morph(lambda x: f"{x['sum']}//{x['product']}", inp) return out def combine(inp: Sequence[Artefact[str]]) -> Artefact[bytes]: def enc(inp: str) -> bytes: return inp.encode() out = [funsies.morph(enc, x, out=Encoding.blob) for x in inp] return funsies.utils.concat(*out) with funsies.ManagedFun(nworkers=1) as db: num1 = funsies.put(b"1 2 3 4 5") num2 = funsies.put(b"11 10 11 10 11") outputs = dynamic.sac( split, apply, combine, num1, num2, out=Encoding.blob, ) outputs = funsies.morph(lambda x: x, outputs) nodes, artefacts, labels, links = _graphviz.export(db, [outputs.hash]) assert len(artefacts) == 4 # not yet generated subdag parents print(len(artefacts)) funsies.execute(outputs) funsies.wait_for(outputs, timeout=1.0) nodes, artefacts, labels, links = _graphviz.export(db, [outputs.hash]) assert len(artefacts) == 22 # generated subdag parents assert funsies.take(outputs) == b"12//1112//2014//3314//4016//55"
def test_truncate() -> None: """Test truncation.""" with Fun(MockServer()): db, store = get_connection() inp = "\n".join([f"{k}" for k in range(10)]) dat1 = put(inp.encode()) trunc = utils.truncate(dat1, 2, 3) run_op(db, store, trunc.parent) assert take(trunc) == ("\n".join(inp.split("\n")[2:-3])).encode()
def test_parametric_store_recall() -> None: """Test storing and recalling parametrics.""" serv = MockServer() with Fun(serv, options(distributed=False)): a = put(3) b = put(4) s = reduce(lambda x, y: x + y, a, b) s2 = morph(lambda x: 3 * x, s) execute(s2) assert take(s2) == 21 # parametrize p.commit("math", inp=dict(a=a, b=b), out=dict(s=s, s2=s2)) with Fun(serv, options(distributed=False)): out = p.recall("math", inp=dict(a=5, b=8)) execute(out["s2"]) assert take(out["s2"]) == 39
def test_parametric_eval() -> None: """Test that parametric evaluate properly.""" with Fun(MockServer(), options(distributed=False)) as db: dat = put(b"bla bla") step1 = morph(capitalize, dat) step2 = shell("cat file1 file2", inp=dict(file1=step1, file2=dat)) final = shell("cat file1 file3", inp={ "file1": step1, "file3": step2.stdout }) execute(final.stdout) # b'BLA BLABLA BLAbla bla' param = _p.make_parametric(db, "param", {"input": dat}, {"output": final.stdout}) dat2 = put(b"lol lol") out = param.evaluate(db, {"input": dat2}) execute(out["output"]) assert take(out["output"]) == b"LOL LOLLOL LOLlol lol"
def test_dag_execute() -> None: """Test execution of a _dag.""" with Fun(MockServer(), defaults=options(distributed=False)): dat = put(b"bla bla") step1 = morph(lambda x: x.decode().upper().encode(), dat) step2 = shell("cat file1 file2", inp=dict(file1=step1, file2=dat)) output = step2.stdout # make queue execute(output) out = take(output) assert out == b"BLA BLAbla bla"
def test_waiting_on_map_reduce() -> None: """Test waiting on the (linked) result of map-reduce.""" def split(a: bytes, b: bytes) -> list[dict[str, int]]: a = a.split() b = b.split() out = [] for ia, ib in zip(a, b): out += [{ "sum": int(ia.decode()) + int(ib.decode()), "product": int(ia.decode()) * int(ib.decode()), }] return out def apply(inp: Artefact) -> Artefact: out = funsies.morph(lambda x: f"{x['sum']}//{x['product']}", inp) return out def combine(inp: Sequence[Artefact]) -> Artefact: out = [ funsies.morph(lambda y: y.encode(), x, out=Encoding.blob) for x in inp ] return funsies.utils.concat(*out) with funsies.ManagedFun(nworkers=1): num1 = funsies.put(b"1 2 3 4 5") num2 = funsies.put(b"11 10 11 10 11") outputs = dynamic.sac( split, apply, combine, num1, num2, out=Encoding.blob, ) funsies.execute(outputs) funsies.wait_for(outputs, timeout=1.0) assert funsies.take(outputs) == b"12//1112//2014//3314//4016//55"
def test_map_reduce() -> None: """Test simple map-reduce.""" def split(a: bytes, b: bytes) -> list[dict[str, int]]: a = a.split() b = b.split() out = [] for ia, ib in zip(a, b): out += [{ "sum": int(ia.decode()) + int(ib.decode()), "product": int(ia.decode()) * int(ib.decode()), }] return out def apply(inp: Artefact) -> Artefact: out = funsies.morph(lambda x: f"{x['sum']}//{x['product']}", inp) return out def combine(inp: Sequence[Artefact]) -> Artefact: out = [ funsies.morph(lambda y: y.encode(), x, out=Encoding.blob) for x in inp ] return funsies.utils.concat(*out) with funsies.Fun(MockServer(), funsies.options(distributed=False)): num1 = funsies.put(b"1 2 3 4 5") num2 = funsies.put(b"11 10 11 10 11") outputs = dynamic.sac( split, apply, combine, num1, num2, out=Encoding.blob, ) funsies.execute(outputs) assert funsies.take(outputs) == b"12//1112//2014//3314//4016//55"
def test_parametric_store_recall_optional() -> None: """Test storing a parametric with optional parameters.""" serv = MockServer() with Fun(serv, options(distributed=False)): a = put(3) b = put("fun") s = reduce(lambda x, y: x * y, a, b) s2 = morph(lambda x: x.upper(), s) # parametrize p.commit("fun", inp=dict(a=a, b=b), out=dict(s=s2)) with Fun(serv, options(distributed=False)): out = p.recall("fun", inp=dict(a=5)) execute(out["s"]) assert take(out["s"]) == "FUNFUNFUNFUNFUN" # nested out = p.recall("fun", inp=dict(b="lol")) out = p.recall("fun", inp=dict(b=out["s"], a=2)) execute(out["s"]) assert take(out["s"]) == "LOLLOLLOLLOLLOLLOL"
def test_infer_errs() -> None: """Test inference applied to functions.""" with f.Fun(MockServer()): a = f.put(b"bla bla") b = f.put(3) with pytest.raises(TypeError): f.py(lambda x, y, z: (x, y), a, a, b) # should NOT raise f.py( lambda x, y, z: (x, y), a, a, b, out=[types.Encoding.blob, types.Encoding.blob], ) def i1o2(x: bytes) -> Tuple[bytes, bytes]: return x, x def i2o1(x: bytes, y: bytes) -> bytes: return x with pytest.raises(TypeError): out = f.morph(i1o2, a) # type:ignore # noqa:F841 with pytest.raises(TypeError): out = f.reduce(i1o2, a) # type:ignore # noqa:F841 with pytest.raises(TypeError): out = f.reduce(lambda x, y: x, a, b) # type:ignore # noqa:F841 # If we pass out= then the inference is skipped out = f.morph(i1o2, a, out=types.Encoding.blob) # type:ignore # noqa:F841 out = f.reduce(i1o2, a, out=types.Encoding.blob) # type:ignore # noqa:F841
def test_parametrize() -> None: """Test that parametrization works.""" with Fun(MockServer(), options(distributed=False)) as db: dat = put(b"bla bla") dat2 = put(b"bla bla bla") step1 = morph(capitalize, dat) step2 = shell("cat file1 file2", inp=dict(file1=step1, file2=dat)) final = shell("cat file1 file3", inp={ "file1": step1, "file3": step2.stdout }) pinp = {"input": dat} pout = {"final.stdout": final.stdout, "step1": step1} new_inp = {"input": dat2} ops = _p._parametrize_subgraph(db, pinp, pout) edges = _p._subgraph_edges(db, ops) sorted_ops = _p._subgraph_toposort(ops, edges) pinp2 = dict([(k, v.hash) for k, v in pinp.items()]) pout2 = dict([(k, v.hash) for k, v in pout.items()]) new_out = _p._do_parametrize(db, sorted_ops, pinp2, pout2, new_inp) # re-run with dat2, check if the same. step1 = morph(capitalize, dat2) step2 = shell("cat file1 file2", inp=dict(file1=step1, file2=dat2)) final = shell("cat file1 file3", inp={ "file1": step1, "file3": step2.stdout }) assert new_out["final.stdout"] == final.stdout assert new_out["step1"] == step1
def test_data_race(nworkers: int) -> None: """Test a data race when execute calls are interleaved.""" with f.ManagedFun(nworkers=nworkers): dat = f.put(b"bla bla") step1 = f.morph(lambda x: x.decode().upper().encode(), dat) step2 = f.shell( "cat file1 file2; grep 'bla' file2 file1 > file3; date >> file3", inp=dict(file1=step1, file2=dat), out=["file2", "file3"], ) f.execute(step1) f.execute(step2) f.wait_for(step1, timeout=20.0) f.wait_for(step2, timeout=20.0)
def test_dag_execute_same_root() -> None: """Test execution of two dags that share the same origin.""" with Fun(MockServer(), defaults=options(distributed=False)): dat = put(b"bla bla") step1 = morph(lambda x: x.decode().upper().encode(), dat) step2 = shell("cat file1 file2", inp=dict(file1=step1, file2=dat)) step2b = shell("cat file1", inp=dict(file1=step1)) execute(step2) out = take(step2.stdout) assert out == b"BLA BLAbla bla" execute(step2b) out = take(step2b.stdout) assert out == b"BLA BLA"
def test_parametric() -> None: """Test that parametric DAGs work.""" with Fun(MockServer(), options(distributed=False)) as db: dat = put(b"bla bla") step1 = morph(capitalize, dat) step2 = shell("cat file1 file2", inp=dict(file1=step1, file2=dat)) final = shell("cat file1 file3", inp={ "file1": step1, "file3": step2.stdout }) param = _p.make_parametric(db, "param", {"input": dat}, {"output": final.stdout}) param2 = _p.Parametric.grab(db, param.hash) assert param == param2
def test_dag_large() -> None: """Test that DAG building doesn't do extra work for large operations.""" with Fun(MockServer()) as db: outputs = [] for i in range(100): dat = put(f"bla{i}".encode()) step1 = morph(lambda x: x.decode().upper().encode(), dat) step2 = shell( "cat file1 file2", inp=dict(file1=step1, file2="something"), out=["file2"], ) outputs += [concat(step1, step1, step2.stdout, join=b" ")] final = concat(*outputs, join=b"\n") _dag.build_dag(db, final.hash) assert len(_dag._dag_dependents(db, final.hash, hash_t("root"))) == 100
def test_template_complicated() -> None: """Test templating with funky types.""" with Fun(MockServer()): db, store = get_connection() t = "wazzaa, {{ mustache }}!" result = template(t, {"mustache": put(b"people")}) run_op(db, store, result.parent) assert take(result) == b"wazzaa, people!" t = "{{a}}{{b}}{{c}}" result = template(t, dict(a=2, b="cool", c="4me")) run_op(db, store, result.parent) assert take(result) == b"2cool4me" t = "" result = template(t, dict(a=2, b="cool", c="4me")) run_op(db, store, result.parent) assert take(result) == b""
def test_dag_build() -> None: """Test simple DAG build.""" with Fun(MockServer()) as db: dat = put(b"bla bla") step1 = morph(lambda x: x.decode().upper().encode(), dat) step2 = shell("cat file1 file2", inp=dict(file1=step1, file2=dat)) output = step2.stdout _dag.build_dag(db, output.hash) assert len(db.smembers(join(DAG_OPERATIONS, output.hash))) == 2 # test deletion _dag.delete_all_dags(db) assert len(db.smembers(join(DAG_OPERATIONS, output.hash))) == 0 # test new _dag _dag.build_dag(db, step1.hash) assert len(db.smembers(join(DAG_OPERATIONS, step1.hash))) == 1 assert len(_dag.descendants(db, step1.parent)) == 1
def test_exec_all() -> None: """Test execute_all.""" with Fun(MockServer(), defaults=options(distributed=False)): results = [] def div_by(x: float) -> float: return 10.0 / x for i in range(10, -1, -1): val = put(float(i)) results += [morph(div_by, val)] with pytest.raises(UnwrapError): take(results[0]) err = utils.execute_all(results) print(take(results[0])) v = take(err, strict=False) assert isinstance(v, Error) assert v.kind == ErrorKind.ExceptionRaised
def test_artefact_disk_distributed() -> None: """Test whether artefacts on disk works on different nodes.""" # funsies import funsies as f with tempfile.TemporaryDirectory() as td: with f.ManagedFun(nworkers=1, data_url=f"file://{td}"): dat = f.put(b"bla bla") step1 = f.morph(lambda x: x.decode().upper().encode(), dat) step2 = f.shell("cat file1 file2", inp=dict(file1=step1, file2=dat)) step2b = f.shell("cat file1", inp=dict(file1=step1)) f.execute(step2) f.wait_for(step2, 1.0) out = f.take(step2.stdout) assert out == b"BLA BLAbla bla" f.execute(step2b) f.wait_for(step2b, 1.0) out = f.take(step2b.stdout) assert out == b"BLA BLA"
def test_dag_efficient() -> None: """Test that DAG building doesn't do extra work.""" with Fun(MockServer()) as db: dat = put(b"bla bla") step1 = morph(lambda x: x.decode().upper().encode(), dat) step2 = shell("cat file1 file2", inp=dict(file1=step1, file2=dat), out=["file2"]) step2b = shell("echo 'not'", inp=dict(file1=step1)) merge = shell("cat file1 file2", inp=dict(file1=step1, file2=step2b.stdout), out=["file2"]) _dag.build_dag(db, step2.stdout.hash) # check that step2 only has stdout has no dependents assert len(_dag._dag_dependents(db, step2.stdout.hash, step2.hash)) == 0 assert len(_dag._dag_dependents(db, step2.stdout.hash, step1.parent)) == 1 _dag.build_dag(db, merge.hash) # check that however, the merged one has two dependents for step1 assert len(_dag._dag_dependents(db, merge.hash, step1.parent)) == 2
# Recursive application of merge sort # split -> generates two list or raises # recurse(x) for each values of split # merge(left, right) split, lambda element: funsies_mergesort(element), lambda lr: f.reduce(merge, lr[0], lr[1]), # type:ignore art, out=Encoding.json, ) return f.reduce( # if the subdag fails, it's because split raised. In this case, we # just forward the arguments ignore_error, result, art, strict=False, out=Encoding.json, ) # run the workflow to_be_sorted = [random.randint(0, 99) for _ in range(120)] with f.Fun(): inp = f.put(to_be_sorted) out = funsies_mergesort(inp) print("output:", out.hash) f.execute(out) f.wait_for(out) print(f.take(out))
def test_integration(reference: str, nworkers: int) -> None: """Test full integration.""" # make a temp file and copy reference database dir = tempfile.mkdtemp() if not make_reference: shutil.copy(os.path.join(ref_dir, reference, "appendonly.aof"), dir) shutil.copytree(os.path.join(ref_dir, reference, "data"), os.path.join(dir, "data")) shutil.copy(os.path.join(ref_dir, "redis.conf"), dir) # data url datadir = f"file://{os.path.join(dir, 'data')}" # Dictionary for test data test_data: dict[str, Any] = {} def update_data(a: dict[int, int], b: list[int]) -> dict[int, int]: for i in b: a[i] = a.get(i, 0) + 1 return a def sum_data(x: dict[int, int]) -> int: return sum([int(k) * v for k, v in x.items()]) def make_secret(x: int) -> str: return secrets.token_hex(x) # Start funsie script with ManagedFun( nworkers=nworkers, directory=dir, data_url=datadir, redis_args=["redis.conf"], ) as db: integers = put([5, 4, 8, 9, 9, 10, 1, 3]) init_data = put({100: 9}) test_data["init_data"] = init_data nbytes = put(4) s1 = reduce(update_data, init_data, integers) num = morph(sum_data, s1) date = shell("date").stdout test_data["date"] = date rand = morph(make_secret, nbytes) s4 = template( "date:{{date}}\n" + "some random bytes:{{random}}\n" + "a number: {{num}}\n" + "a string: {{string}}\n", { "date": date, "random": rand, "num": num, "string": "wazza" }, name="a template", ) test_data["s4"] = s4 execute(s4) wait_for(s4, 5) # check that the db doesn't itself include data for k in db.keys(): assert b"data" not in k if make_reference: folder = os.path.join(ref_dir, reference) os.makedirs(folder, exist_ok=True) for name, artefact in test_data.items(): with open(os.path.join(folder, name), "wb") as f: execute(artefact) wait_for(artefact, 10.0) out = take(artefact) data2 = _serdes.encode(artefact.kind, out) assert isinstance(data2, bytes) f.write(data2) shutil.copy( os.path.join(dir, "appendonly.aof"), os.path.join(folder, "appendonly.aof"), ) shutil.copytree( os.path.join(dir, "data"), os.path.join(folder, "data"), ) else: # Test against reference dbs for name, artefact in test_data.items(): execute(artefact) wait_for(artefact, 10.0) with open(os.path.join(ref_dir, reference, name), "rb") as f: data = f.read() out = take(artefact) data_ref = _serdes.encode(artefact.kind, out) assert isinstance(data_ref, bytes) assert data == data_ref # delete tempdir shutil.rmtree(dir)
return {"structure": as_str, "energy": energy} def sort_by_energy(*elements: dict[str, Any]) -> list[dict[str, Any]]: out = [el for el in elements] out = sorted(out, key=lambda x: x["energy"]) # type:ignore return out out = [] for s in structures: out += [f.morph(to_dict, s, out=Encoding.json)] # elements to dicts return f.reduce(sort_by_energy, *out) # transform to a sorted list with f.Fun(): # put smiles in db smiles = f.put(b"C(O)CCCC(O)") # Generate 3d conformers with openbabel gen3d = f.shell( "obabel input.smi --gen3d --ff mmff94 --minimize -O struct.mol", inp={"input.smi": smiles}, out=["struct.mol"], ) # abort if molecule is empty struct = not_empty(gen3d.out["struct.mol"]) # Generate conformers. confab = f.shell( "obabel input.mol -O conformers.xyz --confab --verbose", inp={"input.mol": struct}, out=["conformers.xyz"],
def test_nested_map_reduce(nworkers: int) -> None: """Test nested map-reduce.""" # ------------------------------------------------------------------------ # Inner def sum_inputs(*inp: int) -> int: out = 0 for el in inp: out += el return out def split_inner(inp: str) -> list[int]: a = inp.split(" ") return [int(el) for el in a] def apply_inner(inp: Artefact) -> Artefact: return funsies.reduce(sum_inputs, inp, 1) def combine_inner(inp: Sequence[Artefact]) -> Artefact: return funsies.reduce(sum_inputs, *inp) # ------------------------------------------------------------------------ # outer def split_outer(inp: list[str], fac: int) -> list[str]: out = [x + f" {fac}" for x in inp] return out def apply_outer(inp: Artefact) -> Artefact: outputs = dynamic.sac( split_inner, apply_inner, combine_inner, inp, out=Encoding.json, ) return outputs def combine_outer(inp: Sequence[Artefact]) -> Artefact: out = [ funsies.morph(lambda y: f"{y}".encode(), x, out=Encoding.blob) for x in inp ] return funsies.utils.concat(*out, join=b",,") with funsies.ManagedFun(nworkers=nworkers): num1 = funsies.put("1 2 3 4 5") outputs = dynamic.sac(split_inner, apply_inner, combine_inner, num1, out=Encoding.json) funsies.execute(outputs) funsies.wait_for(outputs, timeout=30.0) assert funsies.take(outputs) == 20 # Now try the nested one num = funsies.put(["1 2", "3 4 7", "10 12", "1"]) factor = funsies.put(-2) # split -> 1 2 -2|3 4 7 -2|10 12 -2| 1 -2 # apply -> split2 -> 1, 2,-2 | 3,4,7,-2|10,12,-2|1,-2 # apply2 -> 2, 3,-1 | 4,5,8,-1|11,13,-1|2,-1 # combine2 -> 4|16|23|1 # combine -> 4,,16,,23,,1 ans = b"4,,16,,23,,1" outputs = dynamic.sac( split_outer, apply_outer, combine_outer, num, factor, out=Encoding.blob, ) funsies.execute(outputs) funsies.wait_for(outputs, timeout=30.0) assert funsies.take(outputs) == ans