Update README.md: add pipeline descripiton

README.md

@@ -103,6 +103,49 @@ assert tst == 1
 
 ```
 
+----
+
+#### Сборка конвейеров преобразований:
+
+Пусть, имеется несколько методов-потребителей, которые необходимо вызывать последовательно:  
+
+```python
+
+from breakshaft.convertor import ConvRepo
+
+repo = ConvRepo()
+
+# Объявляем A и B, а также методы преобразований - как в прошлом примере
+
+type cons2ret = str  # избегаем использования builtin-типов, чтобы избежать простых коллизий 
+
+
+def consumer1(dep: A) -> B:
+    return B(float(42))
+
+
+def consumer2(dep: B) -> cons2ret:
+    return str(dep.b)
+
+
+def consumer3(dep: cons2ret) -> int:
+    return int(float(dep))
+
+
+pipeline = repo.create_pipeline(
+    (B,),
+    [consumer1, consumer2, consumer3],
+    force_commutative=True,
+    allow_sync=True,
+    allow_async=False,
+    force_async=False
+)
+
+dat = pipeline(B(42))
+assert dat == 42
+```
+
+
 ----
 
 #### Как получить граф преобразований:

src/breakshaft/convertor.py

@@ -1,9 +1,10 @@
 from __future__ import annotations
-from typing import Optional, Callable, Unpack, TypeVarTuple, TypeVar, Awaitable, Any
+from typing import Optional, Callable, Unpack, TypeVarTuple, TypeVar, Awaitable, Any, Sequence
 
 from .graph_walker import GraphWalker
 from .models import ConversionPoint, Callgraph
 from .renderer import ConvertorRenderer, InTimeGenerationConvertorRenderer
+from .util import extract_return_type
 
 Tin = TypeVarTuple('Tin')
 Tout = TypeVar('Tout')
@@ -27,6 +28,31 @@ class ConvRepo:
         self.walker = graph_walker
         self.renderer = renderer
 
+    def create_pipeline(self,
+                        from_types: Sequence[type],
+                        fns: Sequence[Callable],
+                        force_commutative: bool = True,
+                        allow_async: bool = True,
+                        allow_sync: bool = True,
+                        force_async: bool = False
+                        ):
+        filtered_injectors = self.filtered_injectors(allow_async, allow_sync)
+        pipeline_callseq = []
+        orig_from_types = tuple(from_types)
+        from_types = tuple(from_types)
+
+        for fn in fns:
+            injects = extract_return_type(fn)
+
+            callseq = self.get_callseq(filtered_injectors, frozenset(from_types), fn, force_commutative)
+
+            pipeline_callseq += callseq
+
+            if injects is not None:
+                from_types += (injects,)
+
+        return self.renderer.render(orig_from_types, pipeline_callseq, force_async=force_async)
+
     @property
     def convertor_set(self):
         return self._convertor_set
@@ -46,15 +72,7 @@ class ConvRepo:
         ret += [variant.injector]
         return ret
 
-    def get_conversion(self,
-                       from_types: tuple[type[Unpack[Tin]]],
-                       fn: Callable[..., Tout],
-                       force_commutative: bool = True,
-                       allow_async: bool = True,
-                       allow_sync: bool = True,
-                       force_async: bool = False
-                       ) -> Callable[[Unpack[Tin]], Tout] | Awaitable[Callable[[Unpack[Tin]], Tout]]:
-        if not allow_async or force_async:
+    def filtered_injectors(self, allow_async: bool, allow_sync: bool) -> frozenset[ConversionPoint]:
         filtered_injectors: frozenset[ConversionPoint] = frozenset()
         for inj in self.convertor_set:
             if inj.is_async and not allow_async:
@@ -62,14 +80,18 @@ class ConvRepo:
             if not inj.is_async and not allow_sync:
                 continue
             filtered_injectors |= {inj}
-        else:
-            filtered_injectors = frozenset(self.convertor_set)
+        return filtered_injectors
 
-        cg = self.walker.generate_callgraph(filtered_injectors, frozenset(from_types), fn)
+    def get_callseq(self,
+                    injectors: frozenset[ConversionPoint],
+                    from_types: frozenset[type], fn: Callable,
+                    force_commutative: bool) -> list[ConversionPoint]:
+
+        cg = self.walker.generate_callgraph(injectors, from_types, fn)
         if cg is None:
             raise ValueError(f'Unable to compute conversion graph on {from_types}->{fn.__qualname__}')
 
-        exploded = self.walker.explode_callgraph_branches(cg, frozenset(from_types))
+        exploded = self.walker.explode_callgraph_branches(cg, from_types)
 
         selected = self.walker.filter_exploded_callgraph_branch(exploded)
         if len(selected) == 0:
@@ -79,6 +101,25 @@ class ConvRepo:
             raise ValueError('Conversion path is not commutative')
 
         callseq = self._callseq_from_callgraph(Callgraph(frozenset([selected[0]])))
+
+        if len(callseq) > 0:
+            injects = extract_return_type(fn)
+            callseq[-1] = callseq[-1].copy_with(injects=injects)
+
+        return callseq
+
+    def get_conversion(self,
+                       from_types: Sequence[type[Unpack[Tin]]],
+                       fn: Callable[..., Tout],
+                       force_commutative: bool = True,
+                       allow_async: bool = True,
+                       allow_sync: bool = True,
+                       force_async: bool = False
+                       ) -> Callable[[Unpack[Tin]], Tout] | Awaitable[Callable[[Unpack[Tin]], Tout]]:
+
+        filtered_injectors = self.filtered_injectors(allow_async, allow_sync)
+        callseq = self.get_callseq(filtered_injectors, frozenset(from_types), fn, force_commutative)
+
         return self.renderer.render(from_types, callseq, force_async=force_async)
 
     def mark_injector(self, *, rettype: Optional[type] = None):

src/breakshaft/graph_walker.py

@@ -3,7 +3,7 @@ from types import NoneType
 from typing import Callable, Optional
 
 from .models import ConversionPoint, Callgraph, CallgraphVariant, TransformationPoint, CompositionDirection
-from .util import extract_func_argtypes, all_combinations, extract_func_argtypes_seq
+from .util import extract_func_argtypes, all_combinations, extract_func_argtypes_seq, extract_return_type
 
 
 class GraphWalker:
@@ -15,6 +15,7 @@ class GraphWalker:
                            consumer_fn: Callable) -> Optional[Callgraph]:
 
         branches: frozenset[Callgraph] = frozenset()
+        rettype = extract_return_type(consumer_fn)
 
         # Хак, чтобы вынудить систему поставить первым преобразованием требуемый consumer
         # Новый TypeAliasType каждый раз будет иметь эксклюзивный хэш, вне зависимости от содержимого

src/breakshaft/models.py

@@ -20,6 +20,13 @@ class ConversionPoint:
     requires: tuple[type, ...]
     opt_args: tuple[type, ...]
 
+    def copy_with(self, **kwargs):
+        fn = kwargs.get('fn', self.fn)
+        injects = kwargs.get('injects', self.injects)
+        requires = kwargs.get('requires', self.requires)
+        opt_args = kwargs.get('opt_args', self.opt_args)
+        return ConversionPoint(fn, injects, requires, opt_args)
+
     def __hash__(self):
         return hash((self.fn, self.injects, self.requires))
 

src/breakshaft/renderer.py

@@ -51,6 +51,28 @@ class ConversionArgRenderData:
     typehash: str
 
 
+def deduplicate_callseq(conversion_models: list[ConversionRenderData]) -> list[ConversionRenderData]:
+    deduplicated_conv_models: list[ConversionRenderData] = []
+    for conv_model in conversion_models:
+        if conv_model not in deduplicated_conv_models:
+            deduplicated_conv_models.append(conv_model)
+            continue
+
+        argnames = list(map(lambda x: x[1], conv_model.funcargs))
+        argument_changed = False
+        found_model = False
+        for m in deduplicated_conv_models:
+            if not found_model and m == conv_model:
+                found_model = True
+
+            if found_model and m.inj_hash in argnames:
+                argument_changed = True
+                break
+        if argument_changed:
+            deduplicated_conv_models.append(conv_model)
+    return deduplicated_conv_models
+
+
 class InTimeGenerationConvertorRenderer(ConvertorRenderer):
     templateLoader: jinja2.BaseLoader
     templateEnv: jinja2.Environment
@@ -72,7 +94,7 @@ class InTimeGenerationConvertorRenderer(ConvertorRenderer):
                force_async: bool = False) -> Callable:
 
         fnmap = {}
-        conversion_models = []
+        conversion_models: list[ConversionRenderData] = []
         ret_hash = 0
         is_async = force_async
         for call_id, call in enumerate(callseq):
@@ -84,12 +106,13 @@ class InTimeGenerationConvertorRenderer(ConvertorRenderer):
 
             fnmap[hash(call.fn)] = call.fn
             conv = ConversionRenderData.from_inj(call, provided_types)
-            if conv not in conversion_models:
             conversion_models.append(conv)
             if call.is_async:
                 is_async = True
 
-        ret_hash = hash(callseq[-1].injects)
+        conversion_models = deduplicate_callseq(conversion_models)
+
+        ret_hash = hashname(callseq[-1].injects)
 
         conv_args = []
         for i, from_type in enumerate(from_types):

src/breakshaft/util.py

@@ -1,6 +1,11 @@
 import inspect
 from itertools import product
-from typing import Callable, get_type_hints, TypeVar, Any
+from typing import Callable, get_type_hints, TypeVar, Any, Optional
+
+
+def extract_return_type(func: Callable) -> Optional[type]:
+    hints = get_type_hints(func)
+    return hints.get('return')
 
 
 def extract_func_args(func: Callable) -> list[tuple[str, type]]:

tests/test_pipeline.py

@@ -0,0 +1,108 @@
+from dataclasses import dataclass
+
+from src.breakshaft.convertor import ConvRepo
+
+
+@dataclass
+class A:
+    a: int
+
+
+@dataclass
+class B:
+    b: float
+
+
+type optC = str
+
+
+def test_default_consumer_args():
+    repo = ConvRepo()
+
+    @repo.mark_injector()
+    def b_to_a(b: B) -> A:
+        return A(int(b.b))
+
+    @repo.mark_injector()
+    def a_to_b(a: A) -> B:
+        return B(float(a.a))
+
+    @repo.mark_injector()
+    def int_to_a(i: int) -> A:
+        return A(i)
+
+    type ret1 = tuple[int, str]
+
+    def consumer1(dep: A, opt_dep: optC = '42') -> ret1:
+        return dep.a, opt_dep
+
+    def consumer2(dep: A, dep1: ret1) -> optC:
+        return str((dep.a, dep1))
+
+    p1 = repo.create_pipeline(
+        (B,),
+        [consumer1, consumer2],
+        force_commutative=True,
+        allow_sync=True,
+        allow_async=False,
+        force_async=False
+    )
+    res = p1(B(42.1))
+    assert res == "(42, (42, '42'))"
+
+    p2 = repo.create_pipeline(
+        (B,),
+        [consumer1, consumer2, consumer1],
+        force_commutative=True,
+        allow_sync=True,
+        allow_async=False,
+        force_async=False
+    )
+    res = p2(B(42.1))
+    assert res == (42, "(42, (42, '42'))")
+
+
+def test_pipeline_with_subgraph_duplicates():
+    repo = ConvRepo()
+
+    b_to_a_calls = [0]
+
+    @repo.mark_injector()
+    def b_to_a(b: B) -> A:
+        b_to_a_calls[0] += 1
+        return A(int(b.b))
+
+    @repo.mark_injector()
+    def a_to_b(a: A) -> B:
+        return B(float(a.a))
+
+    @repo.mark_injector()
+    def int_to_a(i: int) -> A:
+        return A(i)
+
+    type ret1 = tuple[int, str]
+
+    cons1_calls = [0]
+    cons2_calls = [0]
+
+    def consumer1(dep: A, opt_dep: optC = '42') -> A:
+        cons1_calls[0] += 1
+        return A(dep.a + int(opt_dep))
+
+    def consumer2(dep: A) -> optC:
+        cons2_calls[0] += 1
+        return str(dep.a)
+
+    p1 = repo.create_pipeline(
+        (B,),
+        [consumer1, consumer2, consumer1, consumer2, consumer1, consumer2, consumer1, consumer2, consumer1],
+        force_commutative=True,
+        allow_sync=True,
+        allow_async=False,
+        force_async=False
+    )
+    res = p1(B(42.1))
+    assert res.a == 42 + (42 * 31)
+    assert b_to_a_calls[0] == 1
+    assert cons1_calls[0] == 5
+    assert cons2_calls[0] == 4