instance-repo 1.4.0.dev4__tar.gz → 1.4.1.dev0__tar.gz

{instance_repo-1.4.0.dev4 → instance_repo-1.4.1.dev0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: instance-repo
-Version: 1.4.0.dev4
+Version: 1.4.1.dev0
 Summary: InstanceRepo SDK — 评测 Instance 统一存储客户端（SDK-only；控制面由 apiserver 承载）
 Requires-Python: >=3.10
 Description-Content-Type: text/markdown

{instance_repo-1.4.0.dev4 → instance_repo-1.4.1.dev0}/instance_repo/__init__.py

@@ -20,7 +20,7 @@ from .errors import (
     RegistrationFailed, TransportError,
 )
 
-__version__ = "1.4.0.dev4"
+__version__ = "1.4.1.dev0"
 
 __all__ = [
     "Repo", "load_instances",

{instance_repo-1.4.0.dev4 → instance_repo-1.4.1.dev0}/instance_repo/cli.py

@@ -451,9 +451,16 @@ def build_parser() -> argparse.ArgumentParser:
 
 
 def _split_ref(ref: str, parts: int) -> list[str]:
+    """切分位置参数 ref:尾部 ``parts-2`` 段是 version/instance-id 槽位,其余(≥1 段)
+    全部是 dataset 名——多段 dataset 名 (``A/B/C``) 因此可以原样出现在 ref 里。
+
+    ``parts=2`` 时 ref 整体就是 dataset 名(无尾部槽位)。
+    """
     segs = ref.split("/")
-    if len(segs) != parts:
-        raise SystemExit(f"ref must have {parts} segments, got {ref!r}")
+    minimum = parts - 1
+    if len(segs) < minimum:
+        raise SystemExit(
+            f"ref must have at least {minimum} segments, got {ref!r}")
     return segs
 
 
@@ -475,15 +482,18 @@ def _split_instance_count(split) -> int:
 
 
 def _resolve_split_ref(args) -> tuple[str, str]:
-    """定位 dataset + split：位置 ref（{L1}/{L2}/{split}）或 --dataset/--split。"""
+    """定位 dataset + split:位置 ref(``{dataset}/{split}``)或 --dataset/--split。
+
+    多段 dataset 名时 split 取最后一节,前面所有段拼回 dataset 名。
+    """
     if args.dataset:
         if not args.split:
             raise SystemExit("给定 --dataset 时 split 需要 --split")
         return args.dataset, args.split
     if args.ref:
         segs = _split_ref(args.ref, 3)
-        return f"{segs[0]}/{segs[1]}", segs[2]
-    raise SystemExit("split get/create/set-run-type 需要 {L1}/{L2}/{split} 或 "
+        return "/".join(segs[:-1]), segs[-1]
+    raise SystemExit("split get/create/set-run-type 需要 {dataset}/{split} 或 "
                      "--dataset/--split")
 
 
@@ -493,15 +503,15 @@ def _split_dataset(dataset: str) -> tuple[str, str]:
         name = P.DatasetName.parse(dataset)
     except InstanceRepoError as exc:
         raise SystemExit(
-            f"--dataset must be one or two canonical segments, got "
+            f"--dataset must be one or more canonical segments, got "
             f"{dataset!r}") from exc
     return name.l1, name.l2
 
 
 _TARGET_SHAPE = {
-    2: ("{L1}/{L2}", "--dataset"),
-    3: ("{L1}/{L2}/{version}", "--dataset --version"),
-    4: ("{L1}/{L2}/{version}/{id}", "--dataset --version --instance-id"),
+    2: ("{dataset}", "--dataset"),
+    3: ("{dataset}/{version}", "--dataset --version"),
+    4: ("{dataset}/{version}/{id}", "--dataset --version --instance-id"),
 }
 
 
@@ -532,9 +542,14 @@ def _resolve_target(args, parts: int) -> tuple[str, str | None, str | None]:
             f"（ref={ref!r}）")
     if ref:
         segs = _split_ref(ref, parts)
-        return ("/".join(segs[:2]),
-                segs[2] if parts >= 3 else None,
-                segs[3] if parts >= 4 else None)
+        # 尾部 parts-2 段是 version/instance-id 槽位,其余全部是 dataset 名
+        # (多段 dataset 名按原样保留)。
+        slots = parts - 2
+        if slots == 0:
+            return ("/".join(segs), None, None)
+        if slots == 1:
+            return ("/".join(segs[:-1]), segs[-1], None)
+        return ("/".join(segs[:-2]), segs[-2], segs[-1])
     if not given:
         raise SystemExit(
             f"需要位置参数 ref（{shape}）或 flag 形式（{flag_hint}）")

{instance_repo-1.4.0.dev4 → instance_repo-1.4.1.dev0}/instance_repo/paths.py

@@ -21,7 +21,7 @@ from dataclasses import dataclass
 from .errors import SchemaError
 
 
-def _dataset_name_parts(value: object) -> tuple[str, str, tuple[str, ...], str]:
+def _dataset_name_parts(value: object) -> tuple[str, str, str, tuple[str, ...], str]:
     if not isinstance(value, str):
         raise SchemaError(f"dataset must be a string, got {type(value).__name__}")
     segments = tuple(value.split("/"))
@@ -34,17 +34,20 @@ def _dataset_name_parts(value: object) -> tuple[str, str, tuple[str, ...], str]:
         or any(unicodedata.category(char) == "Cc" for char in segment)
         for segment in segments
     )
-    if len(segments) not in (1, 2) or invalid_segment:
+    if not segments or invalid_segment:
         raise SchemaError(
-            f"dataset must contain one or two canonical segments, got {value!r}")
+            f"dataset must contain one or more canonical segments, got {value!r}")
     l1 = segments[0]
-    l2 = segments[1] if len(segments) == 2 else ""
-    return l1, l2, segments, segments[-1]
+    # 多段名（3+ 段）里 l2 是 l1 之下的完整后缀路径（含 "/"），保持
+    # "l2 为空 ⇔ 根(L1)" 的不变量与 {l1,l2} 家族索引语义不变。
+    l2 = "/".join(segments[1:])
+    namespace = "/".join(segments[:-1])
+    return l1, l2, namespace, segments, segments[-1]
 
 
 @dataclass(frozen=True)
 class DatasetName:
-    """Canonical one- or two-segment dataset name.
+    """Canonical dataset name: one segment (L1) or more (L1/L2/...).
 
     Direct construction is supported only when every derived field matches
     ``canonical``; otherwise construction raises :class:`SchemaError`.
@@ -53,24 +56,26 @@ class DatasetName:
     canonical: str
     l1: str
     l2: str
+    namespace: str
     segments: tuple[str, ...]
     leaf: str
 
     def __post_init__(self) -> None:
         expected = _dataset_name_parts(self.canonical)
-        actual = (self.l1, self.l2, self.segments, self.leaf)
+        actual = (self.l1, self.l2, self.namespace, self.segments, self.leaf)
         if actual != expected:
             raise SchemaError(
                 "DatasetName fields must match canonical: expected "
-                f"(l1, l2, segments, leaf)={expected!r}, got {actual!r}")
+                f"(l1, l2, namespace, segments, leaf)={expected!r}, got {actual!r}")
 
     @classmethod
     def parse(cls, value: str) -> DatasetName:
-        l1, l2, segments, leaf = _dataset_name_parts(value)
+        l1, l2, namespace, segments, leaf = _dataset_name_parts(value)
         return cls(
             canonical=value,
             l1=l1,
             l2=l2,
+            namespace=namespace,
             segments=segments,
             leaf=leaf,
         )
@@ -82,7 +87,11 @@ class DatasetName:
 
 
 def split_dataset(dataset: str) -> tuple[str, str]:
-    """Return ``(L1, L2)``; single-segment datasets use an empty L2."""
+    """Return ``(L1, L2)``; single-segment datasets use an empty L2.
+
+    For multi-level names (3+ segments) ``L2`` is the full suffix path under
+    ``L1`` (segments joined by ``/``), e.g. ``"A/B/C"`` → ``("A", "B/C")``.
+    """
     name = DatasetName.parse(dataset)
     return name.l1, name.l2
 

{instance_repo-1.4.0.dev4 → instance_repo-1.4.1.dev0}/instance_repo/store/oss_discovery.py

@@ -221,7 +221,7 @@ def parse_dataset_level(
 
     Args:
         content_store: :class:`~instance_repo.store.oss.OssContentStore` instance.
-        dataset: Canonical dataset name with one or two segments.
+        dataset: Canonical dataset name with one or more segments.
         oss_prefix: Root prefix inside the bucket (default ``"swe/datasets"``).
         max_versions: Upper bound on the number of versions to probe.  When
             exceeded the result is truncated and ``truncated=True`` is set.

{instance_repo-1.4.0.dev4 → instance_repo-1.4.1.dev0}/instance_repo.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: instance-repo
-Version: 1.4.0.dev4
+Version: 1.4.1.dev0
 Summary: InstanceRepo SDK — 评测 Instance 统一存储客户端（SDK-only；控制面由 apiserver 承载）
 Requires-Python: >=3.10
 Description-Content-Type: text/markdown

{instance_repo-1.4.0.dev4 → instance_repo-1.4.1.dev0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "instance-repo"
-version = "1.4.0.dev4"
+version = "1.4.1.dev0"
 description = "InstanceRepo SDK — 评测 Instance 统一存储客户端（SDK-only；控制面由 apiserver 承载）"
 requires-python = ">=3.10"
 readme = "README.md"