Add Zarr reader for CLI and library

aidrin/compute/executor.py

@@ -15,7 +15,7 @@
 """
 
 from dataclasses import asdict, is_dataclass
-from typing import Any, Callable, Dict, Optional
+from typing import Any, Callable, Dict, List, Optional
 
 from aidrin.compute import client
 from aidrin.compute.profiles import RemoteTarget
@@ -138,11 +138,16 @@ def summarize_dataset(
         file_path: str,
         file_type: Optional[str] = None,
         max_features: Optional[int] = None,
+        selected_keys: Optional[List[str]] = None,
     ) -> Dict[str, Any]:
-        return self._call(
-            "summarize",
-            {"file_path": file_path, "file_type": file_type, "max_features": max_features},
-        )
+        payload: Dict[str, Any] = {
+            "file_path": file_path,
+            "file_type": file_type,
+            "max_features": max_features,
+        }
+        if selected_keys:
+            payload["selected_keys"] = selected_keys
+        return self._call("summarize", payload)
 
     def run_data_quality(
         self,
@@ -151,17 +156,18 @@ def run_data_quality(
         file_name: Optional[str] = None,
         verbose: bool = False,
         strip_visualizations: bool = True,
+        selected_keys: Optional[List[str]] = None,
     ) -> Dict[str, Any]:
-        return self._call(
-            "data_quality",
-            {
-                "file_path": file_path,
-                "file_type": file_type,
-                "file_name": file_name,
-                "verbose": verbose,
-                "strip_visualizations": strip_visualizations,
-            },
-        )
+        payload: Dict[str, Any] = {
+            "file_path": file_path,
+            "file_type": file_type,
+            "file_name": file_name,
+            "verbose": verbose,
+            "strip_visualizations": strip_visualizations,
+        }
+        if selected_keys:
+            payload["selected_keys"] = selected_keys
+        return self._call("data_quality", payload)
 
     def run_batch_metrics(
         self,

aidrin/file_handling/file_parser.py

@@ -8,13 +8,16 @@
 from aidrin.file_handling.readers.json_reader import jsonReader
 from aidrin.file_handling.readers.npz_reader import npzReader
 from aidrin.file_handling.readers.parquet_reader import parquetReader
+from aidrin.file_handling.readers.zarr_reader import zarrReader
 
 # Notes:
 # To add support for new file types:
 # - Add a new subclass of BaseFileReader with a .read() method
 #       (and optionally .parse(), .filter()) to 'file_readers'.
 # - Register the class in READER_MAP.
 # - Add a display name and extension to SUPPORTED_FILE_TYPES for the front end.
+# - Formats that are CLI/library/Globus-only (e.g. Zarr directories) go in
+#   GLOBUS_FILE_TYPES but not SUPPORTED_FILE_TYPES (local upload dropdown).
 
 # Reader Map. Used to create file type specific parsing
 READER_MAP = {
@@ -24,10 +27,11 @@
     ".json": jsonReader,
     ".h5": hdf5Reader,
     ".parquet": parquetReader,
+    ".zarr": zarrReader,
     # Add additional file types here
 }
 
-# Supported file types. Read on front end to create select features.
+# Supported file types. Read on front end to create local upload select features.
 SUPPORTED_FILE_TYPES = [
     (".csv", "CSV"),
     (".xls, .xlsb, .xlsx, .xlsm", "Excel"),
@@ -39,6 +43,15 @@
     # (file_type,file_type_name)
 ]
 
+# Globus (and other path-based surfaces) may include directory-shaped formats
+# that are not offered in the local browser upload dropdown.
+GLOBUS_FILE_TYPES = SUPPORTED_FILE_TYPES + [
+    (".zarr", "Zarr"),
+]
+
+# File types that accept selected_keys for multi-array / multi-dataset selection.
+_SELECTION_FILE_TYPES = {".h5", ".zarr"}
+
 # logger config
 file_upload_time_log = logging.getLogger("file_upload")
 
@@ -204,6 +217,7 @@ def read_file(file_info, columns=None):
             -file_path: str, relative or absolute path of the file.
             -file_name: str, file name.
             -file_type: str, file format. Passed from front end select value.
+            - optional selected_keys (4th element) for HDF5/Zarr path selection.
     columns: list of str, optional
         When given, only these columns are returned.
     Returns
@@ -250,7 +264,7 @@ def read_file(file_info, columns=None):
 
         # Slow path: parse the source once.
         reader_cls = READER_MAP[file_type]
-        if file_type == ".h5":
+        if file_type in _SELECTION_FILE_TYPES:
             df = reader_cls(
                 file_path, file_upload_time_log, selected_keys=selected_keys
             ).read()

aidrin/file_handling/readers/root_reader.py

@@ -0,0 +1,17 @@
+from aidrin.file_handling.readers.structured import (
+    INVENTORY_UNSUPPORTED,
+    InventoryResult,
+    StructuredFileReader,
+    make_inventory,
+)
+
+
+class rootReader(StructuredFileReader):
+    """ROOT file reader (scaffold). v1 will support TTrees only via uproot."""
+
+    def inventory(self) -> InventoryResult:
+        return make_inventory(INVENTORY_UNSUPPORTED)
+
+    def read(self):
+        self.logger.warning("ROOT read is not implemented yet.")
+        return None

aidrin/file_handling/readers/structured.py

@@ -0,0 +1,81 @@
+"""
+Shared contract for structured file readers (HDF5, Zarr, ROOT).
+
+Structured formats may hold many arrays or tables under paths. Readers classify
+the on-disk layout via ``inventory()`` before ``read()`` loads tabular data.
+
+Inventory ``type`` values
+-------------------------
+empty
+    No readable datasets or arrays.
+single_dataset
+    One dataset; auto-read without a picker.
+multi_dataset
+    Incompatible or grouped layout; caller must select paths.
+legacy
+    Internal auto-read path for compatible multi-array layouts (and HDF5
+    PyTables stores). Never surface this label in API or UI responses.
+unsupported
+    Reader stub or format not implemented yet. Never treat as auto-read.
+"""
+
+from __future__ import annotations
+
+from typing import Any, TypedDict
+
+from aidrin.file_handling.readers.base_reader import BaseFileReader
+
+INVENTORY_EMPTY = "empty"
+INVENTORY_SINGLE = "single_dataset"
+INVENTORY_MULTI = "multi_dataset"
+INVENTORY_LEGACY = "legacy"
+INVENTORY_UNSUPPORTED = "unsupported"
+
+# Types that may be shown to users when driving picker or error messages.
+USER_FACING_INVENTORY_TYPES = frozenset(
+    {INVENTORY_EMPTY, INVENTORY_SINGLE, INVENTORY_MULTI}
+)
+
+
+class DatasetEntry(TypedDict):
+    path: str
+    shape: tuple[int, ...]
+    ndim: int
+    dtype: str
+    size: int
+
+
+class PickerGroup(TypedDict, total=False):
+    name: str
+    paths: list[str]
+    prefix: str
+
+
+class InventoryResult(TypedDict):
+    type: str
+    datasets: list[DatasetEntry]
+    groups: list[dict[str, Any]]
+
+
+def make_inventory(
+    inv_type: str,
+    datasets: list[DatasetEntry] | None = None,
+    groups: list[dict[str, Any]] | None = None,
+) -> InventoryResult:
+    """Build a normalized inventory dict."""
+    return {
+        "type": inv_type,
+        "datasets": datasets or [],
+        "groups": groups or [],
+    }
+
+
+class StructuredFileReader(BaseFileReader):
+    """Base for hierarchical array/tree formats with inventory and metadata hooks."""
+
+    def inventory(self) -> InventoryResult:
+        return make_inventory(INVENTORY_UNSUPPORTED)
+
+    def get_metadata(self) -> dict[str, Any]:
+        """Embedded file metadata for FAIR assessment (``.zattrs``, ROOT headers, etc.)."""
+        return {}