atdata 0.1.2a4__tar.gz → 0.1.3b3__tar.gz

{atdata-0.1.2a4 → atdata-0.1.3b3}/PKG-INFO

@@ -1,13 +1,16 @@
 Metadata-Version: 2.4
 Name: atdata
-Version: 0.1.2a4
+Version: 0.1.3b3
 Summary: A loose federation of distributed, typed datasets
 Author-email: Maxine Levesque <hello@maxine.science>
 License-File: LICENSE
 Requires-Python: >=3.12
+Requires-Dist: fastparquet>=2024.11.0
 Requires-Dist: msgpack>=1.1.2
 Requires-Dist: numpy>=2.3.4
 Requires-Dist: ormsgpack>=1.11.0
+Requires-Dist: pandas>=2.3.3
+Requires-Dist: tqdm>=4.67.1
 Requires-Dist: webdataset>=1.0.2
 Description-Content-Type: text/markdown
 

{atdata-0.1.2a4 → atdata-0.1.3b3}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "atdata"
-version = "0.1.2a4"
+version = "0.1.3b3"
 description = "A loose federation of distributed, typed datasets"
 readme = "README.md"
 authors = [
@@ -8,9 +8,12 @@ authors = [
 ]
 requires-python = ">=3.12"
 dependencies = [
+    "fastparquet>=2024.11.0",
     "msgpack>=1.1.2",
     "numpy>=2.3.4",
     "ormsgpack>=1.11.0",
+    "pandas>=2.3.3",
+    "tqdm>=4.67.1",
     "webdataset>=1.0.2",
 ]
 

{atdata-0.1.2a4 → atdata-0.1.3b3}/src/atdata/__init__.py

@@ -10,5 +10,11 @@ from .dataset import (
     packable,
 )
 
+from .lens import (
+    Lens,
+    LensNetwork,
+    lens,
+)
+
 
 #

{atdata-0.1.2a4 → atdata-0.1.3b3}/src/atdata/dataset.py

@@ -5,21 +5,34 @@
 
 import webdataset as wds
 
-import functools
-from dataclasses import dataclass
+from pathlib import Path
 import uuid
+import functools
 
-import numpy as np
-
+import dataclasses
+import types
+from dataclasses import (
+    dataclass,
+    asdict,
+)
 from abc import (
     ABC,
     abstractmethod,
 )
+
+from tqdm import tqdm
+import numpy as np
+import pandas as pd
+
+import typing
 from typing import (
     Any,
     Optional,
     Dict,
     Sequence,
+    Iterable,
+    Callable,
+    Union,
     #
     Self,
     Generic,
@@ -40,14 +53,20 @@ from numpy.typing import (
 import msgpack
 import ormsgpack
 from . import _helpers as eh
+from .lens import Lens, LensNetwork
 
 
 ##
 # Typing help
 
+Pathlike = str | Path
+
 WDSRawSample: TypeAlias = Dict[str, Any]
 WDSRawBatch: TypeAlias = Dict[str, Any]
 
+SampleExportRow: TypeAlias = Dict[str, Any]
+SampleExportMap: TypeAlias = Callable[['PackableSample'], SampleExportRow]
+
 
 ##
 # Main base classes
@@ -94,6 +113,25 @@ def _make_packable( x ):
         return eh.array_to_bytes( x )
     return x
 
+def _is_possibly_ndarray_type( t ):
+    """Checks if a type annotation is possibly an NDArray."""
+
+    # Directly an NDArray
+    if t == NDArray:
+        # print( 'is an NDArray' )
+        return True
+    
+    # Check for Optionals (i.e., NDArray | None)
+    if isinstance( t, types.UnionType ):
+        t_parts = t.__args__
+        if any( x == NDArray
+                for x in t_parts ):
+            return True
+    
+    # Not an NDArray
+    return False
+
+@dataclass
 class PackableSample( ABC ):
     """A sample that can be packed and unpacked with msgpack"""
 
@@ -101,10 +139,13 @@ class PackableSample( ABC ):
         """TODO Stupid kludge because of __post_init__ nonsense for wrapped classes"""
 
         # Auto-convert known types when annotated
-        for var_name, var_type in vars( self.__class__ )['__annotations__'].items():
+        # for var_name, var_type in vars( self.__class__ )['__annotations__'].items():
+        for field in dataclasses.fields( self ):
+            var_name = field.name
+            var_type = field.type
 
             # Annotation for this variable is to be an NDArray
-            if var_type == NDArray:
+            if _is_possibly_ndarray_type( var_type ):
                 # ... so, we'll always auto-convert to numpy
 
                 var_cur_value = getattr( self, var_name )
@@ -120,6 +161,9 @@ class PackableSample( ABC ):
                 #     setattr( self, var_name, var_cur_value.to_numpy )
 
                 elif isinstance( var_cur_value, bytes ):
+                    # TODO This does create a constraint that serialized bytes
+                    # in a field that might be an NDArray are always interpreted
+                    # as being the NDArray interpretation
                     setattr( self, var_name, eh.bytes_to_array( var_cur_value ) )
 
     def __post_init__( self ):
@@ -189,7 +233,7 @@ class SampleBatch( Generic[DT] ):
     @property
     def sample_type( self ) -> Type:
         """The type of each sample in this batch"""
-        return self.__orig_class__.__args__[0]
+        return typing.get_args( self.__orig_class__)[0]
 
     def __getattr__( self, name ):
         # Aggregate named params of sample type
@@ -217,6 +261,8 @@ class SampleBatch( Generic[DT] ):
 ST = TypeVar( 'ST', bound = PackableSample )
 # BT = TypeVar( 'BT' )
 
+RT = TypeVar( 'RT', bound = PackableSample )
+
 # TODO For python 3.13
 # BT = TypeVar( 'BT', default = None )
 # IT = TypeVar( 'IT', default = Any )
@@ -235,7 +281,8 @@ class Dataset( Generic[ST] ):
     @property
     def sample_type( self ) -> Type:
         """The type of each returned sample from this `Dataset`'s iterator"""
-        return self.__orig_class__.__args__[0]
+        # TODO Figure out why linting fails here
+        return typing.get_args( self.__orig_class__ )[0]
     @property
     def batch_type( self ) -> Type:
         """The type of a batch built from `sample_class`"""
@@ -253,6 +300,17 @@ class Dataset( Generic[ST] ):
         super().__init__()
         self.url = url
 
+        # Allow addition of automatic transformation of raw underlying data
+        self._output_lens: Lens | None = None
+
+    def as_type( self, other: Type[RT] ) -> 'Dataset[RT]':
+        """TODO"""
+        ret = Dataset[other]( self.url )
+        # Get the singleton lens registry
+        lenses = LensNetwork()
+        ret._output_lens = lenses.transform( self.sample_type, ret.sample_type )
+        return ret
+
     # @classmethod
     # def register( cls, uri: str,
     #             sample_class: Type,
@@ -278,15 +336,15 @@ class Dataset( Generic[ST] ):
             A full (non-lazy) list of the individual ``tar`` files within the
             source WebDataset.
         """
-        pipe = wds.DataPipeline(
-            wds.SimpleShardList( self.url ),
-            wds.map( lambda x: x['url'] )
+        pipe = wds.pipeline.DataPipeline(
+            wds.shardlists.SimpleShardList( self.url ),
+            wds.filters.map( lambda x: x['url'] )
         )
         return list( pipe )
     
     def ordered( self,
                 batch_size: int | None = 1,
-            ) -> wds.DataPipeline:
+            ) -> Iterable[ST]:
         """Iterate over the dataset in order
         
         Args:
@@ -302,30 +360,30 @@ class Dataset( Generic[ST] ):
 
         if batch_size is None:
             # TODO Duplication here
-            return wds.DataPipeline(
-                wds.SimpleShardList( self.url ),
-                wds.split_by_worker,
+            return wds.pipeline.DataPipeline(
+                wds.shardlists.SimpleShardList( self.url ),
+                wds.shardlists.split_by_worker,
                 #
-                wds.tarfile_to_samples(),
+                wds.tariterators.tarfile_to_samples(),
                 # wds.map( self.preprocess ),
-                wds.map( self.wrap ),
+                wds.filters.map( self.wrap ),
             )
 
-        return wds.DataPipeline(
-            wds.SimpleShardList( self.url ),
-            wds.split_by_worker,
+        return wds.pipeline.DataPipeline(
+            wds.shardlists.SimpleShardList( self.url ),
+            wds.shardlists.split_by_worker,
             #
-            wds.tarfile_to_samples(),
+            wds.tariterators.tarfile_to_samples(),
             # wds.map( self.preprocess ),
-            wds.batched( batch_size ),
-            wds.map( self.wrap_batch ),
+            wds.filters.batched( batch_size ),
+            wds.filters.map( self.wrap_batch ),
         )
 
     def shuffled( self,
                 buffer_shards: int = 100,
                 buffer_samples: int = 10_000,
                 batch_size: int | None = 1,
-            ) -> wds.DataPipeline:
+            ) -> Iterable[ST]:
         """Iterate over the dataset in random order
         
         Args:
@@ -342,30 +400,88 @@ class Dataset( Generic[ST] ):
 
         if batch_size is None:
             # TODO Duplication here
-            return wds.DataPipeline(
-                wds.SimpleShardList( self.url ),
-                wds.shuffle( buffer_shards ),
-                wds.split_by_worker,
+            return wds.pipeline.DataPipeline(
+                wds.shardlists.SimpleShardList( self.url ),
+                wds.filters.shuffle( buffer_shards ),
+                wds.shardlists.split_by_worker,
                 #
-                wds.tarfile_to_samples(),
+                wds.tariterators.tarfile_to_samples(),
                 # wds.shuffle( buffer_samples ),
                 # wds.map( self.preprocess ),
-                wds.shuffle( buffer_samples ),
-                wds.map( self.wrap ),
+                wds.filters.shuffle( buffer_samples ),
+                wds.filters.map( self.wrap ),
             )
 
-        return wds.DataPipeline(
-            wds.SimpleShardList( self.url ),
-            wds.shuffle( buffer_shards ),
-            wds.split_by_worker,
+        return wds.pipeline.DataPipeline(
+            wds.shardlists.SimpleShardList( self.url ),
+            wds.filters.shuffle( buffer_shards ),
+            wds.shardlists.split_by_worker,
             #
-            wds.tarfile_to_samples(),
+            wds.tariterators.tarfile_to_samples(),
             # wds.shuffle( buffer_samples ),
             # wds.map( self.preprocess ),
-            wds.shuffle( buffer_samples ),
-            wds.batched( batch_size ),
-            wds.map( self.wrap_batch ),
+            wds.filters.shuffle( buffer_samples ),
+            wds.filters.batched( batch_size ),
+            wds.filters.map( self.wrap_batch ),
         )
+    
+    # TODO Rewrite to eliminate `pandas` dependency directly calling
+    # `fastparquet`
+    def to_parquet( self, path: Pathlike,
+                sample_map: Optional[SampleExportMap] = None,
+                maxcount: Optional[int] = None,
+                **kwargs,
+            ):
+        """Save dataset contents to a `parquet` file at `path`
+
+        `kwargs` sent to `pandas.to_parquet`
+        """
+        ##
+
+        # Normalize args
+        path = Path( path )
+        if sample_map is None:
+            sample_map = asdict
+        
+        verbose = kwargs.get( 'verbose', False )
+
+        it = self.ordered( batch_size = None )
+        if verbose:
+            it = tqdm( it )
+
+        #
+
+        if maxcount is None:
+            # Load and save full dataset
+            df = pd.DataFrame( [ sample_map( x )
+                                 for x in self.ordered( batch_size = None ) ] )
+            df.to_parquet( path, **kwargs )
+        
+        else:
+            # Load and save dataset in segments of size `maxcount`
+
+            cur_segment = 0
+            cur_buffer = []
+            path_template = (path.parent / f'{path.stem}-%06d.{path.suffix}').as_posix()
+
+            for x in self.ordered( batch_size = None ):
+                cur_buffer.append( sample_map( x ) )
+                
+                if len( cur_buffer ) >= maxcount:
+                    # Write current segment
+                    cur_path = path_template.format( cur_segment )
+                    df = pd.DataFrame( cur_buffer )
+                    df.to_parquet( cur_path, **kwargs )
+
+                    cur_segment += 1
+                    cur_buffer = []
+                
+            if len( cur_buffer ) > 0:
+                # Write one last segment with remainder
+                cur_path = path_template.format( cur_segment )
+                df = pd.DataFrame( cur_buffer )
+                df.to_parquet( cur_path, **kwargs )
+
 
     # Implemented by specific subclasses
 
@@ -388,20 +504,24 @@ class Dataset( Generic[ST] ):
         assert 'msgpack' in sample
         assert type( sample['msgpack'] ) == bytes
         
-        return self.sample_type.from_bytes( sample['msgpack'] )
+        if self._output_lens is None:
+            return self.sample_type.from_bytes( sample['msgpack'] )
+
+        source_sample = self._output_lens.source_type.from_bytes( sample['msgpack'] )
+        return self._output_lens( source_sample )
     
-        try:
-            assert type( sample ) == dict
-            return cls.sample_class( **{
-                k: v
-                for k, v in sample.items() if k != '__key__'
-            } )
+        # try:
+        #     assert type( sample ) == dict
+        #     return cls.sample_class( **{
+        #         k: v
+        #         for k, v in sample.items() if k != '__key__'
+        #     } )
         
-        except Exception as e:
-            # Sample constructor failed -- revert to default
-            return AnySample(
-                value = sample,
-            )
+        # except Exception as e:
+        #     # Sample constructor failed -- revert to default
+        #     return AnySample(
+        #         value = sample,
+        #     )
 
     def wrap_batch( self, batch: WDSRawBatch ) -> SampleBatch[ST]:
         """Wrap a `batch` of samples into the appropriate dataset-specific type
@@ -410,10 +530,17 @@ class Dataset( Generic[ST] ):
         """
 
         assert 'msgpack' in batch
-        batch_unpacked = [ self.sample_type.from_bytes( bs )
-                           for bs in batch['msgpack'] ]
-        return SampleBatch[self.sample_type]( batch_unpacked )
 
+        if self._output_lens is None:
+            batch_unpacked = [ self.sample_type.from_bytes( bs )
+                               for bs in batch['msgpack'] ]
+            return SampleBatch[self.sample_type]( batch_unpacked )
+
+        batch_source = [ self._output_lens.source_type.from_bytes( bs )
+                         for bs in batch['msgpack'] ]
+        batch_view = [ self._output_lens( s )
+                       for s in batch_source ]
+        return SampleBatch[self.sample_type]( batch_view )
 
     # # @classmethod
     # def wrap_batch( self, batch: WDSRawBatch ) -> BT:
@@ -449,6 +576,9 @@ def packable( cls ):
     
     ##
 
+    class_name = cls.__name__
+    class_annotations = cls.__annotations__
+
     # Add in dataclass niceness to original class
     as_dataclass = dataclass( cls )
 
@@ -458,8 +588,9 @@ def packable( cls ):
         def __post_init__( self ):
             return PackableSample.__post_init__( self )
     
-    as_packable.__name__ = cls.__name__
-    as_packable.__annotations__ = cls.__annotations__
+    # TODO This doesn't properly carry over the original
+    as_packable.__name__ = class_name
+    as_packable.__annotations__ = class_annotations
 
     ##
 

atdata-0.1.3b3/src/atdata/lens.py

@@ -0,0 +1,200 @@
+"""Lenses between typed datasets"""
+
+##
+# Imports
+
+import functools
+import inspect
+
+from typing import (
+    TypeAlias,
+    Type,
+    TypeVar,
+    Tuple,
+    Dict,
+    Callable,
+    Optional,
+    Generic,
+    #
+    TYPE_CHECKING
+)
+
+if TYPE_CHECKING:
+    from .dataset import PackableSample
+
+
+##
+# Typing helpers
+
+DatasetType: TypeAlias = Type['PackableSample']
+LensSignature: TypeAlias = Tuple[DatasetType, DatasetType]
+
+S = TypeVar( 'S', bound = 'PackableSample' )
+V = TypeVar( 'V', bound = 'PackableSample' )
+type LensGetter[S, V] = Callable[[S], V]
+type LensPutter[S, V] = Callable[[V, S], S]
+
+
+##
+# Shortcut decorators
+
+class Lens( Generic[S, V] ):
+    """TODO"""
+
+    # @property
+    # def source_type( self ) -> Type[S]:
+    #     """The source type (S) for the lens; what is put to"""
+    #     # TODO Figure out why linting fails here
+    #     return self.__orig_class__.__args__[0]
+
+    # @property
+    # def view_type( self ) -> Type[V]:
+    #     """The view type (V) for the lens; what is get'd from"""
+    #     # TODO FIgure out why linting fails here
+    #     return self.__orig_class__.__args__[1]
+
+    def __init__( self, get: LensGetter[S, V],
+                put: Optional[LensPutter[S, V]] = None
+            ) -> None:
+        """TODO"""
+        ##
+
+        # Check argument validity
+
+        sig = inspect.signature( get )
+        input_types = list( sig.parameters.values() )
+        assert len( input_types ) == 1, \
+            'Wrong number of input args for lens: should only have one'
+
+        # Update function details for this object as returned by annotation
+        functools.update_wrapper( self, get )
+
+        self.source_type: Type[PackableSample] = input_types[0].annotation
+        self.view_type = sig.return_annotation
+
+        # Store the getter
+        self._getter = get
+        
+        # Determine and store the putter
+        if put is None:
+            # Trivial putter does not update the source
+            def _trivial_put( v: V, s: S ) -> S:
+                return s
+            put = _trivial_put
+        self._putter = put
+    
+    #
+
+    def putter( self, put: LensPutter[S, V] ) -> LensPutter[S, V]:
+        """TODO"""
+        ##
+        self._putter = put
+        return put
+    
+    # Methods to actually execute transformations
+
+    def put( self, v: V, s: S ) -> S:
+        """TODO"""
+        return self._putter( v, s )
+
+    def get( self, s: S ) -> V:
+        """TODO"""
+        return self( s )
+
+    # Convenience to enable calling the lens as its getter
+    
+    def __call__( self, s: S ) -> V:
+        return self._getter( s )
+
+# TODO Figure out how to properly parameterize this
+# def _lens_factory[S, V]( register: bool = True ):
+#     """Register the annotated function `f` as the getter of a sample lens"""
+
+#     # The actual lens decorator taking a lens getter function to a lens object
+#     def _decorator( f: LensGetter[S, V] ) -> Lens[S, V]:
+#         ret = Lens[S, V]( f )
+#         if register:
+#             _network.register( ret )
+#         return ret
+    
+#     # Return the lens decorator
+#     return _decorator
+
+# # For convenience
+# lens = _lens_factory
+
+def lens(  f: LensGetter[S, V] ) -> Lens[S, V]:
+    ret = Lens[S, V]( f )
+    _network.register( ret )
+    return ret
+
+
+##
+# Global registry of used lenses
+
+# _registered_lenses: Dict[LensSignature, Lens] = dict()
+# """TODO"""
+
+class LensNetwork:
+    """TODO"""
+
+    _instance = None
+    """The singleton instance"""
+
+    def __new__(cls, *args, **kwargs):
+        if cls._instance is None:
+            # If no instance exists, create a new one
+            cls._instance = super().__new__(cls)
+        return cls._instance  # Return the existing (or newly created) instance
+
+    def __init__(self):
+        if not hasattr(self, '_initialized'):  # Check if already initialized
+            self._registry: Dict[LensSignature, Lens] = dict()
+            self._initialized = True
+    
+    def register( self, _lens: Lens ):
+        """Set `lens` as the canonical view between its source and view types"""
+    
+        # sig = inspect.signature( _lens.get )
+        # input_types = list( sig.parameters.values() )
+        # assert len( input_types ) == 1, \
+        #     'Wrong number of input args for lens: should only have one'
+        
+        # input_type = input_types[0].annotation
+        # print( input_type )
+        # output_type = sig.return_annotation
+
+        # self._registry[input_type, output_type] = _lens
+        # print( _lens.source_type )
+        self._registry[_lens.source_type, _lens.view_type] = _lens
+    
+    def transform( self, source: DatasetType, view: DatasetType ) -> Lens:
+        """TODO"""
+
+        # TODO Handle compositional closure
+        ret = self._registry.get( (source, view), None )
+        if ret is None:
+            raise ValueError( f'No registered lens from source {source} to view {view}' )
+        
+        return ret
+
+
+# Create global singleton registry instance
+_network = LensNetwork()
+
+# def lens( f: LensPutter ) -> Lens:
+#     """Register the annotated function `f` as a sample lens"""
+#     ##
+    
+#     sig = inspect.signature( f )
+
+#     input_types = list( sig.parameters.values() )
+#     output_type = sig.return_annotation
+    
+#     _registered_lenses[]
+
+#     f.lens = Lens(
+
+#     )
+
+#     return f

{atdata-0.1.2a4 → atdata-0.1.3b3}/tests/test_dataset.py

@@ -50,6 +50,12 @@ class NumpyTestSampleDecorated:
     label: int
     image: NDArray
 
+@atdata.packable
+class NumpyOptionalSampleDecorated:
+    label: int
+    image: NDArray
+    embeddings: NDArray | None = None
+
 test_cases = [
     {
         'SampleType': BasicTestSample,
@@ -59,6 +65,7 @@ test_cases = [
             'value': 1024.768,
         },
         'sample_wds_stem': 'basic_test',
+        'test_parquet': True,
     },
     {
         'SampleType': NumpyTestSample,
@@ -68,6 +75,7 @@ test_cases = [
             'image': np.random.randn( 1024, 1024 ),
         },
         'sample_wds_stem': 'numpy_test',
+        'test_parquet': False,
     },
     {
         'SampleType': BasicTestSampleDecorated,
@@ -77,6 +85,7 @@ test_cases = [
             'value': 1024.768,
         },
         'sample_wds_stem': 'basic_test_decorated',
+        'test_parquet': True,
     },
     {
         'SampleType': NumpyTestSampleDecorated,
@@ -86,6 +95,29 @@ test_cases = [
             'image': np.random.randn( 1024, 1024 ),
         },
         'sample_wds_stem': 'numpy_test_decorated',
+        'test_parquet': False,
+    },
+    {
+        'SampleType': NumpyOptionalSampleDecorated,
+        'sample_data':
+        {
+            'label': 9_001,
+            'image': np.random.randn( 1024, 1024 ),
+            'embeddings': np.random.randn( 512 ),
+        },
+        'sample_wds_stem': 'numpy_optional_decorated',
+        'test_parquet': False,
+    },
+    {
+        'SampleType': NumpyOptionalSampleDecorated,
+        'sample_data':
+        {
+            'label': 9_001,
+            'image': np.random.randn( 1024, 1024 ),
+            'embeddings': None,
+        },
+        'sample_wds_stem': 'numpy_optional_decorated_none',
+        'test_parquet': False,
     },
 ]
 
@@ -175,7 +207,7 @@ def test_wds(
     ).as_posix()
     file_wds_pattern = file_pattern.format( shard_id = '%06d' )
 
-    with wds.ShardWriter(
+    with wds.writer.ShardWriter(
         pattern = file_wds_pattern,
         maxcount = shard_maxcount,
     ) as sink:
@@ -323,5 +355,56 @@ def test_wds(
     assert iterations_run == n_iterate, \
         "Only found {iterations_run} samples, not {n_iterate}"
 
+#
+
+@pytest.mark.parametrize(
+    ('SampleType', 'sample_data', 'sample_wds_stem', 'test_parquet'),
+    [ (
+        case['SampleType'],
+        case['sample_data'],
+        case['sample_wds_stem'],
+        case['test_parquet']
+      )
+      for case in test_cases ]
+)
+def test_parquet_export(
+            SampleType: Type[atdata.PackableSample],
+            sample_data: atds.MsgpackRawSample,
+            sample_wds_stem: str,
+            test_parquet: bool,
+            tmp_path
+        ):
+    """Test our ability to export a dataset to `parquet` format"""
+
+    # Skip irrelevant test cases
+    if not test_parquet:
+        return
+
+    ## Testing hyperparameters
+
+    n_copies_dataset = 1_000
+    n_per_file = 100
+
+    ## Start out by writing tar dataset
+
+    wds_filename = (tmp_path / f'{sample_wds_stem}.tar').as_posix()
+    with wds.writer.TarWriter( wds_filename ) as sink:
+        for _ in range( n_copies_dataset ):
+            new_sample = SampleType.from_data( sample_data )
+            sink.write( new_sample.as_wds )
+    
+    ## Now export to `parquet`
+
+    dataset = atdata.Dataset[SampleType]( wds_filename )
+    parquet_filename = tmp_path / f'{sample_wds_stem}.parquet'
+    dataset.to_parquet( parquet_filename )
+
+    parquet_filename = tmp_path / f'{sample_wds_stem}-segments.parquet'
+    dataset.to_parquet( parquet_filename, maxcount = n_per_file )
+
+    ## Double-check our `parquet` export
+    
+    # TODO
+
 
 ##

atdata-0.1.3b3/tests/test_lens.py

@@ -0,0 +1,166 @@
+"""Test lens functionality."""
+
+##
+# Imports
+
+import pytest
+
+from dataclasses import dataclass
+import webdataset as wds
+import atdata
+
+import numpy as np
+from numpy.typing import NDArray
+
+
+##
+# Tests
+
+def test_lens():
+    """Test a lens between sample types"""
+
+    # Set up the lens scenario
+
+    @atdata.packable
+    class Source:
+        name: str
+        age: int
+        height: float
+
+    @atdata.packable
+    class View:
+        name: str
+        height: float
+    
+    @atdata.lens
+    def polite( s: Source ) -> View:
+        return View(
+            name = s.name,
+            height = s.height,
+        )
+    
+    @polite.putter
+    def polite_update( v: View, s: Source ) -> Source:
+        return Source(
+            name = v.name,
+            height = v.height,
+            #
+            age = s.age,
+        )
+    
+    # Test with an example sample
+
+    test_source = Source(
+        name = 'Hello World',
+        age = 42,
+        height = 182.9,
+    )
+    correct_view = View(
+        name = test_source.name,
+        height = test_source.height,
+    )
+
+    test_view = polite( test_source )
+    assert test_view == correct_view, \
+        f'Incorrect lens behavior: {test_view}, and not {correct_view}'
+
+    # This lens should be well-behaved
+
+    update_view = View(
+        name = 'Now Taller',
+        height = 192.9,
+    )
+
+    x = polite( polite.put( update_view, test_source ) )
+    assert x == update_view, \
+        f'Violation of GetPut: {x} =/= {update_view}'
+    
+    y = polite.put( polite( test_source ), test_source )
+    assert y == test_source, \
+        f'Violation of PutGet: {y} =/= {test_source}'
+    
+    # TODO Test PutPut
+
+def test_conversion( tmp_path ):
+    """Test automatic interconversion between sample types"""
+
+    @dataclass
+    class Source( atdata.PackableSample ):
+        name: str
+        height: float
+        favorite_pizza: str
+        favorite_image: NDArray
+    
+    @dataclass
+    class View( atdata.PackableSample ):
+        name: str
+        favorite_pizza: str
+        favorite_image: NDArray
+    
+    @atdata.lens
+    def polite( s: Source ) -> View:
+        return View(
+            name = s.name,
+            favorite_pizza = s.favorite_pizza,
+            favorite_image = s.favorite_image,
+        )
+    
+    lens_network = atdata.LensNetwork()
+    print( lens_network._registry )
+    
+    # Map a test sample through the view
+    test_source = Source(
+        name = 'Larry',
+        height = 42.,
+        favorite_pizza = 'pineapple',
+        favorite_image = np.random.randn( 224, 224 )
+    )
+    test_view = polite( test_source )
+
+    # Create a test dataset
+
+    k_test = 100
+    test_filename = (
+        tmp_path
+        / 'test-source.tar'
+    ).as_posix()
+
+    with wds.writer.TarWriter( test_filename ) as dest:
+        for i in range( k_test ):
+            # Create a new copied sample
+            cur_sample = Source(
+                name = test_source.name,
+                height = test_source.height,
+                favorite_pizza = test_source.favorite_pizza,
+                favorite_image = test_source.favorite_image,
+            )
+            dest.write( cur_sample.as_wds )
+    
+    # Try reading the test dataset
+
+    ds = (
+        atdata.Dataset[Source]( test_filename )
+            .as_type( View )
+    )
+
+    assert ds.sample_type == View, \
+        'Auto-mapped'
+
+    sample: View | None = None
+    for sample in ds.ordered( batch_size = None ):
+        # Load only the first sample
+        break
+
+    assert sample is not None, \
+        'Did not load any samples from `Source` dataset'
+
+    assert sample.name == test_view.name, \
+        f'Divergence on auto-mapped dataset: `name` should be {test_view.name}, but is {sample.name}'
+    # assert sample.height == test_view.height, \
+    #     f'Divergence on auto-mapped dataset: `height` should be {test_view.height}, but is {sample.height}'
+    assert sample.favorite_pizza == test_view.favorite_pizza, \
+        f'Divergence on auto-mapped dataset: `favorite_pizza` should be {test_view.favorite_pizza}, but is {sample.favorite_pizza}'
+    assert np.all( sample.favorite_image == test_view.favorite_image ), \
+        f'Divergence on auto-mapped dataset: `favorite_image`'
+
+##