datachain 0.5.1__py3-none-any.whl → 0.6.1__py3-none-any.whl

datachain/lib/udf.py

@@ -1,14 +1,15 @@
 import sys
 import traceback
 from collections.abc import Iterable, Iterator, Mapping, Sequence
-from dataclasses import dataclass
 from typing import TYPE_CHECKING, Any, Callable, Optional
 
+import attrs
 from fsspec.callbacks import DEFAULT_CALLBACK, Callback
 from pydantic import BaseModel
 
 from datachain.dataset import RowDict
 from datachain.lib.convert.flatten import flatten
+from datachain.lib.data_model import DataValue
 from datachain.lib.file import File
 from datachain.lib.signal_schema import SignalSchema
 from datachain.lib.utils import AbstractUDF, DataChainError, DataChainParamsError
@@ -18,16 +19,14 @@ from datachain.query.batch import (
     NoBatching,
     Partition,
     RowsOutputBatch,
-    UDFInputBatch,
 )
-from datachain.query.schema import ColumnParameter, UDFParameter
 
 if TYPE_CHECKING:
     from typing_extensions import Self
 
     from datachain.catalog import Catalog
     from datachain.lib.udf_signature import UdfSignature
-    from datachain.query.batch import RowsOutput, UDFInput
+    from datachain.query.batch import RowsOutput
 
 
 class UdfError(DataChainParamsError):
@@ -45,11 +44,21 @@ UDFOutputSpec = Mapping[str, ColumnType]
 UDFResult = dict[str, Any]
 
 
-@dataclass
+@attrs.define
 class UDFProperties:
-    """Container for basic UDF properties."""
+    udf: "UDFAdapter"
 
-    params: list[UDFParameter]
+    def get_batching(self, use_partitioning: bool = False) -> BatchingStrategy:
+        return self.udf.get_batching(use_partitioning)
+
+    @property
+    def batch(self):
+        return self.udf.batch
+
+
+@attrs.define(slots=False)
+class UDFAdapter:
+    inner: "UDFBase"
     output: UDFOutputSpec
     batch: int = 1
 
@@ -62,20 +71,10 @@ class UDFProperties:
             return Batch(self.batch)
         raise ValueError(f"invalid batch size {self.batch}")
 
-    def signal_names(self) -> Iterable[str]:
-        return self.output.keys()
-
-
-class UDFAdapter:
-    def __init__(
-        self,
-        inner: "UDFBase",
-        properties: UDFProperties,
-    ):
-        self.inner = inner
-        self.properties = properties
-        self.signal_names = properties.signal_names()
-        self.output = properties.output
+    @property
+    def properties(self):
+        # For backwards compatibility.
+        return UDFProperties(self)
 
     def run(
         self,
@@ -87,72 +86,14 @@ class UDFAdapter:
         download_cb: Callback = DEFAULT_CALLBACK,
         processed_cb: Callback = DEFAULT_CALLBACK,
     ) -> Iterator[Iterable[UDFResult]]:
-        self.inner.catalog = catalog
-        if hasattr(self.inner, "setup") and callable(self.inner.setup):
-            self.inner.setup()
-
-        for batch in udf_inputs:
-            if isinstance(batch, RowsOutputBatch):
-                n_rows = len(batch.rows)
-                inputs: UDFInput = UDFInputBatch(
-                    [RowDict(zip(udf_fields, row)) for row in batch.rows]
-                )
-            else:
-                n_rows = 1
-                inputs = RowDict(zip(udf_fields, batch))
-            output = self.run_once(catalog, inputs, is_generator, cache, cb=download_cb)
-            processed_cb.relative_update(n_rows)
-            yield output
-
-        if hasattr(self.inner, "teardown") and callable(self.inner.teardown):
-            self.inner.teardown()
-
-    def run_once(
-        self,
-        catalog: "Catalog",
-        arg: "UDFInput",
-        is_generator: bool = False,
-        cache: bool = False,
-        cb: Callback = DEFAULT_CALLBACK,
-    ) -> Iterable[UDFResult]:
-        if isinstance(arg, UDFInputBatch):
-            udf_inputs = [
-                self.bind_parameters(catalog, row, cache=cache, cb=cb)
-                for row in arg.rows
-            ]
-            udf_outputs = self.inner.run_once(udf_inputs, cache=cache, download_cb=cb)
-            return self._process_results(arg.rows, udf_outputs, is_generator)
-        if isinstance(arg, RowDict):
-            udf_inputs = self.bind_parameters(catalog, arg, cache=cache, cb=cb)
-            udf_outputs = self.inner.run_once(udf_inputs, cache=cache, download_cb=cb)
-            if not is_generator:
-                # udf_outputs is generator already if is_generator=True
-                udf_outputs = [udf_outputs]
-            return self._process_results([arg], udf_outputs, is_generator)
-        raise ValueError(f"Unexpected UDF argument: {arg}")
-
-    def bind_parameters(self, catalog: "Catalog", row: "RowDict", **kwargs) -> list:
-        return [p.get_value(catalog, row, **kwargs) for p in self.properties.params]
-
-    def _process_results(
-        self,
-        rows: Sequence["RowDict"],
-        results: Sequence[Sequence[Any]],
-        is_generator=False,
-    ) -> Iterable[UDFResult]:
-        """Create a list of dictionaries representing UDF results."""
-
-        # outputting rows
-        if is_generator:
-            # each row in results is a tuple of column values
-            return (dict(zip(self.signal_names, row)) for row in results)
-
-        # outputting signals
-        row_ids = [row["sys__id"] for row in rows]
-        return [
-            {"sys__id": row_id} | dict(zip(self.signal_names, signals))
-            for row_id, signals in zip(row_ids, results)
-        ]
+        yield from self.inner.run(
+            udf_fields,
+            udf_inputs,
+            catalog,
+            cache,
+            download_cb,
+            processed_cb,
+        )
 
 
 class UDFBase(AbstractUDF):
@@ -203,17 +144,12 @@ class UDFBase(AbstractUDF):
         ```
     """
 
-    is_input_batched = False
     is_output_batched = False
-    is_input_grouped = False
-    params_spec: Optional[list[str]]
     catalog: "Optional[Catalog]"
 
     def __init__(self):
-        self.params = None
+        self.params: Optional[SignalSchema] = None
         self.output = None
-        self.params_spec = None
-        self.output_spec = None
         self.catalog = None
         self._func = None
 
@@ -241,11 +177,6 @@ class UDFBase(AbstractUDF):
     ):
         self.params = params
         self.output = sign.output_schema
-
-        params_spec = self.params.to_udf_spec()
-        self.params_spec = list(params_spec.keys())
-        self.output_spec = self.output.to_udf_spec()
-
         self._func = func
 
     @classmethod
@@ -273,48 +204,27 @@ class UDFBase(AbstractUDF):
     def name(self):
         return self.__class__.__name__
 
+    @property
+    def signal_names(self) -> Iterable[str]:
+        return self.output.to_udf_spec().keys()
+
     def to_udf_wrapper(self, batch: int = 1) -> UDFAdapter:
-        assert self.params_spec is not None
-        properties = UDFProperties(
-            [ColumnParameter(p) for p in self.params_spec], self.output_spec, batch
+        return UDFAdapter(
+            self,
+            self.output.to_udf_spec(),
+            batch,
         )
-        return UDFAdapter(self, properties)
-
-    def validate_results(self, results, *args, **kwargs):
-        return results
 
-    def run_once(self, rows, cache, download_cb):
-        if self.is_input_batched:
-            objs = zip(*self._parse_rows(rows, cache, download_cb))
-        else:
-            objs = self._parse_rows([rows], cache, download_cb)[0]
-
-        result_objs = self.process_safe(objs)
-
-        if not self.is_output_batched:
-            result_objs = [result_objs]
-
-        # Generator expression is required, otherwise the value will be materialized
-        res = (self._flatten_row(row) for row in result_objs)
-
-        if not self.is_output_batched:
-            res = list(res)
-            assert (
-                len(res) == 1
-            ), f"{self.name} returns {len(res)} rows while it's not batched"
-            if isinstance(res[0], tuple):
-                res = res[0]
-        elif (
-            self.is_input_batched
-            and self.is_output_batched
-            and not self.is_input_grouped
-        ):
-            res = list(res)
-            assert len(res) == len(
-                rows
-            ), f"{self.name} returns {len(res)} rows while {len(rows)} expected"
-
-        return res
+    def run(
+        self,
+        udf_fields: "Sequence[str]",
+        udf_inputs: "Iterable[Any]",
+        catalog: "Catalog",
+        cache: bool,
+        download_cb: Callback = DEFAULT_CALLBACK,
+        processed_cb: Callback = DEFAULT_CALLBACK,
+    ) -> Iterator[Iterable[UDFResult]]:
+        raise NotImplementedError
 
     def _flatten_row(self, row):
         if len(self.output.values) > 1 and not isinstance(row, BaseModel):
@@ -328,17 +238,28 @@ class UDFBase(AbstractUDF):
     def _obj_to_list(obj):
         return flatten(obj) if isinstance(obj, BaseModel) else [obj]
 
-    def _parse_rows(self, rows, cache, download_cb):
-        objs = []
-        for row in rows:
-            obj_row = self.params.row_to_objs(row)
-            for obj in obj_row:
-                if isinstance(obj, File):
-                    obj._set_stream(
-                        self.catalog, caching_enabled=cache, download_cb=download_cb
-                    )
-            objs.append(obj_row)
-        return objs
+    def _parse_row(
+        self, row_dict: RowDict, cache: bool, download_cb: Callback
+    ) -> list[DataValue]:
+        assert self.params
+        row = [row_dict[p] for p in self.params.to_udf_spec()]
+        obj_row = self.params.row_to_objs(row)
+        for obj in obj_row:
+            if isinstance(obj, File):
+                assert self.catalog is not None
+                obj._set_stream(
+                    self.catalog, caching_enabled=cache, download_cb=download_cb
+                )
+        return obj_row
+
+    def _prepare_row(self, row, udf_fields, cache, download_cb):
+        row_dict = RowDict(zip(udf_fields, row))
+        return self._parse_row(row_dict, cache, download_cb)
+
+    def _prepare_row_and_id(self, row, udf_fields, cache, download_cb):
+        row_dict = RowDict(zip(udf_fields, row))
+        udf_input = self._parse_row(row_dict, cache, download_cb)
+        return row_dict["sys__id"], *udf_input
 
     def process_safe(self, obj_rows):
         try:
@@ -358,23 +279,128 @@ class UDFBase(AbstractUDF):
 class Mapper(UDFBase):
     """Inherit from this class to pass to `DataChain.map()`."""
 
+    def run(
+        self,
+        udf_fields: "Sequence[str]",
+        udf_inputs: "Iterable[Sequence[Any]]",
+        catalog: "Catalog",
+        cache: bool,
+        download_cb: Callback = DEFAULT_CALLBACK,
+        processed_cb: Callback = DEFAULT_CALLBACK,
+    ) -> Iterator[Iterable[UDFResult]]:
+        self.catalog = catalog
+        self.setup()
+
+        for row in udf_inputs:
+            id_, *udf_args = self._prepare_row_and_id(
+                row, udf_fields, cache, download_cb
+            )
+            result_objs = self.process_safe(udf_args)
+            udf_output = self._flatten_row(result_objs)
+            output = [{"sys__id": id_} | dict(zip(self.signal_names, udf_output))]
+            processed_cb.relative_update(1)
+            yield output
+
+        self.teardown()
+
 
 class BatchMapper(UDFBase):
     """Inherit from this class to pass to `DataChain.batch_map()`."""
 
-    is_input_batched = True
     is_output_batched = True
 
+    def run(
+        self,
+        udf_fields: Sequence[str],
+        udf_inputs: Iterable[RowsOutputBatch],
+        catalog: "Catalog",
+        cache: bool,
+        download_cb: Callback = DEFAULT_CALLBACK,
+        processed_cb: Callback = DEFAULT_CALLBACK,
+    ) -> Iterator[Iterable[UDFResult]]:
+        self.catalog = catalog
+        self.setup()
+
+        for batch in udf_inputs:
+            n_rows = len(batch.rows)
+            row_ids, *udf_args = zip(
+                *[
+                    self._prepare_row_and_id(row, udf_fields, cache, download_cb)
+                    for row in batch.rows
+                ]
+            )
+            result_objs = list(self.process_safe(udf_args))
+            n_objs = len(result_objs)
+            assert (
+                n_objs == n_rows
+            ), f"{self.name} returns {n_objs} rows, but {n_rows} were expected"
+            udf_outputs = (self._flatten_row(row) for row in result_objs)
+            output = [
+                {"sys__id": row_id} | dict(zip(self.signal_names, signals))
+                for row_id, signals in zip(row_ids, udf_outputs)
+            ]
+            processed_cb.relative_update(n_rows)
+            yield output
+
+        self.teardown()
+
 
 class Generator(UDFBase):
     """Inherit from this class to pass to `DataChain.gen()`."""
 
     is_output_batched = True
 
+    def run(
+        self,
+        udf_fields: "Sequence[str]",
+        udf_inputs: "Iterable[Sequence[Any]]",
+        catalog: "Catalog",
+        cache: bool,
+        download_cb: Callback = DEFAULT_CALLBACK,
+        processed_cb: Callback = DEFAULT_CALLBACK,
+    ) -> Iterator[Iterable[UDFResult]]:
+        self.catalog = catalog
+        self.setup()
+
+        for row in udf_inputs:
+            udf_args = self._prepare_row(row, udf_fields, cache, download_cb)
+            result_objs = self.process_safe(udf_args)
+            udf_outputs = (self._flatten_row(row) for row in result_objs)
+            output = (dict(zip(self.signal_names, row)) for row in udf_outputs)
+            processed_cb.relative_update(1)
+            yield output
+
+        self.teardown()
+
 
 class Aggregator(UDFBase):
     """Inherit from this class to pass to `DataChain.agg()`."""
 
-    is_input_batched = True
     is_output_batched = True
-    is_input_grouped = True
+
+    def run(
+        self,
+        udf_fields: "Sequence[str]",
+        udf_inputs: Iterable[RowsOutputBatch],
+        catalog: "Catalog",
+        cache: bool,
+        download_cb: Callback = DEFAULT_CALLBACK,
+        processed_cb: Callback = DEFAULT_CALLBACK,
+    ) -> Iterator[Iterable[UDFResult]]:
+        self.catalog = catalog
+        self.setup()
+
+        for batch in udf_inputs:
+            udf_args = zip(
+                *[
+                    self._prepare_row(row, udf_fields, cache, download_cb)
+                    for row in batch.rows
+                ]
+            )
+            result_objs = self.process_safe(udf_args)
+            udf_outputs = (self._flatten_row(row) for row in result_objs)
+            output = (dict(zip(self.signal_names, row)) for row in udf_outputs)
+            processed_cb.relative_update(len(batch.rows))
+            yield output
+
+        self.teardown()

datachain/lib/utils.py

@@ -23,3 +23,8 @@ class DataChainError(Exception):
 class DataChainParamsError(DataChainError):
     def __init__(self, message):
         super().__init__(message)
+
+
+class DataChainColumnError(DataChainParamsError):
+    def __init__(self, col_name, msg):
+        super().__init__(f"Error for column {col_name}: {msg}")

datachain/query/__init__.py

@@ -1,12 +1,11 @@
 from .dataset import DatasetQuery
 from .params import param
-from .schema import C, DatasetRow, LocalFilename, Object, Stream
+from .schema import C, LocalFilename, Object, Stream
 from .session import Session
 
 __all__ = [
     "C",
     "DatasetQuery",
-    "DatasetRow",
     "LocalFilename",
     "Object",
     "Session",

datachain/query/batch.py

@@ -11,8 +11,6 @@ from datachain.data_storage.warehouse import SELECT_BATCH_SIZE
 if TYPE_CHECKING:
     from sqlalchemy import Select
 
-    from datachain.dataset import RowDict
-
 
 @dataclass
 class RowsOutputBatch:
@@ -22,14 +20,6 @@ class RowsOutputBatch:
 RowsOutput = Union[Sequence, RowsOutputBatch]
 
 
-@dataclass
-class UDFInputBatch:
-    rows: Sequence["RowDict"]
-
-
-UDFInput = Union["RowDict", UDFInputBatch]
-
-
 class BatchingStrategy(ABC):
     """BatchingStrategy provides means of batching UDF executions."""
 
@@ -107,7 +97,6 @@ class Partition(BatchingStrategy):
 
         ordered_query = query.order_by(None).order_by(
             PARTITION_COLUMN_ID,
-            "sys__id",
             *query._order_by_clauses,
         )
 

datachain/query/dataset.py

@@ -392,7 +392,7 @@ class UDFStep(Step, ABC):
 
     def populate_udf_table(self, udf_table: "Table", query: Select) -> None:
         use_partitioning = self.partition_by is not None
-        batching = self.udf.properties.get_batching(use_partitioning)
+        batching = self.udf.get_batching(use_partitioning)
         workers = self.workers
         if (
             not workers
@@ -591,10 +591,6 @@ class UDFSignal(UDFStep):
             return query, []
         table = self.catalog.warehouse.create_pre_udf_table(query)
         q: Select = sqlalchemy.select(*table.c)
-        if query._order_by_clauses:
-            # we are adding ordering only if it's explicitly added by user in
-            # query part before adding signals
-            q = q.order_by(table.c.sys__id)
         return q, [table]
 
     def create_result_query(
@@ -630,11 +626,6 @@ class UDFSignal(UDFStep):
             else:
                 res = sqlalchemy.select(*cols1).select_from(subq)
 
-            if query._order_by_clauses:
-                # if ordering is used in query part before adding signals, we
-                # will have it as order by id from select from pre-created udf table
-                res = res.order_by(subq.c.sys__id)
-
             if self.partition_by is not None:
                 subquery = res.subquery()
                 res = sqlalchemy.select(*subquery.c).select_from(subquery)
@@ -666,13 +657,6 @@ class RowGenerator(UDFStep):
     def create_result_query(
         self, udf_table, query: Select
     ) -> tuple[QueryGeneratorFunc, list["sqlalchemy.Column"]]:
-        if not query._order_by_clauses:
-            # if we are not selecting all rows in UDF, we need to ensure that
-            # we get the same rows as we got as inputs of UDF since selecting
-            # without ordering can be non deterministic in some databases
-            c = query.selected_columns
-            query = query.order_by(c.sys__id)
-
         udf_table_query = udf_table.select().subquery()
         udf_table_cols: list[sqlalchemy.Label[Any]] = [
             label(c.name, c) for c in udf_table_query.columns
@@ -957,24 +941,24 @@ class SQLJoin(Step):
 
 
 @frozen
-class GroupBy(Step):
-    """Group rows by a specific column."""
-
-    cols: PartitionByType
+class SQLGroupBy(SQLClause):
+    cols: Sequence[Union[str, ColumnElement]]
+    group_by: Sequence[Union[str, ColumnElement]]
 
-    def clone(self) -> "Self":
-        return self.__class__(self.cols)
+    def apply_sql_clause(self, query) -> Select:
+        if not self.cols:
+            raise ValueError("No columns to select")
+        if not self.group_by:
+            raise ValueError("No columns to group by")
 
-    def apply(
-        self, query_generator: QueryGenerator, temp_tables: list[str]
-    ) -> StepResult:
-        query = query_generator.select()
-        grouped_query = query.group_by(*self.cols)
+        subquery = query.subquery()
 
-        def q(*columns):
-            return grouped_query.with_only_columns(*columns)
+        cols = [
+            subquery.c[str(c)] if isinstance(c, (str, C)) else c
+            for c in [*self.group_by, *self.cols]
+        ]
 
-        return step_result(q, grouped_query.selected_columns)
+        return sqlalchemy.select(*cols).select_from(subquery).group_by(*self.group_by)
 
 
 def _validate_columns(
@@ -1130,25 +1114,14 @@ class DatasetQuery:
             query.steps = query.steps[-1:] + query.steps[:-1]
 
         result = query.starting_step.apply()
-        group_by = None
         self.dependencies.update(result.dependencies)
 
         for step in query.steps:
-            if isinstance(step, GroupBy):
-                if group_by is not None:
-                    raise TypeError("only one group_by allowed")
-                group_by = step
-                continue
-
             result = step.apply(
                 result.query_generator, self.temp_table_names
             )  # a chain of steps linked by results
             self.dependencies.update(result.dependencies)
 
-        if group_by:
-            result = group_by.apply(result.query_generator, self.temp_table_names)
-            self.dependencies.update(result.dependencies)
-
         return result.query_generator
 
     @staticmethod
@@ -1410,9 +1383,13 @@ class DatasetQuery:
         return query.as_scalar()
 
     @detach
-    def group_by(self, *cols: ColumnElement) -> "Self":
+    def group_by(
+        self,
+        cols: Sequence[ColumnElement],
+        group_by: Sequence[ColumnElement],
+    ) -> "Self":
         query = self.clone()
-        query.steps.append(GroupBy(cols))
+        query.steps.append(SQLGroupBy(cols, group_by))
         return query
 
     @detach
@@ -1591,6 +1568,8 @@ class DatasetQuery:
             )
             version = version or dataset.latest_version
 
+            self.session.add_dataset_version(dataset=dataset, version=version)
+
             dr = self.catalog.warehouse.dataset_rows(dataset)
 
             self.catalog.warehouse.copy_table(dr.get_table(), query.select())

datachain/query/dispatch.py

@@ -114,7 +114,6 @@ class UDFDispatcher:
     catalog: Optional[Catalog] = None
     task_queue: Optional[multiprocess.Queue] = None
     done_queue: Optional[multiprocess.Queue] = None
-    _batch_size: Optional[int] = None
 
     def __init__(
         self,
@@ -154,17 +153,6 @@ class UDFDispatcher:
         self.done_queue = None
         self.ctx = get_context("spawn")
 
-    @property
-    def batch_size(self):
-        if self._batch_size is None:
-            if hasattr(self.udf, "properties") and hasattr(
-                self.udf.properties, "batch"
-            ):
-                self._batch_size = self.udf.properties.batch
-            else:
-                self._batch_size = 1
-        return self._batch_size
-
     def _create_worker(self) -> "UDFWorker":
         if not self.catalog:
             id_generator = self.id_generator_class(