maxframe 1.0.0rc1__cp38-cp38-macosx_10_9_universal2.whl → 1.0.0rc3__cp38-cp38-macosx_10_9_universal2.whl

maxframe/{odpsio → io/odpsio}/tests/test_schema.py

@@ -18,9 +18,9 @@ import pyarrow as pa
 import pytest
 from odps import types as odps_types
 
-from ... import dataframe as md
-from ... import tensor as mt
-from ...core import OutputType
+from .... import dataframe as md
+from .... import tensor as mt
+from ....core import OutputType
 from ..schema import (
     arrow_schema_to_odps_schema,
     build_dataframe_table_meta,
@@ -143,17 +143,17 @@ def test_pandas_to_odps_schema_index(wrap_obj):
     data = pd.Index(np.random.randint(0, 100, 100))
 
     test_idx = _wrap_maxframe_obj(data, wrap=wrap_obj)
-    schema, meta = pandas_to_odps_schema(test_idx, unknown_as_string=True)
-    assert [c.name for c in schema.columns] == ["_idx_0"]
-    assert [c.type.name for c in schema.columns] == ["bigint"]
-    assert meta.type == OutputType.index
-    assert meta.table_column_names == []
-    assert meta.table_index_column_names == ["_idx_0"]
-    assert meta.pd_column_level_names == []
-    assert meta.pd_index_level_names == [None]
-
-    with pytest.raises(AssertionError):
-        pandas_to_odps_schema(test_idx, unknown_as_string=True, ignore_index=True)
+    for ignore_idx in (False, True):
+        schema, meta = pandas_to_odps_schema(
+            test_idx, unknown_as_string=True, ignore_index=ignore_idx
+        )
+        assert [c.name for c in schema.columns] == ["_idx_0"]
+        assert [c.type.name for c in schema.columns] == ["bigint"]
+        assert meta.type == OutputType.index
+        assert meta.table_column_names == []
+        assert meta.table_index_column_names == ["_idx_0"]
+        assert meta.pd_column_level_names == []
+        assert meta.pd_index_level_names == [None]
 
     data = pd.MultiIndex.from_arrays(
         [np.random.choice(list("ABC"), 100), np.random.randint(0, 10, 100)],
@@ -177,6 +177,7 @@ def test_pandas_to_odps_schema_scalar(wrap_obj):
     test_scalar = _wrap_maxframe_obj(data, wrap=wrap_obj)
     if wrap_obj != "no":
         test_scalar.op.data = None
+
     schema, meta = pandas_to_odps_schema(test_scalar, unknown_as_string=True)
     assert schema.columns[0].name == "_idx_0"
     assert schema.columns[0].type.name == "double"
@@ -186,9 +187,6 @@ def test_pandas_to_odps_schema_scalar(wrap_obj):
     assert meta.pd_column_level_names == []
     assert meta.pd_index_level_names == [None]
 
-    with pytest.raises(AssertionError):
-        pandas_to_odps_schema(test_scalar, unknown_as_string=True, ignore_index=True)
-
 
 def test_odps_arrow_schema_conversion():
     odps_schema = odps_types.OdpsSchema(
@@ -211,10 +209,11 @@ def test_odps_arrow_schema_conversion():
             odps_types.Column("col16", "struct<a1: string, a2: map<string, bigint>>"),
             odps_types.Column("col17", "CHAR(15)"),
             odps_types.Column("col18", "VARCHAR(15)"),
+            odps_types.Column("col19", "decimal"),
         ]
     )
     arrow_schema = odps_schema_to_arrow_schema(odps_schema)
-    assert arrow_schema.names == [f"col{i}" for i in range(1, 19)]
+    assert arrow_schema.names == [f"col{i}" for i in range(1, 20)]
     assert arrow_schema.types == [
         pa.string(),
         pa.binary(),
@@ -234,6 +233,7 @@ def test_odps_arrow_schema_conversion():
         pa.struct([("a1", pa.string()), ("a2", pa.map_(pa.string(), pa.int64()))]),
         pa.string(),
         pa.string(),
+        pa.decimal128(38, 18),
     ]
 
     expected_odps_schema = odps_types.OdpsSchema(
@@ -256,6 +256,7 @@ def test_odps_arrow_schema_conversion():
             odps_types.Column("col16", "struct<a1: string, a2: map<string, bigint>>"),
             odps_types.Column("col17", "string"),
             odps_types.Column("col18", "string"),
+            odps_types.Column("col19", "decimal(38, 18)"),
         ]
     )
 

maxframe/{odpsio → io/odpsio}/tests/test_tableio.py

@@ -12,22 +12,37 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+import datetime
+
 import numpy as np
 import pandas as pd
 import pyarrow as pa
+import pytest
 from odps import ODPS
 
-from ...tests.utils import flaky, tn
-from ...utils import config_odps_default_options
-from ..tableio import HaloTableIO
+from ....config import options
+from ....tests.utils import flaky, tn
+from ....utils import config_odps_default_options
+from ..tableio import ODPSTableIO
+
+
+@pytest.fixture
+def switch_table_io(request):
+    old_use_common_table = options.use_common_table
+    try:
+        options.use_common_table = request.param
+        yield
+    finally:
+        options.use_common_table = old_use_common_table
 
 
 @flaky(max_runs=3)
-def test_empty_table_io():
+@pytest.mark.parametrize("switch_table_io", [False, True], indirect=True)
+def test_empty_table_io(switch_table_io):
     config_odps_default_options()
 
     o = ODPS.from_environments()
-    halo_table_io = HaloTableIO(o)
+    table_io = ODPSTableIO(o)
 
     # test read from empty table
     empty_table_name = tn("test_empty_table_halo_read")
@@ -35,42 +50,53 @@ def test_empty_table_io():
     tb = o.create_table(empty_table_name, "col1 string", lifecycle=1)
 
     try:
-        with halo_table_io.open_reader(empty_table_name) as reader:
+        with table_io.open_reader(empty_table_name) as reader:
             assert len(reader.read_all()) == 0
     finally:
         tb.drop()
 
 
 @flaky(max_runs=3)
-def test_table_io_without_parts():
+@pytest.mark.parametrize("switch_table_io", [False, True], indirect=True)
+def test_table_io_without_parts(switch_table_io):
     config_odps_default_options()
 
     o = ODPS.from_environments()
-    halo_table_io = HaloTableIO(o)
+    table_io = ODPSTableIO(o)
 
     # test read and write tables without partition
     no_part_table_name = tn("test_no_part_halo_write")
     o.delete_table(no_part_table_name, if_exists=True)
-    tb = o.create_table(
-        no_part_table_name, ",".join(f"{c} double" for c in "abcde"), lifecycle=1
-    )
+    col_desc = ",".join(f"{c} double" for c in "abcde") + ", f datetime"
+    tb = o.create_table(no_part_table_name, col_desc, lifecycle=1)
 
     try:
         pd_data = pd.DataFrame(np.random.rand(100, 5), columns=list("abcde"))
-        with halo_table_io.open_writer(no_part_table_name) as writer:
+        date_val = [
+            (
+                datetime.datetime.now().replace(microsecond=0)
+                + datetime.timedelta(seconds=i)
+            )
+            for i in range(100)
+        ]
+        pd_data["f"] = pd.Series(date_val, dtype="datetime64[ms]").dt.tz_localize(
+            options.local_timezone
+        )
+        with table_io.open_writer(no_part_table_name) as writer:
             writer.write(pa.Table.from_pandas(pd_data, preserve_index=False))
-        with halo_table_io.open_reader(no_part_table_name) as reader:
+        with table_io.open_reader(no_part_table_name) as reader:
             pd.testing.assert_frame_equal(reader.read_all().to_pandas(), pd_data)
     finally:
         tb.drop()
 
 
 @flaky(max_runs=3)
-def test_table_io_with_range_reader():
+@pytest.mark.parametrize("switch_table_io", [False, True], indirect=True)
+def test_table_io_with_range_reader(switch_table_io):
     config_odps_default_options()
 
     o = ODPS.from_environments()
-    halo_table_io = HaloTableIO(o)
+    table_io = ODPSTableIO(o)
 
     # test read and write tables without partition
     no_part_table_name = tn("test_no_part_halo_write")
@@ -81,15 +107,15 @@ def test_table_io_with_range_reader():
 
     try:
         pd_data = pd.DataFrame(np.random.rand(100, 5), columns=list("abcde"))
-        with halo_table_io.open_writer(no_part_table_name) as writer:
+        with table_io.open_writer(no_part_table_name) as writer:
             writer.write(pa.Table.from_pandas(pd_data, preserve_index=False))
 
-        with halo_table_io.open_reader(
+        with table_io.open_reader(
             no_part_table_name, start=None, stop=100, row_batch_size=10
         ) as reader:
             pd.testing.assert_frame_equal(reader.read_all().to_pandas(), pd_data)
 
-        with halo_table_io.open_reader(
+        with table_io.open_reader(
             no_part_table_name,
             start=-2,
             stop=-52,
@@ -105,11 +131,12 @@ def test_table_io_with_range_reader():
 
 
 @flaky(max_runs=3)
-def test_table_io_with_parts():
+@pytest.mark.parametrize("switch_table_io", [False, True], indirect=True)
+def test_table_io_with_parts(switch_table_io):
     config_odps_default_options()
 
     o = ODPS.from_environments()
-    halo_table_io = HaloTableIO(o)
+    table_io = ODPSTableIO(o)
 
     # test read and write tables with partition
     parted_table_name = tn("test_parted_halo_write")
@@ -122,11 +149,11 @@ def test_table_io_with_parts():
 
     try:
         pd_data = pd.DataFrame(np.random.rand(100, 5), columns=list("abcde"))
-        with halo_table_io.open_writer(parted_table_name, "pt=test") as writer:
+        with table_io.open_writer(parted_table_name, "pt=test") as writer:
             writer.write(pa.Table.from_pandas(pd_data, preserve_index=False))
-        with halo_table_io.open_reader(parted_table_name, "pt=test") as reader:
+        with table_io.open_reader(parted_table_name, "pt=test") as reader:
             pd.testing.assert_frame_equal(reader.read_all().to_pandas(), pd_data)
-        with halo_table_io.open_reader(
+        with table_io.open_reader(
             parted_table_name, "pt=test", partition_columns=True
         ) as reader:
             expected_data = pd_data.copy()

maxframe/{odpsio → io/odpsio}/tests/test_volumeio.py

@@ -15,7 +15,7 @@
 import pytest
 from odps import ODPS
 
-from ...tests.utils import tn
+from ....tests.utils import tn
 from ..volumeio import ODPSVolumeReader, ODPSVolumeWriter
 
 
@@ -69,19 +69,17 @@ def create_volume(request, oss_config):
             oss_config.oss_bucket.batch_delete_objects(keys)
 
 
-@pytest.mark.parametrize("create_volume", ["parted", "external"], indirect=True)
+@pytest.mark.parametrize("create_volume", ["external"], indirect=True)
 def test_read_write_volume(create_volume):
     test_vol_dir = "test_vol_dir"
 
     odps_entry = ODPS.from_environments()
 
     writer = ODPSVolumeWriter(odps_entry, create_volume, test_vol_dir)
-    write_session_id = writer.create_write_session()
 
     writer = ODPSVolumeWriter(odps_entry, create_volume, test_vol_dir)
-    writer.write_file("file1", b"content1", write_session_id)
-    writer.write_file("file2", b"content2", write_session_id)
-    writer.commit(["file1", "file2"], write_session_id)
+    writer.write_file("file1", b"content1")
+    writer.write_file("file2", b"content2")
 
     reader = ODPSVolumeReader(odps_entry, create_volume, test_vol_dir)
     assert reader.read_file("file1") == b"content1"

maxframe/io/odpsio/volumeio.py

@@ -0,0 +1,57 @@
+# Copyright 1999-2024 Alibaba Group Holding Ltd.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import inspect
+from typing import Iterator, List, Union
+
+from odps import ODPS
+
+
+class ODPSVolumeReader:
+    def __init__(self, odps_entry: ODPS, volume_name: str, volume_dir: str):
+        self._odps_entry = odps_entry
+        self._volume = odps_entry.get_volume(volume_name)
+        self._volume_dir = volume_dir
+
+    def list_files(self) -> List[str]:
+        def _get_file_name(vol_file):
+            if hasattr(vol_file, "name"):
+                return vol_file.name
+            return vol_file.path.rsplit("/", 1)[-1]
+
+        return [
+            _get_file_name(f)
+            for f in self._odps_entry.list_volume_files(
+                f"/{self._volume.name}/{self._volume_dir}"
+            )
+        ]
+
+    def read_file(self, file_name: str) -> bytes:
+        with self._volume.open_reader(self._volume_dir + "/" + file_name) as reader:
+            return reader.read()
+
+
+class ODPSVolumeWriter:
+    def __init__(self, odps_entry: ODPS, volume_name: str, volume_dir: str):
+        self._odps_entry = odps_entry
+        self._volume = odps_entry.get_volume(volume_name)
+        self._volume_dir = volume_dir
+
+    def write_file(self, file_name: str, data: Union[bytes, Iterator[bytes]]):
+        with self._volume.open_writer(self._volume_dir + "/" + file_name) as writer:
+            if not inspect.isgenerator(data):
+                writer.write(data)
+            else:
+                for chunk in data:
+                    writer.write(chunk)

maxframe/learn/contrib/xgboost/classifier.py

@@ -14,7 +14,7 @@
 
 import numpy as np
 
-from ....tensor import argmax
+from ....tensor import argmax, transpose, vstack
 from ..utils import make_import_error_func
 from .core import XGBScikitLearnBase, xgboost
 
@@ -42,7 +42,10 @@ else:
             sample_weight_eval_set=None,
             base_margin_eval_set=None,
             num_class=None,
+            **kw,
         ):
+            session = kw.pop("session", None)
+            run_kwargs = kw.pop("run_kwargs", dict())
             dtrain, evals = wrap_evaluation_matrices(
                 None,
                 X,
@@ -68,6 +71,8 @@ else:
                 evals=evals,
                 evals_result=self.evals_result_,
                 num_class=num_class,
+                session=session,
+                run_kwargs=run_kwargs,
             )
             self._Booster = result
             return self
@@ -83,4 +88,23 @@ else:
         def predict_proba(self, data, ntree_limit=None, flag=False, **kw):
             if ntree_limit is not None:
                 raise NotImplementedError("ntree_limit is not currently supported")
-            return predict(self.get_booster(), data, flag=flag, **kw)
+            prediction = predict(self.get_booster(), data, flag=flag, **kw)
+
+            if len(prediction.shape) == 2 and prediction.shape[1] == self.n_classes_:
+                # multi-class
+                return prediction
+            if (
+                len(prediction.shape) == 2
+                and self.n_classes_ == 2
+                and prediction.shape[1] >= self.n_classes_
+            ):
+                # multi-label
+                return prediction
+            # binary logistic function
+            classone_probs = prediction
+            classzero_probs = 1.0 - classone_probs
+            return transpose(vstack((classzero_probs, classone_probs)))
+
+        @property
+        def classes_(self) -> np.ndarray:
+            return np.arange(self.n_classes_)

maxframe/learn/contrib/xgboost/core.py

@@ -12,15 +12,67 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from typing import Any, Callable, List, Optional, Tuple
+from typing import Any, Callable, Dict, List, Optional, Tuple
 
 try:
     import xgboost
 except ImportError:
     xgboost = None
 
+from ...core import Model, ModelData
 from .dmatrix import DMatrix
 
+
+class BoosterData(ModelData):
+    __slots__ = ("_evals_result",)
+
+    _evals_result: Dict
+
+    def __init__(self, *args, evals_result=None, **kwargs):
+        super().__init__(*args, **kwargs)
+        self._evals_result = evals_result if evals_result is not None else dict()
+
+    def execute(self, session=None, **kw):
+        # The evals_result should be fetched when BoosterData.execute() is called.
+        result = super().execute(session=session, **kw)
+        if self.op.has_evals_result and self.key == self.op.outputs[0].key:
+            self._evals_result.update(self.op.outputs[1].fetch(session=session))
+        return result
+
+    def predict(
+        self,
+        data,
+        output_margin=False,
+        pred_leaf=False,
+        pred_contribs=False,
+        approx_contribs=False,
+        pred_interactions=False,
+        validate_features=True,
+        training=False,
+        iteration_range=None,
+        strict_shape=False,
+    ):
+        from .predict import predict
+
+        return predict(
+            self,
+            data,
+            output_margin=output_margin,
+            pred_leaf=pred_leaf,
+            pred_contribs=pred_contribs,
+            approx_contribs=approx_contribs,
+            pred_interactions=pred_interactions,
+            validate_features=validate_features,
+            training=training,
+            iteration_range=iteration_range,
+            strict_shape=strict_shape,
+        )
+
+
+class Booster(Model):
+    pass
+
+
 if not xgboost:
     XGBScikitLearnBase = None
 else:
@@ -40,7 +92,9 @@ else:
             **kw,
         ):
             """
-            Fit the regressor.
+            Fit the regressor. Note that fit() is an eager-execution
+            API. The call will be blocked until training finished.
+
             Parameters
             ----------
             X : array_like
@@ -72,6 +126,37 @@ else:
             """
             raise NotImplementedError
 
+        def evals_result(self, **kw) -> Dict:
+            """Return the evaluation results.
+
+            If **eval_set** is passed to the :py:meth:`fit` function, you can call
+            ``evals_result()`` to get evaluation results for all passed **eval_sets**.  When
+            **eval_metric** is also passed to the :py:meth:`fit` function, the
+            **evals_result** will contain the **eval_metrics** passed to the :py:meth:`fit`
+            function.
+
+            The returned evaluation result is a dictionary:
+
+            .. code-block:: python
+
+                {'validation_0': {'logloss': ['0.604835', '0.531479']},
+                 'validation_1': {'logloss': ['0.41965', '0.17686']}}
+
+            Note that evals_result() will be blocked until the train is finished.
+
+            Returns
+            -------
+            evals_result
+
+            """
+            result = super().evals_result()
+            if not self._Booster.op.has_evals_result or len(result) != 0:
+                return result
+            session = kw.pop("session", None)
+            run_kwargs = kw.pop("run_kwargs", dict())
+            self._Booster.execute(session=session, **run_kwargs)
+            return super().evals_result()
+
     def wrap_evaluation_matrices(
         missing: float,
         X: Any,

maxframe/learn/contrib/xgboost/dmatrix.py

@@ -13,7 +13,7 @@
 # limitations under the License.
 
 
-from .... import opcodes as OperandDef
+from .... import opcodes
 from ....core.entity.output_types import get_output_types
 from ....core.operator.base import Operator
 from ....core.operator.core import TileableOperatorMixin
@@ -27,7 +27,7 @@ from ...utils import convert_to_tensor_or_dataframe
 
 
 class ToDMatrix(Operator, TileableOperatorMixin):
-    _op_type_ = OperandDef.TO_DMATRIX
+    _op_type_ = opcodes.TO_DMATRIX
 
     data = KeyField("data", default=None)
     label = KeyField("label", default=None)
@@ -99,10 +99,7 @@ def check_array_like(y: TileableType, name: str) -> TileableType:
     y = convert_to_tensor_or_dataframe(y)
     if isinstance(y, DATAFRAME_TYPE):
         y = y.iloc[:, 0]
-    y = astensor(y)
-    if y.ndim != 1:
-        raise ValueError(f"Expecting 1-d {name}, got: {y.ndim}-d")
-    return y
+    return astensor(y)
 
 
 def to_dmatrix(

maxframe/learn/contrib/xgboost/predict.py

@@ -12,29 +12,32 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import pickle
 
 import numpy as np
 import pandas as pd
 
-from .... import opcodes as OperandDef
+from .... import opcodes
 from ....core.entity.output_types import OutputType
 from ....core.operator.base import Operator
 from ....core.operator.core import TileableOperatorMixin
 from ....dataframe.utils import parse_index
-from ....serialization.serializables import BoolField, BytesField, KeyField, TupleField
+from ....serialization.serializables import (
+    BoolField,
+    KeyField,
+    ReferenceField,
+    TupleField,
+)
 from ....tensor.core import TENSOR_TYPE, TensorOrder
+from .core import BoosterData
 from .dmatrix import check_data
 
 
 class XGBPredict(Operator, TileableOperatorMixin):
-    _op_type_ = OperandDef.XGBOOST_PREDICT
+    _op_type_ = opcodes.XGBOOST_PREDICT
     output_dtype = np.dtype(np.float32)
 
     data = KeyField("data", default=None)
-    model = BytesField(
-        "model", on_serialize=pickle.dumps, on_deserialize=pickle.loads, default=None
-    )
+    model = ReferenceField("model", reference_type=BoosterData, default=None)
     pred_leaf = BoolField("pred_leaf", default=False)
     pred_contribs = BoolField("pred_contribs", default=False)
     approx_contribs = BoolField("approx_contribs", default=False)
@@ -107,6 +110,17 @@ def predict(
     strict_shape=False,
     flag=False,
 ):
+    """
+    Using MaxFrame XGBoost model to predict data.
+
+    Parameters
+    ----------
+    Parameters are the same as `xgboost.train`. The predict() is lazy-execution mode.
+
+    Returns
+    -------
+    results: Booster
+    """
     data = check_data(data)
     # TODO: check model datatype
 

maxframe/learn/contrib/xgboost/regressor.py

@@ -41,11 +41,6 @@ else:
         ):
             session = kw.pop("session", None)
             run_kwargs = kw.pop("run_kwargs", dict())
-            if kw:
-                raise TypeError(
-                    f"fit got an unexpected keyword argument '{next(iter(kw))}'"
-                )
-
             dtrain, evals = wrap_evaluation_matrices(
                 None,
                 X,
@@ -57,6 +52,8 @@ else:
                 base_margin_eval_set,
             )
             params = self.get_xgb_params()
+            if not params.get("objective"):
+                params["objective"] = "reg:squarederror"
             self.evals_result_ = dict()
             result = train(
                 params,
@@ -71,8 +68,4 @@ else:
             return self
 
         def predict(self, data, **kw):
-            session = kw.pop("session", None)
-            run_kwargs = kw.pop("run_kwargs", None)
-            return predict(
-                self.get_booster(), data, session=session, run_kwargs=run_kwargs, **kw
-            )
+            return predict(self.get_booster(), data, **kw)