Add sort dataframe logic on qid (#239)

* add sort dataframe logic on qid for centralized

* move sorting logic to _split_dataframe

* Update xgboost_ray/matrix.py

Co-authored-by: Antoni Baum <[email protected]>
Signed-off-by: atomic <[email protected]>

* refactor sorting logic to method

* logic for more cases of qid type and add integration test

- add logic to include more case of qid data type (array, dataframe)
- add 2 integration tests to cover behavior for sorting qid

* raise exception for the case when qid DataFrame is using unexpected shape

* fix lint

* more lint fix

* add unittest skip for xgboost 0.9

* Apply suggestions from code review

Signed-off-by: Antoni Baum <[email protected]>

* Fix

Signed-off-by: Antoni Baum <[email protected]>

* Test tweak

Signed-off-by: Antoni Baum <[email protected]>

Signed-off-by: atomic <[email protected]>
Signed-off-by: Antoni Baum <[email protected]>
Co-authored-by: Antoni Baum <[email protected]>

xgboost_ray/matrix.py

@@ -48,6 +48,31 @@ def concat_dataframes(dfs: List[Optional[pd.DataFrame]]):
     return pd.concat(filtered, ignore_index=True, copy=False)
 
 
+def ensure_sorted_by_qid(df: pd.DataFrame, qid: Data
+                         ) -> Tuple[Union[np.array, str], pd.DataFrame]:
+    _qid: pd.Series = None
+    if isinstance(qid, str):
+        _qid = df[qid]
+    elif isinstance(qid, np.ndarray):
+        _qid = pd.Series(qid)
+    elif isinstance(qid, pd.DataFrame):
+        if len(df.shape) != 2 and df.shape[1] != 1:
+            raise ValueError(f"qid argument of type pd.DataFrame is expected"
+                             "to contains only 1 column of data "
+                             f"but the qid passed in is of shape {df.shape}.")
+        _qid = qid.iloc[:, 0]
+    elif isinstance(qid, pd.Series):
+        _qid = qid
+    if _qid.is_monotonic:
+        return _qid, df
+    else:
+        if isinstance(qid, str):
+            return qid, df.sort_values([qid])
+        else:  # case when qid is not part of df
+            return _qid.sort_values(), \
+                   df.set_index(_qid).sort_index().reset_index(drop=True)
+
+
 @PublicAPI(stability="beta")
 class RayShardingMode(Enum):
     """Enum for different modes of sharding the data.
@@ -227,6 +252,12 @@ def _split_dataframe(
         `label_upper_bound`
 
         """
+        # sort dataframe by qid if exists (required by DMatrix)
+        if self.qid is not None:
+            _qid, local_data = ensure_sorted_by_qid(local_data, self.qid)
+            if not isinstance(self.qid, str):
+                self.qid = _qid
+
         exclude_cols: Set[str] = set()  # Exclude these columns from `x`
 
         label, exclude = data_source.get_column(local_data, self.label)

xgboost_ray/tests/test_matrix.py

@@ -1,6 +1,8 @@
+import inspect
 import os
 import tempfile
 import unittest
+import xgboost as xgb
 
 import numpy as np
 import pandas as pd
@@ -27,7 +29,7 @@ def setUp(self):
 
     @classmethod
     def setUpClass(cls):
-        ray.init(num_cpus=1, local_mode=True)
+        ray.init(local_mode=True)
 
     @classmethod
     def tearDownClass(cls):
@@ -356,6 +358,56 @@ def testLegacyParams(self):
             label_lower_bound=label_lower_bound,
             label_upper_bound=label_upper_bound)
 
+    @unittest.skipIf("qid" not in inspect.signature(xgb.DMatrix).parameters,
+                     f"not supported in xgb version {xgb.__version__}")
+    def testQidSortedBehaviorXGBoost(self):
+        """Test that data with unsorted qid is sorted in RayDMatrix"""
+        in_x = self.x
+        in_y = self.y
+        unsorted_qid = np.array([1, 2] * 16)
+
+        from xgboost import DMatrix
+        with self.assertRaises(ValueError):
+            DMatrix(**{"data": in_x, "label": in_y, "qid": unsorted_qid})
+        DMatrix(**{
+            "data": in_x,
+            "label": in_y,
+            "qid": np.sort(unsorted_qid)
+        })  # no exception
+        # test RayDMatrix handles sorting automatically
+        mat = RayDMatrix(in_x, in_y, qid=unsorted_qid)
+        params = mat.get_data(rank=0, num_actors=1)
+        DMatrix(**params)
+
+    @unittest.skipIf("qid" not in inspect.signature(xgb.DMatrix).parameters,
+                     f"not supported in xgb version {xgb.__version__}")
+    def testQidSortedParquet(self):
+        from xgboost import DMatrix
+        with tempfile.TemporaryDirectory() as dir:
+            parquet_file1 = os.path.join(dir, "file1.parquet")
+            parquet_file2 = os.path.join(dir, "file2.parquet")
+
+            unsorted_qid1 = np.array([2, 4] * 16)
+            unsorted_qid2 = np.array([1, 3] * 16)
+
+            # parquet 1
+            data_df = pd.DataFrame(self.x, columns=["a", "b", "c", "d"])
+            data_df["label"] = pd.Series(self.y)
+            data_df["group"] = pd.Series(unsorted_qid1)
+            data_df.to_parquet(parquet_file1)
+            # parquet 2
+            data_df = pd.DataFrame(self.x, columns=["a", "b", "c", "d"])
+            data_df["label"] = pd.Series(self.y)
+            data_df["group"] = pd.Series(unsorted_qid2)
+            data_df.to_parquet(parquet_file2)
+            mat = RayDMatrix(
+                [parquet_file1, parquet_file2],
+                columns=["a", "b", "c", "d", "label", "group"],
+                label="label",
+                qid="group")
+            params = mat.get_data(rank=0, num_actors=1)
+            DMatrix(**params)
+
 
 if __name__ == "__main__":
     import pytest