Update indexer instantiation. Allow loc from index with duplicates.

.circleci/config.yml

@@ -0,0 +1,11 @@
+version: 2
+jobs:
+  build:
+    working_directory: ~/sparsity
+    docker:
+      - image: drtools/dask:latest
+    steps:
+      - checkout
+      - run: pip install pytest pytest-cov moto .
+      - run: py.test --cov sparsity --cov-report xml sparsity
+      - run: bash <(curl -s https://codecov.io/bash)

setup.py

@@ -29,7 +29,7 @@
     packages=packages,
     cmdclass=versioneer.get_cmdclass(),
     install_requires=[
-                        'pandas>=0.19.0,<0.23.0',
+                        'pandas>=0.19.0',
                         'scipy>=0.18.1',
                         'numpy>=1.12.0',
                         's3fs>=0.1.0'

sparsity/indexing.py

@@ -1,5 +1,13 @@
 from pandas.core.indexing import _LocIndexer, _iLocIndexer
 
+def get_indexers_list():
+
+    return [
+        ('iloc', _CsrILocationIndexer),
+        ('loc', _CsrLocIndexer),
+    ]
+
+
 class _CsrLocIndexer(_LocIndexer):
 
     def __getitem__(self, item):
@@ -10,6 +18,7 @@ def _slice(self, slice, axis=0, kind=None):
             raise NotImplementedError()
         return self.obj._slice(slice)
 
+
 class _CsrILocationIndexer(_iLocIndexer):
 
     def __getitem__(self, item):
@@ -18,4 +27,4 @@ def __getitem__(self, item):
     def _slice(self, slice, axis=0, kind=None):
         if axis != 0:
             raise NotImplementedError()
-        return self.obj._slice(slice)
+        return self.obj._slice(slice)

sparsity/sparse_frame.py

@@ -1,4 +1,5 @@
 # coding=utf-8
+import functools
 import traceback
 import uuid
 import warnings
@@ -22,7 +23,8 @@
     trail_db = True
 except:
     trail_db = False
-from sparsity.indexing import _CsrILocationIndexer, _CsrLocIndexer
+from sparsity.indexing import _CsrILocationIndexer, _CsrLocIndexer, \
+    get_indexers_list
 
 
 def _is_empty(data):
@@ -45,9 +47,6 @@ class SparseFrame(object):
     Simple sparse table based on scipy.sparse.csr_matrix
     """
 
-    __slots__ = ["_index", "_columns", "_data", "shape",
-                 'ndim', 'iloc', 'loc', 'empty']
-
     def __init__(self, data, index=None, columns=None, **kwargs):
         if len(data.shape) > 2:
             raise ValueError("Only two dimensional data supported")
@@ -86,8 +85,17 @@ def __init__(self, data, index=None, columns=None, **kwargs):
 
         # register indexers
         self.ndim = 2
-        self.iloc = _CsrILocationIndexer(self, 'iloc')
-        self.loc = _CsrLocIndexer(self, 'loc')
+
+    @classmethod
+    def _create_indexer(cls, name, indexer):
+        """Create an indexer like _name in the class."""
+        if getattr(cls, name, None) is None:
+            _v = tuple(map(int, pd.__version__.split('.')))
+            if _v >= (0, 23, 0):
+                _indexer = functools.partial(indexer, name)
+            else:
+                _indexer = functools.partial(indexer, name=name)
+            setattr(cls, name, property(_indexer, doc=indexer.__doc__))
 
     def _init_values(self, data, kwargs):
         if isinstance(data, pd.DataFrame):
@@ -219,10 +227,21 @@ def _take(self, *args, **kwargs):
         """
         return self.take(*args, **kwargs)
 
-    def _xs(self, key, *args, **kwargs):
+    def _xs(self, key, *args, axis=0, **kwargs):
         """Used for label based indexing."""
-        loc = self.index.get_loc(key)
-        return SparseFrame(self.data[loc], index=[key], columns=self.columns)
+        if axis == 0:
+            loc = self.index.get_loc(key)
+            new_data = self.data[loc]
+            return SparseFrame(new_data,
+                               index=[key] * new_data.shape[0],
+                               columns=self.columns)
+        else:
+            loc = self.columns.get_loc(key)
+            new_data = self.data[:, loc]
+            return SparseFrame(new_data,
+                               columns=[key] * new_data.shape[1],
+                               index=self.index)
+
 
     @property
     def index(self):
@@ -558,7 +577,7 @@ def drop(self, labels, axis=0):
             labels = [labels]
         if axis == 1:
             mask = np.logical_not(self.columns.isin(labels))
-            sf = self[self.columns[mask].tolist()]
+            sf = self.loc[:, self.columns[mask].tolist()]
         else:
             raise NotImplementedError
         return sf
@@ -572,7 +591,13 @@ def drop_duplicate_idx(self, **kwargs):
     def __getitem__(self, item):
         if not isinstance(item, (tuple, list)):
             item = [item]
-        return self.reindex_axis(item, axis=1)
+        if len(item) > 0:
+            return self.reindex_axis(item, axis=1)
+        else:
+            data = np.empty(shape=(self.shape[0], 0))
+            return SparseFrame(data, index=self.index,
+                               columns=self.columns[[]])
+
 
     def dropna(self):
         """Drop nans from index."""
@@ -609,7 +634,7 @@ def set_index(self, column=None, idx=None, level=None, inplace=False):
                 isinstance(self._index, pd.MultiIndex):
             new_idx = self.index.get_level_values(level)
         elif column is not None:
-            new_idx = np.asarray(self[column].data.todense()).reshape(-1)
+            new_idx = np.asarray(self.loc[:, column].data.todense()).reshape(-1)
 
         if inplace:
             self._index = _ensure_index(new_idx)
@@ -647,6 +672,30 @@ def _get_axis_name(self, axis):
             raise ValueError('No axis named {} for {}'
                              .format(axis, self.__class__))
 
+    def _reindex_with_indexers(self, reindexers, **kwargs):
+        """allow_dups indicates an internal call here """
+
+        # reindex doing multiple operations on different axes if indicated
+        new_data = self.copy()
+        for axis in sorted(reindexers.keys()):
+            index, indexer = reindexers[axis]
+
+            if index is None:
+                continue
+
+            if axis == 0:
+                new_mat = new_data.data[indexer, :]
+                new_data = SparseFrame(new_mat, index=index,
+                                       columns=self.columns)
+            elif axis == 1:
+                new_mat = new_data.data[:, indexer]
+                new_data = SparseFrame(new_mat, columns=index,
+                                       index=self.index)
+            else:
+                raise ValueError('Only supported axes are 0 and 1.')
+
+        return new_data
+
     def reindex(self, labels=None, index=None, columns=None, axis=None,
                 *args, **kwargs):
         """Conform SparseFrame to new index.
@@ -923,3 +972,6 @@ def _check_categories_order(categories1, categories2, categorical_column_name,
                 mismatch_type=mismatch_type
             )
         )
+
+for _name, _indexer in get_indexers_list():
+    SparseFrame._create_indexer(_name, _indexer)

sparsity/test/test_dask_sparse_frame.py

@@ -10,7 +10,6 @@
 import sparsity as sp
 import sparsity.dask as dsp
 from dask.local import get_sync
-from sparsity import sparse_one_hot
 from sparsity.dask.reshape import one_hot_encode
 import pandas.util.testing as pdt
 
@@ -63,29 +62,23 @@ def test_loc(iindexer, correct_shape):
     assert res.shape == correct_shape
 
 def test_dask_loc(clickstream):
-    sf = dd.from_pandas(clickstream, npartitions=10) \
-        .map_partitions(
-        sparse_one_hot,
-        categories={'page_id': list('ABCDE')},
-        meta=list
-    )
-
+    sf = one_hot_encode(dd.from_pandas(clickstream, npartitions=10),
+                        categories={'page_id': list('ABCDE'),
+                                    'other_categorical': list('FGHIJ')},
+                        index_col=['index', 'id'])
     res = sf.loc['2016-01-15':'2016-02-15']
-    res = sp.SparseFrame.concat(res.compute(get=get_sync).tolist())
-    assert res.index.date.max() == dt.date(2016, 2, 15)
-    assert res.index.date.min() == dt.date(2016, 1, 15)
+    res = res.compute()
+    assert res.index.levels[0].max().date() == dt.date(2016, 2, 15)
+    assert res.index.levels[0].min().date() == dt.date(2016, 1, 15)
 
 
 def test_dask_multi_index_loc(clickstream):
-    sf = dd.from_pandas(clickstream, npartitions=10) \
-        .map_partitions(
-            sparse_one_hot,
-            index_col=['index', 'id'],
-            categories={'page_id': list('ABCDE')},
-            meta=list
-    )
+    sf = one_hot_encode(dd.from_pandas(clickstream, npartitions=10),
+                        categories={'page_id': list('ABCDE'),
+                                    'other_categorical': list('FGHIJ')},
+                        index_col=['index', 'id'])
     res = sf.loc['2016-01-15':'2016-02-15']
-    res = sp.SparseFrame.vstack(res.compute(get=get_sync).tolist())
+    res = res.compute()
     assert res.index.get_level_values(0).date.min() == dt.date(2016, 1, 15)
     assert res.index.get_level_values(0).date.max() == dt.date(2016, 2, 15)
 
@@ -234,7 +227,8 @@ def test_assign_column():
     dsf = dsf.assign(new=ds)
     assert dsf._meta.empty
     sf = dsf.compute()
-    assert np.all(sf.todense() == f.assign(new=s))
+    assert np.all((sf.todense() == f.assign(new=s)).values)
+
 
 @pytest.mark.parametrize('arg_dict', [
     dict(divisions=[0, 30, 50, 70, 99]),

sparsity/test/test_sparse_frame.py

@@ -2,7 +2,6 @@
 import datetime as dt
 import os
 
-#import dask.dataframe as dd
 from contextlib import contextmanager
 
 import numpy as np
@@ -202,8 +201,8 @@ def test_loc_multi_index(sf_midx, sf_midx_int):
     assert np.all(sf_midx.loc[dt_slice].data.todense() ==
                   np.identity(5)[:3])
 
-    assert np.all(sf_midx_int.loc[1].todense() == sf_midx.data[:4,:])
-    assert np.all(sf_midx_int.loc[0].todense() == sf_midx.data[4, :])
+    assert np.all(sf_midx_int.loc[1].todense().values == sf_midx.data[:4,:])
+    assert np.all(sf_midx_int.loc[0].todense().values == sf_midx.data[4, :])
 
 
 def test_set_index(sf_midx):
@@ -619,13 +618,18 @@ def test_npz_io_s3(complex_example):
 
 
 def test_getitem():
-    sf = SparseFrame(np.identity(10), columns=list('abcdefghij'))
+    id_ = np.identity(10)
+    sf = SparseFrame(id_, columns=list('abcdefghij'))
     assert sf['a'].data.todense()[0] == 1
     assert sf['j'].data.todense()[9] == 1
+    assert np.all(sf[['a', 'b']].data.todense() == np.asmatrix(id_[:, [0, 1]]))
     tmp = sf[['j', 'a']].data.todense()
     assert tmp[9, 0] == 1
     assert tmp[0, 1] == 1
     assert (sf[list('abcdefghij')].data.todense() == np.identity(10)).all()
+    assert sf[[]].shape == (10, 0)
+    assert len(sf[[]].columns) == 0
+    assert isinstance(sf.columns, type(sf[[]].columns))
 
 
 def test_vstack():
@@ -895,4 +899,17 @@ def test_empty_elemwise():
     assert np.all(res == sf.data.todense())
 
     with pytest.raises(ValueError):
-        res = sf.add(sf_empty, fill_value=None)
+        res = sf.add(sf_empty, fill_value=None)
+
+
+def test_loc_duplicate_index():
+    sf = SparseFrame(np.identity(5),
+                     columns=list('UUXYZ'),
+                     index=list('AAABB'))
+    assert len(sf.loc['A'].index) == 3
+    assert len(sf.loc['B'].index) == 2
+    assert np.all(sf.loc['A'].todense().values == np.identity(5)[:3])
+    assert np.all(sf.loc['B'].todense().values == np.identity(5)[3:])
+
+    assert len(sf.loc[:, 'U'].columns) == 2
+    assert np.all(sf.loc[:, 'U'].todense().values == np.identity(5)[:, :2])