forked from wizardforcel/pandas-doc-zh
-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathinternals.html
233 lines (210 loc) · 25.9 KB
/
internals.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
<span id="id1"></span><h1><span class="yiyi-st" id="yiyi-52">Internals</span></h1>
<blockquote>
<p>原文：<a href="http://pandas.pydata.org/pandas-docs/stable/internals.html">http://pandas.pydata.org/pandas-docs/stable/internals.html</a></p>
<p>译者：<a href="https://github.com/wizardforcel">飞龙</a> <a href="http://usyiyi.cn/">UsyiyiCN</a></p>
<p>校对：（虚位以待）</p>
</blockquote>
<p><span class="yiyi-st" id="yiyi-53">本节将介绍一些大熊猫内部。</span></p>
<div class="section" id="indexing">
<h2><span class="yiyi-st" id="yiyi-54">Indexing</span></h2>
<p><span class="yiyi-st" id="yiyi-55">在大熊猫中有一些实现的对象可以作为轴标签的有效容器：</span></p>
<ul class="simple">
<li><span class="yiyi-st" id="yiyi-56"><code class="docutils literal"><span class="pre">Index</span></code>：通用的“有序集合”对象，对象类型的ndarray，不考虑其内容。</span><span class="yiyi-st" id="yiyi-57">标签必须是可哈希的（并且可能是不可变的）和唯一的。</span><span class="yiyi-st" id="yiyi-58">在Cython中填充标签的位置以执行<code class="docutils literal"><span class="pre">O(1)</span></code>查找。</span></li>
<li><span class="yiyi-st" id="yiyi-59"><code class="docutils literal"><span class="pre">Int64Index</span></code>：针对64位整数数据（例如时间戳）高度优化的<code class="docutils literal"><span class="pre">Index</span></code>版本</span></li>
<li><span class="yiyi-st" id="yiyi-60"><code class="docutils literal"><span class="pre">Float64Index</span></code>：针对64位浮点数据进行高度优化的<code class="docutils literal"><span class="pre">Index</span></code>版本</span></li>
<li><span class="yiyi-st" id="yiyi-61"><code class="docutils literal"><span class="pre">MultiIndex</span></code>：标准层次索引对象</span></li>
<li><span class="yiyi-st" id="yiyi-62"><code class="docutils literal"><span class="pre">DatetimeIndex</span></code>：带有<code class="docutils literal"><span class="pre">Timestamp</span></code>框元素的索引对象（impl是int64值）</span></li>
<li><span class="yiyi-st" id="yiyi-63"><code class="docutils literal"><span class="pre">TimedeltaIndex</span></code>：具有<code class="docutils literal"><span class="pre">Timedelta</span></code>框元素的索引对象（impl是in64值）</span></li>
<li><span class="yiyi-st" id="yiyi-64"><code class="docutils literal"><span class="pre">PeriodIndex</span></code>：包含Period元素的索引对象</span></li>
</ul>
<p><span class="yiyi-st" id="yiyi-65">有一些功能使得创建常规索引变得容易：</span></p>
<ul class="simple">
<li><span class="yiyi-st" id="yiyi-66"><code class="docutils literal"><span class="pre">date_range</span></code>：从时间规则或DateOffset生成的固定频率日期范围。</span><span class="yiyi-st" id="yiyi-67">Python datetime对象的ndarray</span></li>
<li><span class="yiyi-st" id="yiyi-68"><code class="docutils literal"><span class="pre">period_range</span></code>：从时间规则或DateOffset生成的固定频率日期范围。</span><span class="yiyi-st" id="yiyi-69"><code class="docutils literal"><span class="pre">Period</span></code>对象的数组，表示时间段</span></li>
</ul>
<p><span class="yiyi-st" id="yiyi-70">首先具有<code class="docutils literal"><span class="pre">Index</span></code>类的动机是启用不同的索引实现。</span><span class="yiyi-st" id="yiyi-71">这意味着，用户可以实现一个自定义<code class="docutils literal"><span class="pre">Index</span></code>子类，它可能更适合于特定应用程序，而不是在pandas中提供的。</span></p>
<p><span class="yiyi-st" id="yiyi-72">从内部实现的角度来看，<code class="docutils literal"><span class="pre">Index</span></code>必须定义的相关方法是以下一个或多个（取决于新对象内部与<code class="docutils literal"><span class="pre">Index</span></code>功能）：</span></p>
<ul class="simple">
<li><span class="yiyi-st" id="yiyi-73"><code class="docutils literal"><span class="pre">get_loc</span></code>：为标签返回“indexer”（整数，或在某些情况下为slice对象）</span></li>
<li><span class="yiyi-st" id="yiyi-74"><code class="docutils literal"><span class="pre">slice_locs</span></code>：返回“范围”到两个标签之间的切片</span></li>
<li><span class="yiyi-st" id="yiyi-75"><code class="docutils literal"><span class="pre">get_indexer</span></code>：计算用于重建索引/数据对齐目的的索引向量。</span><span class="yiyi-st" id="yiyi-76">有关更多信息，请参阅source / docstrings</span></li>
<li><span class="yiyi-st" id="yiyi-77"><code class="docutils literal"><span class="pre">get_indexer_non_unique</span></code>：当索引非唯一时，计算索引向量以进行重建索引/数据对齐。</span><span class="yiyi-st" id="yiyi-78">有关更多信息，请参阅source / docstrings</span></li>
<li><span class="yiyi-st" id="yiyi-79"><code class="docutils literal"><span class="pre">reindex</span></code>：输入索引的任何预转换然后调用<code class="docutils literal"><span class="pre">get_indexer</span></code></span></li>
<li><span class="yiyi-st" id="yiyi-80"><code class="docutils literal"><span class="pre">union</span></code>，<code class="docutils literal"><span class="pre">intersection</span></code>：计算两个Index对象的并集或交集</span></li>
<li><span class="yiyi-st" id="yiyi-81"><code class="docutils literal"><span class="pre">insert</span></code>：在索引中插入一个新标签，产生一个新对象</span></li>
<li><span class="yiyi-st" id="yiyi-82"><code class="docutils literal"><span class="pre">delete</span></code>：删除标签，生成一个新对象</span></li>
<li><span class="yiyi-st" id="yiyi-83"><code class="docutils literal"><span class="pre">drop</span></code>：删除一组标签</span></li>
<li><span class="yiyi-st" id="yiyi-84"><code class="docutils literal"><span class="pre">take</span></code>：类似于ndarray.take</span></li>
</ul>
<div class="section" id="multiindex">
<h3><span class="yiyi-st" id="yiyi-85">MultiIndex</span></h3>
<p><span class="yiyi-st" id="yiyi-86">在内部，<code class="docutils literal"><span class="pre">MultiIndex</span></code>包含以下几项内容：<strong>级别</strong>，整数<strong>标签</strong>和级别<strong>名称</strong>：</span></p>
<div class="highlight-ipython"><div class="highlight"><pre><span></span><span class="gp">In [1]: </span><span class="n">index</span> <span class="o">=</span> <span class="n">pd</span><span class="o">.</span><span class="n">MultiIndex</span><span class="o">.</span><span class="n">from_product</span><span class="p">([</span><span class="nb">range</span><span class="p">(</span><span class="mi">3</span><span class="p">),</span> <span class="p">[</span><span class="s1">'one'</span><span class="p">,</span> <span class="s1">'two'</span><span class="p">]],</span> <span class="n">names</span><span class="o">=</span><span class="p">[</span><span class="s1">'first'</span><span class="p">,</span> <span class="s1">'second'</span><span class="p">])</span>
<span class="gp">In [2]: </span><span class="n">index</span>
<span class="gr">Out[2]: </span>
<span class="go">MultiIndex(levels=[[0, 1, 2], [u'one', u'two']],</span>
<span class="go"> labels=[[0, 0, 1, 1, 2, 2], [0, 1, 0, 1, 0, 1]],</span>
<span class="go"> names=[u'first', u'second'])</span>
<span class="gp">In [3]: </span><span class="n">index</span><span class="o">.</span><span class="n">levels</span>
<span class="gr">Out[3]: </span><span class="n">FrozenList</span><span class="p">([[</span><span class="mi">0</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">2</span><span class="p">],</span> <span class="p">[</span><span class="s1">u'one'</span><span class="p">,</span> <span class="s1">u'two'</span><span class="p">]])</span>
<span class="gp">In [4]: </span><span class="n">index</span><span class="o">.</span><span class="n">labels</span>
<span class="gr">Out[4]: </span><span class="n">FrozenList</span><span class="p">([[</span><span class="mi">0</span><span class="p">,</span> <span class="mi">0</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="mi">2</span><span class="p">],</span> <span class="p">[</span><span class="mi">0</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">0</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">0</span><span class="p">,</span> <span class="mi">1</span><span class="p">]])</span>
<span class="gp">In [5]: </span><span class="n">index</span><span class="o">.</span><span class="n">names</span>
<span class="gr">Out[5]: </span><span class="n">FrozenList</span><span class="p">([</span><span class="s1">u'first'</span><span class="p">,</span> <span class="s1">u'second'</span><span class="p">])</span>
</pre></div>
</div>
<p><span class="yiyi-st" id="yiyi-87">您可能猜测标签确定在索引的每个层上用该位置标识哪个唯一元素。</span><span class="yiyi-st" id="yiyi-88">It’s important to note that sortedness is determined <strong>solely</strong> from the integer labels and does not check (or care) whether the levels themselves are sorted. </span><span class="yiyi-st" id="yiyi-89">幸运的是，构造函数<code class="docutils literal"><span class="pre">from_tuples</span></code>和<code class="docutils literal"><span class="pre">from_arrays</span></code>确保这是真的，但如果你自己计算级别和标签，请小心。</span></p>
</div>
</div>
<div class="section" id="subclassing-pandas-data-structures">
<span id="ref-subclassing-pandas"></span><h2><span class="yiyi-st" id="yiyi-90">Subclassing pandas Data Structures</span></h2>
<div class="admonition warning">
<p class="first admonition-title"><span class="yiyi-st" id="yiyi-91">警告</span></p>
<p><span class="yiyi-st" id="yiyi-92">在考虑子类化<code class="docutils literal"><span class="pre">pandas</span></code>数据结构之前，有一些更简单的替代方法。</span></p>
<ol class="last arabic simple">
<li><span class="yiyi-st" id="yiyi-93">具有<a class="reference internal" href="basics.html#basics-pipe"><span class="std std-ref">pipe</span></a></span></li>
<li><span class="yiyi-st" id="yiyi-94">使用<em>组合</em>。</span><span class="yiyi-st" id="yiyi-95">请参阅<a class="reference external" href="http://en.wikipedia.org/wiki/Composition_over_inheritance">此处</a>。</span></li>
</ol>
</div>
<p><span class="yiyi-st" id="yiyi-96">本节介绍如何子类化<code class="docutils literal"><span class="pre">pandas</span></code>数据结构以满足更具体的需求。</span><span class="yiyi-st" id="yiyi-97">有2分需要注意：</span></p>
<ol class="arabic simple">
<li><span class="yiyi-st" id="yiyi-98">覆盖构造函数属性。</span></li>
<li><span class="yiyi-st" id="yiyi-99">定义原始属性</span></li>
</ol>
<div class="admonition note">
<p class="first admonition-title"><span class="yiyi-st" id="yiyi-100">注意</span></p>
<p class="last"><span class="yiyi-st" id="yiyi-101">你可以在<a class="reference external" href="https://github.com/geopandas/geopandas">geopandas</a>项目中找到一个很好的例子。</span></p>
</div>
<div class="section" id="override-constructor-properties">
<h3><span class="yiyi-st" id="yiyi-102">Override Constructor Properties</span></h3>
<p><span class="yiyi-st" id="yiyi-103">每个数据结构都具有指定数据构造函数的构造函数属性。</span><span class="yiyi-st" id="yiyi-104">通过覆盖这些属性，您可以通过<code class="docutils literal"><span class="pre">pandas</span></code>数据操作来保留定义类。</span></p>
<p><span class="yiyi-st" id="yiyi-105">有3个构造函数要定义：</span></p>
<ul class="simple">
<li><span class="yiyi-st" id="yiyi-106"><code class="docutils literal"><span class="pre">_constructor</span></code>：当操作结果与原始操作结果具有相同的缩放时使用。</span></li>
<li><span class="yiyi-st" id="yiyi-107"><code class="docutils literal"><span class="pre">_constructor_sliced</span></code>：当操作结果具有一个较低维度作为原始值时使用，例如<code class="docutils literal"><span class="pre">DataFrame</span></code>单列切片。</span></li>
<li><span class="yiyi-st" id="yiyi-108"><code class="docutils literal"><span class="pre">_constructor_expanddim</span></code>：当操作结果具有一个较高维度作为原始值时使用，例如<code class="docutils literal"><span class="pre">Series.to_frame()</span></code>和<code class="docutils literal"><span class="pre">DataFrame.to_panel()</span></code>。</span></li>
</ul>
<p><span class="yiyi-st" id="yiyi-109">下表显示了默认情况下<code class="docutils literal"><span class="pre">pandas</span></code>数据结构如何定义构造函数属性。</span></p>
<table border="1" class="docutils">
<colgroup>
<col width="29%">
<col width="25%">
<col width="21%">
<col width="25%">
</colgroup>
<thead valign="bottom">
<tr class="row-odd"><th class="head"><span class="yiyi-st" id="yiyi-110">属性属性</span></th>
<th class="head"><span class="yiyi-st" id="yiyi-111"><code class="docutils literal"><span class="pre">Series</span></code></span></th>
<th class="head"><span class="yiyi-st" id="yiyi-112"><code class="docutils literal"><span class="pre">DataFrame</span></code></span></th>
<th class="head"><span class="yiyi-st" id="yiyi-113"><code class="docutils literal"><span class="pre">Panel</span></code></span></th>
</tr>
</thead>
<tbody valign="top">
<tr class="row-even"><td><span class="yiyi-st" id="yiyi-114"><code class="docutils literal"><span class="pre">_constructor</span></code></span></td>
<td><span class="yiyi-st" id="yiyi-115"><code class="docutils literal"><span class="pre">Series</span></code></span></td>
<td><span class="yiyi-st" id="yiyi-116"><code class="docutils literal"><span class="pre">DataFrame</span></code></span></td>
<td><span class="yiyi-st" id="yiyi-117"><code class="docutils literal"><span class="pre">Panel</span></code></span></td>
</tr>
<tr class="row-odd"><td><span class="yiyi-st" id="yiyi-118"><code class="docutils literal"><span class="pre">_constructor_sliced</span></code></span></td>
<td><span class="yiyi-st" id="yiyi-119"><code class="docutils literal"><span class="pre">NotImplementedError</span></code></span></td>
<td><span class="yiyi-st" id="yiyi-120"><code class="docutils literal"><span class="pre">Series</span></code></span></td>
<td><span class="yiyi-st" id="yiyi-121"><code class="docutils literal"><span class="pre">DataFrame</span></code></span></td>
</tr>
<tr class="row-even"><td><span class="yiyi-st" id="yiyi-122"><code class="docutils literal"><span class="pre">_constructor_expanddim</span></code></span></td>
<td><span class="yiyi-st" id="yiyi-123"><code class="docutils literal"><span class="pre">DataFrame</span></code></span></td>
<td><span class="yiyi-st" id="yiyi-124"><code class="docutils literal"><span class="pre">Panel</span></code></span></td>
<td><span class="yiyi-st" id="yiyi-125"><code class="docutils literal"><span class="pre">NotImplementedError</span></code></span></td>
</tr>
</tbody>
</table>
<p><span class="yiyi-st" id="yiyi-126">下面的示例显示如何定义<code class="docutils literal"><span class="pre">SubclassedSeries</span></code>和<code class="docutils literal"><span class="pre">SubclassedDataFrame</span></code>覆盖构造函数属性。</span></p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">class</span> <span class="nc">SubclassedSeries</span><span class="p">(</span><span class="n">Series</span><span class="p">):</span>
<span class="nd">@property</span>
<span class="k">def</span> <span class="nf">_constructor</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
<span class="k">return</span> <span class="n">SubclassedSeries</span>
<span class="nd">@property</span>
<span class="k">def</span> <span class="nf">_constructor_expanddim</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
<span class="k">return</span> <span class="n">SubclassedDataFrame</span>
<span class="k">class</span> <span class="nc">SubclassedDataFrame</span><span class="p">(</span><span class="n">DataFrame</span><span class="p">):</span>
<span class="nd">@property</span>
<span class="k">def</span> <span class="nf">_constructor</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
<span class="k">return</span> <span class="n">SubclassedDataFrame</span>
<span class="nd">@property</span>
<span class="k">def</span> <span class="nf">_constructor_sliced</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
<span class="k">return</span> <span class="n">SubclassedSeries</span>
</pre></div>
</div>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="gp">>>> </span><span class="n">s</span> <span class="o">=</span> <span class="n">SubclassedSeries</span><span class="p">([</span><span class="mi">1</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="mi">3</span><span class="p">])</span>
<span class="gp">>>> </span><span class="nb">type</span><span class="p">(</span><span class="n">s</span><span class="p">)</span>
<span class="go"><class '__main__.SubclassedSeries'></span>
<span class="gp">>>> </span><span class="n">to_framed</span> <span class="o">=</span> <span class="n">s</span><span class="o">.</span><span class="n">to_frame</span><span class="p">()</span>
<span class="gp">>>> </span><span class="nb">type</span><span class="p">(</span><span class="n">to_framed</span><span class="p">)</span>
<span class="go"><class '__main__.SubclassedDataFrame'></span>
<span class="gp">>>> </span><span class="n">df</span> <span class="o">=</span> <span class="n">SubclassedDataFrame</span><span class="p">({</span><span class="s1">'A'</span><span class="p">,</span> <span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="mi">3</span><span class="p">],</span> <span class="s1">'B'</span><span class="p">:</span> <span class="p">[</span><span class="mi">4</span><span class="p">,</span> <span class="mi">5</span><span class="p">,</span> <span class="mi">6</span><span class="p">],</span> <span class="s1">'C'</span><span class="p">:</span> <span class="p">[</span><span class="mi">7</span><span class="p">,</span> <span class="mi">8</span><span class="p">,</span> <span class="mi">9</span><span class="p">]})</span>
<span class="gp">>>> </span><span class="n">df</span>
<span class="go"> A B C</span>
<span class="go">0 1 4 7</span>
<span class="go">1 2 5 8</span>
<span class="go">2 3 6 9</span>
<span class="gp">>>> </span><span class="nb">type</span><span class="p">(</span><span class="n">df</span><span class="p">)</span>
<span class="go"><class '__main__.SubclassedDataFrame'></span>
<span class="gp">>>> </span><span class="n">sliced1</span> <span class="o">=</span> <span class="n">df</span><span class="p">[[</span><span class="s1">'A'</span><span class="p">,</span> <span class="s1">'B'</span><span class="p">]]</span>
<span class="gp">>>> </span><span class="n">sliced1</span>
<span class="go"> A B</span>
<span class="go">0 1 4</span>
<span class="go">1 2 5</span>
<span class="go">2 3 6</span>
<span class="gp">>>> </span><span class="nb">type</span><span class="p">(</span><span class="n">sliced1</span><span class="p">)</span>
<span class="go"><class '__main__.SubclassedDataFrame'></span>
<span class="gp">>>> </span><span class="n">sliced2</span> <span class="o">=</span> <span class="n">df</span><span class="p">[</span><span class="s1">'A'</span><span class="p">]</span>
<span class="gp">>>> </span><span class="n">sliced2</span>
<span class="go">0 1</span>
<span class="go">1 2</span>
<span class="go">2 3</span>
<span class="go">Name: A, dtype: int64</span>
<span class="gp">>>> </span><span class="nb">type</span><span class="p">(</span><span class="n">sliced2</span><span class="p">)</span>
<span class="go"><class '__main__.SubclassedSeries'></span>
</pre></div>
</div>
</div>
<div class="section" id="define-original-properties">
<h3><span class="yiyi-st" id="yiyi-127">Define Original Properties</span></h3>
<p><span class="yiyi-st" id="yiyi-128">要让原始数据结构具有其他属性，您应该让<code class="docutils literal"><span class="pre">pandas</span></code>知道添加了什么属性。</span><span class="yiyi-st" id="yiyi-129"><code class="docutils literal"><span class="pre">pandas</span></code>将未知属性映射到覆盖<code class="docutils literal"><span class="pre">__getattribute__</span></code>的数据名称。</span><span class="yiyi-st" id="yiyi-130">定义原始属性可以通过以下两种方式之一完成：</span></p>
<ol class="arabic simple">
<li><span class="yiyi-st" id="yiyi-131">为不会传递到操作结果的临时属性定义<code class="docutils literal"><span class="pre">_internal_names</span></code>和<code class="docutils literal"><span class="pre">_internal_names_set</span></code>。</span></li>
<li><span class="yiyi-st" id="yiyi-132">为将传递到操作结果的正常属性定义<code class="docutils literal"><span class="pre">_metadata</span></code>。</span></li>
</ol>
<p><span class="yiyi-st" id="yiyi-133">下面是一个定义2个原始属性的示例，“internal_cache”作为临时属性，“added_property”作为正常属性</span></p>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="k">class</span> <span class="nc">SubclassedDataFrame2</span><span class="p">(</span><span class="n">DataFrame</span><span class="p">):</span>
<span class="c1"># temporary properties</span>
<span class="n">_internal_names</span> <span class="o">=</span> <span class="n">pd</span><span class="o">.</span><span class="n">DataFrame</span><span class="o">.</span><span class="n">_internal_names</span> <span class="o">+</span> <span class="p">[</span><span class="s1">'internal_cache'</span><span class="p">]</span>
<span class="n">_internal_names_set</span> <span class="o">=</span> <span class="nb">set</span><span class="p">(</span><span class="n">_internal_names</span><span class="p">)</span>
<span class="c1"># normal properties</span>
<span class="n">_metadata</span> <span class="o">=</span> <span class="p">[</span><span class="s1">'added_property'</span><span class="p">]</span>
<span class="nd">@property</span>
<span class="k">def</span> <span class="nf">_constructor</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
<span class="k">return</span> <span class="n">SubclassedDataFrame2</span>
</pre></div>
</div>
<div class="highlight-python"><div class="highlight"><pre><span></span><span class="gp">>>> </span><span class="n">df</span> <span class="o">=</span> <span class="n">SubclassedDataFrame2</span><span class="p">({</span><span class="s1">'A'</span><span class="p">,</span> <span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="mi">3</span><span class="p">],</span> <span class="s1">'B'</span><span class="p">:</span> <span class="p">[</span><span class="mi">4</span><span class="p">,</span> <span class="mi">5</span><span class="p">,</span> <span class="mi">6</span><span class="p">],</span> <span class="s1">'C'</span><span class="p">:</span> <span class="p">[</span><span class="mi">7</span><span class="p">,</span> <span class="mi">8</span><span class="p">,</span> <span class="mi">9</span><span class="p">]})</span>
<span class="gp">>>> </span><span class="n">df</span>
<span class="go"> A B C</span>
<span class="go">0 1 4 7</span>
<span class="go">1 2 5 8</span>
<span class="go">2 3 6 9</span>
<span class="gp">>>> </span><span class="n">df</span><span class="o">.</span><span class="n">internal_cache</span> <span class="o">=</span> <span class="s1">'cached'</span>
<span class="gp">>>> </span><span class="n">df</span><span class="o">.</span><span class="n">added_property</span> <span class="o">=</span> <span class="s1">'property'</span>
<span class="gp">>>> </span><span class="n">df</span><span class="o">.</span><span class="n">internal_cache</span>
<span class="go">cached</span>
<span class="gp">>>> </span><span class="n">df</span><span class="o">.</span><span class="n">added_property</span>
<span class="go">property</span>
<span class="go"># properties defined in _internal_names is reset after manipulation</span>
<span class="gp">>>> </span><span class="n">df</span><span class="p">[[</span><span class="s1">'A'</span><span class="p">,</span> <span class="s1">'B'</span><span class="p">]]</span><span class="o">.</span><span class="n">internal_cache</span>
<span class="go">AttributeError: 'SubclassedDataFrame2' object has no attribute 'internal_cache'</span>
<span class="go"># properties defined in _metadata are retained</span>
<span class="gp">>>> </span><span class="n">df</span><span class="p">[[</span><span class="s1">'A'</span><span class="p">,</span> <span class="s1">'B'</span><span class="p">]]</span><span class="o">.</span><span class="n">added_property</span>
<span class="go">property</span>
</pre></div>
</div>
</div>
</div>