This repository has been archived by the owner on May 6, 2021. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 87
/
sparse.html
436 lines (390 loc) · 40.3 KB
/
sparse.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
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
<span id="sparse"></span><h1><span class="yiyi-st" id="yiyi-53">Sparse data structures</span></h1>
<blockquote>
<p>原文：<a href="http://pandas.pydata.org/pandas-docs/stable/sparse.html">http://pandas.pydata.org/pandas-docs/stable/sparse.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>
<div class="admonition note">
<p class="first admonition-title"><span class="yiyi-st" id="yiyi-54">注意</span></p>
<p class="last"><span class="yiyi-st" id="yiyi-55">在0.19.0中已删除<code class="docutils literal"><span class="pre">SparsePanel</span></code>类</span></p>
</div>
<p><span class="yiyi-st" id="yiyi-56">我们实现了“稀疏”版本的Series和DataFrame。</span><span class="yiyi-st" id="yiyi-57">这些在典型的“大多为0”中不稀疏。</span><span class="yiyi-st" id="yiyi-58">相反，您可以将这些对象视为“压缩”，其中省略任何匹配特定值（<code class="docutils literal"><span class="pre">NaN</span></code> /缺失值，尽管可以选择任何值）的数据。</span><span class="yiyi-st" id="yiyi-59">特殊的<code class="docutils literal"><span class="pre">SparseIndex</span></code>对象跟踪数据已被“稀疏化”的位置。</span><span class="yiyi-st" id="yiyi-60">在一个例子中，这将更有意义。</span><span class="yiyi-st" id="yiyi-61">所有标准的熊猫数据结构都有一个<code class="docutils literal"><span class="pre">to_sparse</span></code>方法：</span></p>
<div class="highlight-ipython"><div class="highlight"><pre><span></span><span class="gp">In [1]: </span><span class="n">ts</span> <span class="o">=</span> <span class="n">pd</span><span class="o">.</span><span class="n">Series</span><span class="p">(</span><span class="n">randn</span><span class="p">(</span><span class="mi">10</span><span class="p">))</span>
<span class="gp">In [2]: </span><span class="n">ts</span><span class="p">[</span><span class="mi">2</span><span class="p">:</span><span class="o">-</span><span class="mi">2</span><span class="p">]</span> <span class="o">=</span> <span class="n">np</span><span class="o">.</span><span class="n">nan</span>
<span class="gp">In [3]: </span><span class="n">sts</span> <span class="o">=</span> <span class="n">ts</span><span class="o">.</span><span class="n">to_sparse</span><span class="p">()</span>
<span class="gp">In [4]: </span><span class="n">sts</span>
<span class="gr">Out[4]: </span>
<span class="go">0 0.469112</span>
<span class="go">1 -0.282863</span>
<span class="go">2 NaN</span>
<span class="go">3 NaN</span>
<span class="go">4 NaN</span>
<span class="go">5 NaN</span>
<span class="go">6 NaN</span>
<span class="go">7 NaN</span>
<span class="go">8 -0.861849</span>
<span class="go">9 -2.104569</span>
<span class="go">dtype: float64</span>
<span class="go">BlockIndex</span>
<span class="go">Block locations: array([0, 8], dtype=int32)</span>
<span class="go">Block lengths: array([2, 2], dtype=int32)</span>
</pre></div>
</div>
<p><span class="yiyi-st" id="yiyi-62"><code class="docutils literal"><span class="pre">to_sparse</span></code>方法采用<code class="docutils literal"><span class="pre">kind</span></code>参数（对于稀疏索引，请参见下文）和<code class="docutils literal"><span class="pre">fill_value</span></code>。</span><span class="yiyi-st" id="yiyi-63">所以如果我们有一个大多数为零的系列，我们可以将它转换为稀疏与<code class="docutils literal"><span class="pre">fill_value=0</span></code>：</span></p>
<div class="highlight-ipython"><div class="highlight"><pre><span></span><span class="gp">In [5]: </span><span class="n">ts</span><span class="o">.</span><span class="n">fillna</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span><span class="o">.</span><span class="n">to_sparse</span><span class="p">(</span><span class="n">fill_value</span><span class="o">=</span><span class="mi">0</span><span class="p">)</span>
<span class="gr">Out[5]: </span>
<span class="go">0 0.469112</span>
<span class="go">1 -0.282863</span>
<span class="go">2 0.000000</span>
<span class="go">3 0.000000</span>
<span class="go">4 0.000000</span>
<span class="go">5 0.000000</span>
<span class="go">6 0.000000</span>
<span class="go">7 0.000000</span>
<span class="go">8 -0.861849</span>
<span class="go">9 -2.104569</span>
<span class="go">dtype: float64</span>
<span class="go">BlockIndex</span>
<span class="go">Block locations: array([0, 8], dtype=int32)</span>
<span class="go">Block lengths: array([2, 2], dtype=int32)</span>
</pre></div>
</div>
<p><span class="yiyi-st" id="yiyi-64">稀疏对象存在是为了内存效率的原因。</span><span class="yiyi-st" id="yiyi-65">假设你有一个大的，主要是NA DataFrame：</span></p>
<div class="highlight-ipython"><div class="highlight"><pre><span></span><span class="gp">In [6]: </span><span class="n">df</span> <span class="o">=</span> <span class="n">pd</span><span class="o">.</span><span class="n">DataFrame</span><span class="p">(</span><span class="n">randn</span><span class="p">(</span><span class="mi">10000</span><span class="p">,</span> <span class="mi">4</span><span class="p">))</span>
<span class="gp">In [7]: </span><span class="n">df</span><span class="o">.</span><span class="n">ix</span><span class="p">[:</span><span class="mi">9998</span><span class="p">]</span> <span class="o">=</span> <span class="n">np</span><span class="o">.</span><span class="n">nan</span>
<span class="gp">In [8]: </span><span class="n">sdf</span> <span class="o">=</span> <span class="n">df</span><span class="o">.</span><span class="n">to_sparse</span><span class="p">()</span>
<span class="gp">In [9]: </span><span class="n">sdf</span>
<span class="gr">Out[9]: </span>
<span class="go"> 0 1 2 3</span>
<span class="go">0 NaN NaN NaN NaN</span>
<span class="go">1 NaN NaN NaN NaN</span>
<span class="go">2 NaN NaN NaN NaN</span>
<span class="go">3 NaN NaN NaN NaN</span>
<span class="go">4 NaN NaN NaN NaN</span>
<span class="go">5 NaN NaN NaN NaN</span>
<span class="go">6 NaN NaN NaN NaN</span>
<span class="go">... ... ... ... ...</span>
<span class="go">9993 NaN NaN NaN NaN</span>
<span class="go">9994 NaN NaN NaN NaN</span>
<span class="go">9995 NaN NaN NaN NaN</span>
<span class="go">9996 NaN NaN NaN NaN</span>
<span class="go">9997 NaN NaN NaN NaN</span>
<span class="go">9998 NaN NaN NaN NaN</span>
<span class="go">9999 0.280249 -1.648493 1.490865 -0.890819</span>
<span class="go">[10000 rows x 4 columns]</span>
<span class="gp">In [10]: </span><span class="n">sdf</span><span class="o">.</span><span class="n">density</span>
<span class="gr">Out[10]: </span><span class="mf">0.0001</span>
</pre></div>
</div>
<p><span class="yiyi-st" id="yiyi-66">如你所见，密度（未被“压缩”的值的百分比）非常低。</span><span class="yiyi-st" id="yiyi-67">这个稀疏对象在磁盘（pickled）和Python解释器中占用更少的内存。</span><span class="yiyi-st" id="yiyi-68">在功能上，它们的行为应该与它们的稠密对应物几乎相同。</span></p>
<p><span class="yiyi-st" id="yiyi-69">任何稀疏对象都可以通过调用<code class="docutils literal"><span class="pre">to_dense</span></code>转换回标准密集形式：</span></p>
<div class="highlight-ipython"><div class="highlight"><pre><span></span><span class="gp">In [11]: </span><span class="n">sts</span><span class="o">.</span><span class="n">to_dense</span><span class="p">()</span>
<span class="gr">Out[11]: </span>
<span class="go">0 0.469112</span>
<span class="go">1 -0.282863</span>
<span class="go">2 NaN</span>
<span class="go">3 NaN</span>
<span class="go">4 NaN</span>
<span class="go">5 NaN</span>
<span class="go">6 NaN</span>
<span class="go">7 NaN</span>
<span class="go">8 -0.861849</span>
<span class="go">9 -2.104569</span>
<span class="go">dtype: float64</span>
</pre></div>
</div>
<div class="section" id="sparsearray">
<span id="sparse-array"></span><h2><span class="yiyi-st" id="yiyi-70">SparseArray</span></h2>
<p><span class="yiyi-st" id="yiyi-71"><code class="docutils literal"><span class="pre">SparseArray</span></code>是所有稀疏索引数据结构的基本层。</span><span class="yiyi-st" id="yiyi-72">它是一个1维的ndarray样对象，只存储不同于<code class="docutils literal"><span class="pre">fill_value</span></code>的值：</span></p>
<div class="highlight-ipython"><div class="highlight"><pre><span></span><span class="gp">In [12]: </span><span class="n">arr</span> <span class="o">=</span> <span class="n">np</span><span class="o">.</span><span class="n">random</span><span class="o">.</span><span class="n">randn</span><span class="p">(</span><span class="mi">10</span><span class="p">)</span>
<span class="gp">In [13]: </span><span class="n">arr</span><span class="p">[</span><span class="mi">2</span><span class="p">:</span><span class="mi">5</span><span class="p">]</span> <span class="o">=</span> <span class="n">np</span><span class="o">.</span><span class="n">nan</span><span class="p">;</span> <span class="n">arr</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="o">=</span> <span class="n">np</span><span class="o">.</span><span class="n">nan</span>
<span class="gp">In [14]: </span><span class="n">sparr</span> <span class="o">=</span> <span class="n">pd</span><span class="o">.</span><span class="n">SparseArray</span><span class="p">(</span><span class="n">arr</span><span class="p">)</span>
<span class="gp">In [15]: </span><span class="n">sparr</span>
<span class="gr">Out[15]: </span>
<span class="go">[-1.95566352972, -1.6588664276, nan, nan, nan, 1.15893288864, 0.145297113733, nan, 0.606027190513, 1.33421134013]</span>
<span class="go">Fill: nan</span>
<span class="go">IntIndex</span>
<span class="go">Indices: array([0, 1, 5, 6, 8, 9], dtype=int32)</span>
</pre></div>
</div>
<p><span class="yiyi-st" id="yiyi-73">像索引对象（SparseSeries，SparseDataFrame）一样，通过调用<code class="docutils literal"><span class="pre">to_dense</span></code>可以将<code class="docutils literal"><span class="pre">SparseArray</span></code>转换回常规的ndarray：</span></p>
<div class="highlight-ipython"><div class="highlight"><pre><span></span><span class="gp">In [16]: </span><span class="n">sparr</span><span class="o">.</span><span class="n">to_dense</span><span class="p">()</span>
<span class="gr">Out[16]: </span>
<span class="go">array([-1.9557, -1.6589, nan, nan, nan, 1.1589, 0.1453,</span>
<span class="go"> nan, 0.606 , 1.3342])</span>
</pre></div>
</div>
</div>
<div class="section" id="sparselist">
<span id="sparse-list"></span><h2><span class="yiyi-st" id="yiyi-74">SparseList</span></h2>
<p><span class="yiyi-st" id="yiyi-75"><code class="docutils literal"><span class="pre">SparseList</span></code>类已弃用，将在以后的版本中删除。</span><span class="yiyi-st" id="yiyi-76">有关<code class="docutils literal"><span class="pre">SparseList</span></code>的文档，请参见以前版本的<a class="reference external" href="http://pandas.pydata.org/pandas-docs/version/0.18.1/sparse.html#sparselist">文档</a>。</span></p>
</div>
<div class="section" id="sparseindex-objects">
<h2><span class="yiyi-st" id="yiyi-77">SparseIndex objects</span></h2>
<p><span class="yiyi-st" id="yiyi-78">实现了两种<code class="docutils literal"><span class="pre">SparseIndex</span></code>，<code class="docutils literal"><span class="pre">block</span></code>和<code class="docutils literal"><span class="pre">integer</span></code>。</span><span class="yiyi-st" id="yiyi-79">我们建议使用<code class="docutils literal"><span class="pre">block</span></code>，因为它更节省内存。</span><span class="yiyi-st" id="yiyi-80"><code class="docutils literal"><span class="pre">integer</span></code>格式保留数据不等于填充值的所有位置的数组。</span><span class="yiyi-st" id="yiyi-81"><code class="docutils literal"><span class="pre">block</span></code>格式只跟踪数据块的位置和大小。</span></p>
</div>
<div class="section" id="sparse-dtypes">
<span id="sparse-dtype"></span><h2><span class="yiyi-st" id="yiyi-82">Sparse Dtypes</span></h2>
<p><span class="yiyi-st" id="yiyi-83">稀疏数据应具有与其密集表示相同的dtype。</span><span class="yiyi-st" id="yiyi-84">目前，支持<code class="docutils literal"><span class="pre">float64</span></code>，<code class="docutils literal"><span class="pre">int64</span></code>和<code class="docutils literal"><span class="pre">bool</span></code> dtypes。</span><span class="yiyi-st" id="yiyi-85">根据原始dtype，<code class="docutils literal"><span class="pre">fill_value</span></code>默认更改：</span></p>
<ul class="simple">
<li><span class="yiyi-st" id="yiyi-86"><code class="docutils literal"><span class="pre">float64</span></code>：<code class="docutils literal"><span class="pre">np.nan</span></code></span></li>
<li><span class="yiyi-st" id="yiyi-87"><code class="docutils literal"><span class="pre">int64</span></code>：<code class="docutils literal"><span class="pre">0</span></code></span></li>
<li><span class="yiyi-st" id="yiyi-88"><code class="docutils literal"><span class="pre">bool</span></code>：<code class="docutils literal"><span class="pre">False</span></code></span></li>
</ul>
<div class="highlight-ipython"><div class="highlight"><pre><span></span><span class="gp">In [17]: </span><span class="n">s</span> <span class="o">=</span> <span class="n">pd</span><span class="o">.</span><span class="n">Series</span><span class="p">([</span><span class="mi">1</span><span class="p">,</span> <span class="n">np</span><span class="o">.</span><span class="n">nan</span><span class="p">,</span> <span class="n">np</span><span class="o">.</span><span class="n">nan</span><span class="p">])</span>
<span class="gp">In [18]: </span><span class="n">s</span>
<span class="gr">Out[18]: </span>
<span class="go">0 1.0</span>
<span class="go">1 NaN</span>
<span class="go">2 NaN</span>
<span class="go">dtype: float64</span>
<span class="gp">In [19]: </span><span class="n">s</span><span class="o">.</span><span class="n">to_sparse</span><span class="p">()</span>
<span class="gr">Out[19]: </span>
<span class="go">0 1.0</span>
<span class="go">1 NaN</span>
<span class="go">2 NaN</span>
<span class="go">dtype: float64</span>
<span class="go">BlockIndex</span>
<span class="go">Block locations: array([0], dtype=int32)</span>
<span class="go">Block lengths: array([1], dtype=int32)</span>
<span class="gp">In [20]: </span><span class="n">s</span> <span class="o">=</span> <span class="n">pd</span><span class="o">.</span><span class="n">Series</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">0</span><span class="p">])</span>
<span class="gp">In [21]: </span><span class="n">s</span>
<span class="gr">Out[21]: </span>
<span class="go">0 1</span>
<span class="go">1 0</span>
<span class="go">2 0</span>
<span class="go">dtype: int64</span>
<span class="gp">In [22]: </span><span class="n">s</span><span class="o">.</span><span class="n">to_sparse</span><span class="p">()</span>
<span class="gr">Out[22]: </span>
<span class="go">0 1</span>
<span class="go">1 0</span>
<span class="go">2 0</span>
<span class="go">dtype: int64</span>
<span class="go">BlockIndex</span>
<span class="go">Block locations: array([0], dtype=int32)</span>
<span class="go">Block lengths: array([1], dtype=int32)</span>
<span class="gp">In [23]: </span><span class="n">s</span> <span class="o">=</span> <span class="n">pd</span><span class="o">.</span><span class="n">Series</span><span class="p">([</span><span class="bp">True</span><span class="p">,</span> <span class="bp">False</span><span class="p">,</span> <span class="bp">True</span><span class="p">])</span>
<span class="gp">In [24]: </span><span class="n">s</span>
<span class="gr">Out[24]: </span>
<span class="go">0 True</span>
<span class="go">1 False</span>
<span class="go">2 True</span>
<span class="go">dtype: bool</span>
<span class="gp">In [25]: </span><span class="n">s</span><span class="o">.</span><span class="n">to_sparse</span><span class="p">()</span>
<span class="gr">Out[25]: </span>
<span class="go">0 True</span>
<span class="go">1 False</span>
<span class="go">2 True</span>
<span class="go">dtype: bool</span>
<span class="go">BlockIndex</span>
<span class="go">Block locations: array([0, 2], dtype=int32)</span>
<span class="go">Block lengths: array([1, 1], dtype=int32)</span>
</pre></div>
</div>
<p><span class="yiyi-st" id="yiyi-89">您可以使用<code class="docutils literal"><span class="pre">.astype()</span></code>更改dtype，结果也是稀疏的。</span><span class="yiyi-st" id="yiyi-90">请注意，<code class="docutils literal"><span class="pre">.astype()</span></code>也会影响<code class="docutils literal"><span class="pre">fill_value</span></code>以保持其密集表示。</span></p>
<div class="highlight-ipython"><div class="highlight"><pre><span></span><span class="gp">In [26]: </span><span class="n">s</span> <span class="o">=</span> <span class="n">pd</span><span class="o">.</span><span class="n">Series</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">0</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="gp">In [27]: </span><span class="n">s</span>
<span class="gr">Out[27]: </span>
<span class="go">0 1</span>
<span class="go">1 0</span>
<span class="go">2 0</span>
<span class="go">3 0</span>
<span class="go">4 0</span>
<span class="go">dtype: int64</span>
<span class="gp">In [28]: </span><span class="n">ss</span> <span class="o">=</span> <span class="n">s</span><span class="o">.</span><span class="n">to_sparse</span><span class="p">()</span>
<span class="gp">In [29]: </span><span class="n">ss</span>
<span class="gr">Out[29]: </span>
<span class="go">0 1</span>
<span class="go">1 0</span>
<span class="go">2 0</span>
<span class="go">3 0</span>
<span class="go">4 0</span>
<span class="go">dtype: int64</span>
<span class="go">BlockIndex</span>
<span class="go">Block locations: array([0], dtype=int32)</span>
<span class="go">Block lengths: array([1], dtype=int32)</span>
<span class="gp">In [30]: </span><span class="n">ss</span><span class="o">.</span><span class="n">astype</span><span class="p">(</span><span class="n">np</span><span class="o">.</span><span class="n">float64</span><span class="p">)</span>
<span class="gr">Out[30]: </span>
<span class="go">0 1.0</span>
<span class="go">1 0.0</span>
<span class="go">2 0.0</span>
<span class="go">3 0.0</span>
<span class="go">4 0.0</span>
<span class="go">dtype: float64</span>
<span class="go">BlockIndex</span>
<span class="go">Block locations: array([0], dtype=int32)</span>
<span class="go">Block lengths: array([1], dtype=int32)</span>
</pre></div>
</div>
<p><span class="yiyi-st" id="yiyi-91">如果任何值不能强制到指定的dtype，它会引发。</span></p>
<div class="highlight-ipython"><div class="highlight"><pre><span></span><span class="gp">In [1]: </span><span class="n">ss</span> <span class="o">=</span> <span class="n">pd</span><span class="o">.</span><span class="n">Series</span><span class="p">([</span><span class="mi">1</span><span class="p">,</span> <span class="n">np</span><span class="o">.</span><span class="n">nan</span><span class="p">,</span> <span class="n">np</span><span class="o">.</span><span class="n">nan</span><span class="p">])</span><span class="o">.</span><span class="n">to_sparse</span><span class="p">()</span>
<span class="go">0 1.0</span>
<span class="go">1 NaN</span>
<span class="go">2 NaN</span>
<span class="go">dtype: float64</span>
<span class="go">BlockIndex</span>
<span class="go">Block locations: array([0], dtype=int32)</span>
<span class="go">Block lengths: array([1], dtype=int32)</span>
<span class="gp">In [2]: </span><span class="n">ss</span><span class="o">.</span><span class="n">astype</span><span class="p">(</span><span class="n">np</span><span class="o">.</span><span class="n">int64</span><span class="p">)</span>
<span class="go">ValueError: unable to coerce current fill_value nan to int64 dtype</span>
</pre></div>
</div>
</div>
<div class="section" id="sparse-calculation">
<span id="id1"></span><h2><span class="yiyi-st" id="yiyi-92">Sparse Calculation</span></h2>
<p><span class="yiyi-st" id="yiyi-93">您可以将NumPy <em>ufuncs</em>应用于<code class="docutils literal"><span class="pre">SparseArray</span></code>，并获得<code class="docutils literal"><span class="pre">SparseArray</span></code>作为结果。</span></p>
<div class="highlight-ipython"><div class="highlight"><pre><span></span><span class="gp">In [31]: </span><span class="n">arr</span> <span class="o">=</span> <span class="n">pd</span><span class="o">.</span><span class="n">SparseArray</span><span class="p">([</span><span class="mf">1.</span><span class="p">,</span> <span class="n">np</span><span class="o">.</span><span class="n">nan</span><span class="p">,</span> <span class="n">np</span><span class="o">.</span><span class="n">nan</span><span class="p">,</span> <span class="o">-</span><span class="mf">2.</span><span class="p">,</span> <span class="n">np</span><span class="o">.</span><span class="n">nan</span><span class="p">])</span>
<span class="gp">In [32]: </span><span class="n">np</span><span class="o">.</span><span class="n">abs</span><span class="p">(</span><span class="n">arr</span><span class="p">)</span>
<span class="gr">Out[32]: </span>
<span class="go">[1.0, nan, nan, 2.0, nan]</span>
<span class="go">Fill: nan</span>
<span class="go">IntIndex</span>
<span class="go">Indices: array([0, 3], dtype=int32)</span>
</pre></div>
</div>
<p><span class="yiyi-st" id="yiyi-94"><em>ufunc</em>也适用于<code class="docutils literal"><span class="pre">fill_value</span></code>。</span><span class="yiyi-st" id="yiyi-95">这是需要得到正确的密集结果。</span></p>
<div class="highlight-ipython"><div class="highlight"><pre><span></span><span class="gp">In [33]: </span><span class="n">arr</span> <span class="o">=</span> <span class="n">pd</span><span class="o">.</span><span class="n">SparseArray</span><span class="p">([</span><span class="mf">1.</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="o">-</span><span class="mf">2.</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">],</span> <span class="n">fill_value</span><span class="o">=-</span><span class="mi">1</span><span class="p">)</span>
<span class="gp">In [34]: </span><span class="n">np</span><span class="o">.</span><span class="n">abs</span><span class="p">(</span><span class="n">arr</span><span class="p">)</span>
<span class="gr">Out[34]: </span>
<span class="go">[1.0, 1, 1, 2.0, 1]</span>
<span class="go">Fill: 1</span>
<span class="go">IntIndex</span>
<span class="go">Indices: array([0, 3], dtype=int32)</span>
<span class="gp">In [35]: </span><span class="n">np</span><span class="o">.</span><span class="n">abs</span><span class="p">(</span><span class="n">arr</span><span class="p">)</span><span class="o">.</span><span class="n">to_dense</span><span class="p">()</span>
<span class="gr">Out[35]: </span><span class="n">array</span><span class="p">([</span> <span class="mf">1.</span><span class="p">,</span> <span class="mf">1.</span><span class="p">,</span> <span class="mf">1.</span><span class="p">,</span> <span class="mf">2.</span><span class="p">,</span> <span class="mf">1.</span><span class="p">])</span>
</pre></div>
</div>
</div>
<div class="section" id="interaction-with-scipy-sparse">
<span id="sparse-scipysparse"></span><h2><span class="yiyi-st" id="yiyi-96">Interaction with scipy.sparse</span></h2>
<p><span class="yiyi-st" id="yiyi-97">实验api在稀疏熊猫和scipy.sparse结构之间进行转换。</span></p>
<p><span class="yiyi-st" id="yiyi-98">A <a class="reference internal" href="generated/pandas.SparseSeries.to_coo.html#pandas.SparseSeries.to_coo" title="pandas.SparseSeries.to_coo"><code class="xref py py-meth docutils literal"><span class="pre">SparseSeries.to_coo()</span></code></a> method is implemented for transforming a <code class="docutils literal"><span class="pre">SparseSeries</span></code> indexed by a <code class="docutils literal"><span class="pre">MultiIndex</span></code> to a <code class="docutils literal"><span class="pre">scipy.sparse.coo_matrix</span></code>.</span></p>
<p><span class="yiyi-st" id="yiyi-99">该方法需要具有两个或更多个级别的<code class="docutils literal"><span class="pre">MultiIndex</span></code>。</span></p>
<div class="highlight-ipython"><div class="highlight"><pre><span></span><span class="gp">In [36]: </span><span class="n">s</span> <span class="o">=</span> <span class="n">pd</span><span class="o">.</span><span class="n">Series</span><span class="p">([</span><span class="mf">3.0</span><span class="p">,</span> <span class="n">np</span><span class="o">.</span><span class="n">nan</span><span class="p">,</span> <span class="mf">1.0</span><span class="p">,</span> <span class="mf">3.0</span><span class="p">,</span> <span class="n">np</span><span class="o">.</span><span class="n">nan</span><span class="p">,</span> <span class="n">np</span><span class="o">.</span><span class="n">nan</span><span class="p">])</span>
<span class="gp">In [37]: </span><span class="n">s</span><span class="o">.</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_tuples</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="s1">'a'</span><span class="p">,</span> <span class="mi">0</span><span class="p">),</span>
<span class="gp"> ....:</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="s1">'a'</span><span class="p">,</span> <span class="mi">1</span><span class="p">),</span>
<span class="gp"> ....:</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="s1">'b'</span><span class="p">,</span> <span class="mi">0</span><span class="p">),</span>
<span class="gp"> ....:</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="s1">'b'</span><span class="p">,</span> <span class="mi">1</span><span class="p">),</span>
<span class="gp"> ....:</span> <span class="p">(</span><span class="mi">2</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="s1">'b'</span><span class="p">,</span> <span class="mi">0</span><span class="p">),</span>
<span class="gp"> ....:</span> <span class="p">(</span><span class="mi">2</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="s1">'b'</span><span class="p">,</span> <span class="mi">1</span><span class="p">)],</span>
<span class="gp"> ....:</span> <span class="n">names</span><span class="o">=</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="s1">'C'</span><span class="p">,</span> <span class="s1">'D'</span><span class="p">])</span>
<span class="gp"> ....:</span>
<span class="gp">In [38]: </span><span class="n">s</span>
<span class="gr">Out[38]: </span>
<span class="go">A B C D</span>
<span class="go">1 2 a 0 3.0</span>
<span class="go"> 1 NaN</span>
<span class="go"> 1 b 0 1.0</span>
<span class="go"> 1 3.0</span>
<span class="go">2 1 b 0 NaN</span>
<span class="go"> 1 NaN</span>
<span class="go">dtype: float64</span>
<span class="c"># SparseSeries</span>
<span class="gp">In [39]: </span><span class="n">ss</span> <span class="o">=</span> <span class="n">s</span><span class="o">.</span><span class="n">to_sparse</span><span class="p">()</span>
<span class="gp">In [40]: </span><span class="n">ss</span>
<span class="gr">Out[40]: </span>
<span class="go">A B C D</span>
<span class="go">1 2 a 0 3.0</span>
<span class="go"> 1 NaN</span>
<span class="go"> 1 b 0 1.0</span>
<span class="go"> 1 3.0</span>
<span class="go">2 1 b 0 NaN</span>
<span class="go"> 1 NaN</span>
<span class="go">dtype: float64</span>
<span class="go">BlockIndex</span>
<span class="go">Block locations: array([0, 2], dtype=int32)</span>
<span class="go">Block lengths: array([1, 2], dtype=int32)</span>
</pre></div>
</div>
<p><span class="yiyi-st" id="yiyi-100">在下面的示例中，通过指定第一个和第二个<code class="docutils literal"><span class="pre">MultiIndex</span></code>级别定义行的标签，将<code class="docutils literal"><span class="pre">SparseSeries</span></code>变换为2-d数组的稀疏表示，和第四级定义列的标签。</span><span class="yiyi-st" id="yiyi-101">我们还指定列和行标签应按最终稀疏表示法排序。</span></p>
<div class="highlight-ipython"><div class="highlight"><pre><span></span><span class="gp">In [41]: </span><span class="n">A</span><span class="p">,</span> <span class="n">rows</span><span class="p">,</span> <span class="n">columns</span> <span class="o">=</span> <span class="n">ss</span><span class="o">.</span><span class="n">to_coo</span><span class="p">(</span><span class="n">row_levels</span><span class="o">=</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">column_levels</span><span class="o">=</span><span class="p">[</span><span class="s1">'C'</span><span class="p">,</span> <span class="s1">'D'</span><span class="p">],</span>
<span class="gp"> ....:</span> <span class="n">sort_labels</span><span class="o">=</span><span class="bp">True</span><span class="p">)</span>
<span class="gp"> ....:</span>
<span class="gp">In [42]: </span><span class="n">A</span>
<span class="gr">Out[42]: </span>
<span class="go"><3x4 sparse matrix of type '<type 'numpy.float64'>'</span>
<span class="go"> with 3 stored elements in COOrdinate format></span>
<span class="gp">In [43]: </span><span class="n">A</span><span class="o">.</span><span class="n">todense</span><span class="p">()</span>
<span class="gr">Out[43]: </span>
<span class="go">matrix([[ 0., 0., 1., 3.],</span>
<span class="go"> [ 3., 0., 0., 0.],</span>
<span class="go"> [ 0., 0., 0., 0.]])</span>
<span class="gp">In [44]: </span><span class="n">rows</span>
<span class="gr">Out[44]: </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="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="mi">2</span><span class="p">,</span> <span class="mi">1</span><span class="p">)]</span>
<span class="gp">In [45]: </span><span class="n">columns</span>
<span class="gr">Out[45]: </span><span class="p">[(</span><span class="s1">'a'</span><span class="p">,</span> <span class="mi">0</span><span class="p">),</span> <span class="p">(</span><span class="s1">'a'</span><span class="p">,</span> <span class="mi">1</span><span class="p">),</span> <span class="p">(</span><span class="s1">'b'</span><span class="p">,</span> <span class="mi">0</span><span class="p">),</span> <span class="p">(</span><span class="s1">'b'</span><span class="p">,</span> <span class="mi">1</span><span class="p">)]</span>
</pre></div>
</div>
<p><span class="yiyi-st" id="yiyi-102">指定不同的行和列标签（而不是排序）会产生不同的稀疏矩阵：</span></p>
<div class="highlight-ipython"><div class="highlight"><pre><span></span><span class="gp">In [46]: </span><span class="n">A</span><span class="p">,</span> <span class="n">rows</span><span class="p">,</span> <span class="n">columns</span> <span class="o">=</span> <span class="n">ss</span><span class="o">.</span><span class="n">to_coo</span><span class="p">(</span><span class="n">row_levels</span><span class="o">=</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="s1">'C'</span><span class="p">],</span>
<span class="gp"> ....:</span> <span class="n">column_levels</span><span class="o">=</span><span class="p">[</span><span class="s1">'D'</span><span class="p">],</span>
<span class="gp"> ....:</span> <span class="n">sort_labels</span><span class="o">=</span><span class="bp">False</span><span class="p">)</span>
<span class="gp"> ....:</span>
<span class="gp">In [47]: </span><span class="n">A</span>
<span class="gr">Out[47]: </span>
<span class="go"><3x2 sparse matrix of type '<type 'numpy.float64'>'</span>
<span class="go"> with 3 stored elements in COOrdinate format></span>
<span class="gp">In [48]: </span><span class="n">A</span><span class="o">.</span><span class="n">todense</span><span class="p">()</span>
<span class="gr">Out[48]: </span>
<span class="go">matrix([[ 3., 0.],</span>
<span class="go"> [ 1., 3.],</span>
<span class="go"> [ 0., 0.]])</span>
<span class="gp">In [49]: </span><span class="n">rows</span>
<span class="gr">Out[49]: </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="s1">'a'</span><span class="p">),</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="s1">'b'</span><span class="p">),</span> <span class="p">(</span><span class="mi">2</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="s1">'b'</span><span class="p">)]</span>
<span class="gp">In [50]: </span><span class="n">columns</span>
<span class="gr">Out[50]: </span><span class="p">[</span><span class="mi">0</span><span class="p">,</span> <span class="mi">1</span><span class="p">]</span>
</pre></div>
</div>
<p><span class="yiyi-st" id="yiyi-103">实现方便方法<a class="reference internal" href="generated/pandas.SparseSeries.from_coo.html#pandas.SparseSeries.from_coo" title="pandas.SparseSeries.from_coo"><code class="xref py py-meth docutils literal"><span class="pre">SparseSeries.from_coo()</span></code></a>用于从<code class="docutils literal"><span class="pre">scipy.sparse.coo_matrix</span></code>创建<code class="docutils literal"><span class="pre">SparseSeries</span></code>。</span></p>
<div class="highlight-ipython"><div class="highlight"><pre><span></span><span class="gp">In [51]: </span><span class="kn">from</span> <span class="nn">scipy</span> <span class="kn">import</span> <span class="n">sparse</span>
<span class="gp">In [52]: </span><span class="n">A</span> <span class="o">=</span> <span class="n">sparse</span><span class="o">.</span><span class="n">coo_matrix</span><span class="p">(([</span><span class="mf">3.0</span><span class="p">,</span> <span class="mf">1.0</span><span class="p">,</span> <span class="mf">2.0</span><span class="p">],</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">0</span><span class="p">],</span> <span class="p">[</span><span class="mi">0</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="n">shape</span><span class="o">=</span><span class="p">(</span><span class="mi">3</span><span class="p">,</span> <span class="mi">4</span><span class="p">))</span>
<span class="gp"> ....:</span>
<span class="gp">In [53]: </span><span class="n">A</span>
<span class="gr">Out[53]: </span>
<span class="go"><3x4 sparse matrix of type '<type 'numpy.float64'>'</span>
<span class="go"> with 3 stored elements in COOrdinate format></span>
<span class="gp">In [54]: </span><span class="n">A</span><span class="o">.</span><span class="n">todense</span><span class="p">()</span>
<span class="gr">Out[54]: </span>
<span class="go">matrix([[ 0., 0., 1., 2.],</span>
<span class="go"> [ 3., 0., 0., 0.],</span>
<span class="go"> [ 0., 0., 0., 0.]])</span>
</pre></div>
</div>
<p><span class="yiyi-st" id="yiyi-104">默认行为（<code class="docutils literal"><span class="pre">dense_index=False</span></code>）只返回一个只包含非空条目的<code class="docutils literal"><span class="pre">SparseSeries</span></code>。</span></p>
<div class="highlight-ipython"><div class="highlight"><pre><span></span><span class="gp">In [55]: </span><span class="n">ss</span> <span class="o">=</span> <span class="n">pd</span><span class="o">.</span><span class="n">SparseSeries</span><span class="o">.</span><span class="n">from_coo</span><span class="p">(</span><span class="n">A</span><span class="p">)</span>
<span class="gp">In [56]: </span><span class="n">ss</span>
<span class="gr">Out[56]: </span>
<span class="go">0 2 1.0</span>
<span class="go"> 3 2.0</span>
<span class="go">1 0 3.0</span>
<span class="go">dtype: float64</span>
<span class="go">BlockIndex</span>
<span class="go">Block locations: array([0], dtype=int32)</span>
<span class="go">Block lengths: array([3], dtype=int32)</span>
</pre></div>
</div>
<p><span class="yiyi-st" id="yiyi-105">指定<code class="docutils literal"><span class="pre">dense_index=True</span></code>将产生一个索引，该索引是矩阵的行和列坐标的笛卡尔乘积。</span><span class="yiyi-st" id="yiyi-106">注意，如果稀疏矩阵足够大（和稀疏），这将消耗大量的存储器（相对于<code class="docutils literal"><span class="pre">dense_index=False</span></code>）。</span></p>
<div class="highlight-ipython"><div class="highlight"><pre><span></span><span class="gp">In [57]: </span><span class="n">ss_dense</span> <span class="o">=</span> <span class="n">pd</span><span class="o">.</span><span class="n">SparseSeries</span><span class="o">.</span><span class="n">from_coo</span><span class="p">(</span><span class="n">A</span><span class="p">,</span> <span class="n">dense_index</span><span class="o">=</span><span class="bp">True</span><span class="p">)</span>
<span class="gp">In [58]: </span><span class="n">ss_dense</span>
<span class="gr">Out[58]: </span>
<span class="go">0 0 NaN</span>
<span class="go"> 1 NaN</span>
<span class="go"> 2 1.0</span>
<span class="go"> 3 2.0</span>
<span class="go">1 0 3.0</span>
<span class="go"> 1 NaN</span>
<span class="go"> 2 NaN</span>
<span class="go"> 3 NaN</span>
<span class="go">2 0 NaN</span>
<span class="go"> 1 NaN</span>
<span class="go"> 2 NaN</span>
<span class="go"> 3 NaN</span>
<span class="go">dtype: float64</span>
<span class="go">BlockIndex</span>
<span class="go">Block locations: array([2], dtype=int32)</span>
<span class="go">Block lengths: array([3], dtype=int32)</span>
</pre></div>
</div>
</div>