Sample Code/Day_08_SampleCode.ipynb

{"nbformat":4,"nbformat_minor":0,"metadata":{"@webio":{"lastCommId":null,"lastKernelId":null},"kernelspec":{"display_name":"Python 3","language":"python","name":"python3"},"language_info":{"codemirror_mode":{"name":"ipython","version":3},"file_extension":".py","mimetype":"text/x-python","name":"python","nbconvert_exporter":"python","pygments_lexer":"ipython3","version":"3.7.4"},"colab":{"name":"範例程式.ipynb","provenance":[],"collapsed_sections":[]}},"cells":[{"cell_type":"markdown","metadata":{"id":"soB11lZtKO9X"},"source":["<img width=150 src=\"https://upload.wikimedia.org/wikipedia/commons/thumb/1/1a/NumPy_logo.svg/200px-NumPy_logo.svg.png\"></img>\n","\n","# Part.2-1-08 NumPy 結構化陣列 (Structured Arrays)\n","\n","除了數值資料之外，NumPy陣列也可以儲存複合式資料，也就是包含不同資料型別的元素。這就是結構化陣列 (Structured Arrays) 的功能，可以在陣列資料中指定名稱、型別，以進行後續的資料存取及處理。"]},{"cell_type":"markdown","metadata":{"id":"Y4QZHZ4vKSSM"},"source":["範例目標:<br>\r\n","1. 分辨不同資料型別 dtype，並注意與 python 資料行別的對應\r\n","2. 實做結構化陣列 (Structured Arrays)\r\n"]},{"cell_type":"markdown","metadata":{"id":"431P7lrwK9gh"},"source":["範例重點:<br>\r\n","1. 注意 Numpy 與 python 資料行別的對應\r\n","2. 結構化陣列可透過dictionary型別的資料建立 np.dtype 物件，並指定 dtype 給陣列\r\n","3. RecordArray 提供更多的屬性可以用來存取結構化陣列，但是相對的效能上較差於Structured Array"]},{"cell_type":"code","metadata":{"id":"Bv3Jhb-YKO9j"},"source":["import numpy as np"],"execution_count":null,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"SkC_SAPMKO9k"},"source":["## 1. 資料型別 (dtype)\n","\n","|資料型別|字母|Python資料型別|NumPy通用資料型別|\n","|---|---|---|---|\n","|boolean | '?'|bool|np.bool_|\n","|signed byte | 'b'|bytes|np.bytes_|\n","|unsigned byte | 'B'|bytes|np.bytes_|\n","|signed integer | 'i'|int|np.int_|\n","|unsigned integer | 'u'||np.uint|\n","|floating-point | 'f'|float|np.float_|\n","|complex-floating point | 'c'|complex|np.cfloat|\n","|timedelta | 'm'|datetime.timedelta|np.timedelta64|\n","|datetime | 'M'|datetime.datetime|np.datetime64|\n","|string|'S', 'a'|str|np.str_|\n","|Unicode string | 'U'|str|np.str_|"]},{"cell_type":"markdown","metadata":{"id":"QP-zZMMRKO9l"},"source":["例如在CSV檔中有不同型別的資料要讀取。我們可以指定各個Column的資料型別。"]},{"cell_type":"code","metadata":{"id":"SRTQUaGaKO9m"},"source":["%load structured.txt"],"execution_count":null,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"aiqz2De1KO9m"},"source":["透過 `numpy.dtype` 物件，指定要讀入各Column的資料型別，下面的例子分別示範Python的資料型別及字母代表的型別，其中字母後的數字代表其長度，例如 f8 代表 float64 (8 bytes)，U5代表長度5以下的unicode字串。\n","\n","資料型別也可以使用NumPy的資料型別，例如 `np.int32`。"]},{"cell_type":"code","metadata":{"id":"G44GVCAPKO9p"},"source":["# 使用字母代表的資料型別\n","dt = np.dtype('U5, i8, i8, U3')"],"execution_count":null,"outputs":[]},{"cell_type":"code","metadata":{"id":"j5NlxGkEKO9q","outputId":"72c00d3c-5bb3-433a-beb5-29f95d724c0e"},"source":["a = np.genfromtxt(\"structured.txt\", delimiter=',', dtype=dt)\n","a"],"execution_count":null,"outputs":[{"output_type":"execute_result","data":{"text/plain":["array([('Jay', 1, 2, 'Yes'), ('James', 3, 4, 'No'), ('Joe', 5, 6, 'Yes')],\n","      dtype=[('f0', '<U5'), ('f1', '<i8'), ('f2', '<i8'), ('f3', '<U3')])"]},"metadata":{"tags":[]},"execution_count":4}]},{"cell_type":"markdown","metadata":{"id":"2xHqsdKHKO9v"},"source":["## 2. 結構化陣列 (Structured Arrays)\n","\n","建立結構化陣列可透過dictionary型別的資料建立 `np.dtype` 物件，並指定 `dtype` 給陣列。\n","\n","這邊的資料型別可以使用Python的資料型別、NumPy的資料型別、或是字母代表的型別皆可。在範例中我們混用了3種型別的表示方式。"]},{"cell_type":"code","metadata":{"id":"9eOAyCIJKO9v"},"source":["dt = np.dtype({'names':('Name', 'num1', 'num2', 'True'), 'formats':((np.str_, 5), np.int32, int, 'U3')})"],"execution_count":null,"outputs":[]},{"cell_type":"code","metadata":{"id":"7G6cql4UKO9z","outputId":"d8b9c5d8-3377-4611-e2b0-dc4fcf7ec3d2"},"source":["b = np.genfromtxt(\"structured.txt\", delimiter=',', dtype=dt)\n","b"],"execution_count":null,"outputs":[{"output_type":"execute_result","data":{"text/plain":["array([('Jay', 1, 2, 'Yes'), ('James', 3, 4, 'No'), ('Joe', 5, 6, 'Yes')],\n","      dtype=[('Name', '<U5'), ('num1', '<i4'), ('num2', '<i8'), ('True', '<U3')])"]},"metadata":{"tags":[]},"execution_count":6}]},{"cell_type":"markdown","metadata":{"id":"a1r2ZxvOKO90"},"source":["建立陣列後，可以用索引的方式存取元素資料。"]},{"cell_type":"code","metadata":{"id":"DJ_EVDbvKO90","outputId":"2161a020-b11a-4001-e7f6-0c23e271ba41"},"source":["b[0]"],"execution_count":null,"outputs":[{"output_type":"execute_result","data":{"text/plain":["('Jay', 1, 2, 'Yes')"]},"metadata":{"tags":[]},"execution_count":7}]},{"cell_type":"markdown","metadata":{"id":"qhuHZK48KO92"},"source":["也可以用Column名稱，取得Column所有元素值。"]},{"cell_type":"code","metadata":{"id":"o75ZCY6TKO93","outputId":"c2fc8b4b-7783-4bb6-ed84-cdf69bc19eba"},"source":["b['Name']"],"execution_count":null,"outputs":[{"output_type":"execute_result","data":{"text/plain":["array(['Jay', 'James', 'Joe'], dtype='<U5')"]},"metadata":{"tags":[]},"execution_count":8}]},{"cell_type":"markdown","metadata":{"id":"26u6X7R7KO93"},"source":["或是單筆資料的欄位值。"]},{"cell_type":"code","metadata":{"id":"VnJMg5XVKO94","outputId":"08f42078-e8b4-4531-8f22-9805e19fa9b4"},"source":["b[1]['True']"],"execution_count":null,"outputs":[{"output_type":"execute_result","data":{"text/plain":["'No'"]},"metadata":{"tags":[]},"execution_count":9}]},{"cell_type":"markdown","metadata":{"id":"Di35hzpEKO96"},"source":["也可以進行邏輯操作，取得對應的結果。"]},{"cell_type":"code","metadata":{"id":"AdDKUtpGKO97","outputId":"812f253e-7d34-4329-8cf7-b6fe710f91b3"},"source":["b[b['num2'] >= 3]['Name']"],"execution_count":null,"outputs":[{"output_type":"execute_result","data":{"text/plain":["array(['James', 'Joe'], dtype='<U5')"]},"metadata":{"tags":[]},"execution_count":10}]},{"cell_type":"markdown","metadata":{"id":"Io_ezTptKO98"},"source":["### 建立結構化陣列\n","\n","上面的範例是從檔案中讀取的資料建立結構化陣列，如果要新建立一個結構化陣列，方式跟建立陣列非常類似。\n","\n","下例使用 `zeros()` 初始化陣列，並指定 `dtype`。"]},{"cell_type":"code","metadata":{"id":"x64qrdrsKO99","outputId":"a99a44f9-f33c-4419-d4ce-f12153ac4bd3"},"source":["c = np.zeros(3, dtype=dt)\n","c"],"execution_count":null,"outputs":[{"output_type":"execute_result","data":{"text/plain":["array([('', 0, 0, ''), ('', 0, 0, ''), ('', 0, 0, '')],\n","      dtype=[('Name', '<U5'), ('num1', '<i4'), ('num2', '<i8'), ('True', '<U3')])"]},"metadata":{"tags":[]},"execution_count":11}]},{"cell_type":"code","metadata":{"id":"5gKM986JKO9_","outputId":"5e886b20-8f82-4879-f05e-1e078ca6f205"},"source":["print(c.dtype)"],"execution_count":null,"outputs":[{"output_type":"stream","text":["[('Name', '<U5'), ('num1', '<i4'), ('num2', '<i8'), ('True', '<U3')]\n"],"name":"stdout"}]},{"cell_type":"code","metadata":{"id":"7Yv25pdKKO-A"},"source":["name = ['Chloe', 'Charlotte', 'Clara']\n","num_1 = [11, 12, 13]\n","num_2 = [14, 15, 16]\n","check = ['Y', 'Y', 'N']"],"execution_count":null,"outputs":[]},{"cell_type":"code","metadata":{"id":"Sy7--6zSKO-A"},"source":["c['Name'] = name\n","c['num1'] = num_1\n","c['num2'] = num_2\n","c['True'] = check"],"execution_count":null,"outputs":[]},{"cell_type":"code","metadata":{"id":"P_NlcUhMKO-A","outputId":"5e97a944-6d21-4e18-8d9e-1a87a45c715a"},"source":["print(c)"],"execution_count":null,"outputs":[{"output_type":"stream","text":["[('Chloe', 11, 14, 'Y') ('Charl', 12, 15, 'Y') ('Clara', 13, 16, 'N')]\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"o7WPsiJ_KO-C"},"source":["## 3. RecordArray：`numpy.recarray()`\n","\n","RecordArray 與 Structured Array 非常類似，但是提供更多的屬性可以用來存取結構化陣列。不過 RecordArray 雖然方便但是在效能上會比原來的陣列差。使用方法如下："]},{"cell_type":"code","metadata":{"id":"vplfF-j7KO-D","outputId":"c79b5e38-a497-4057-fbe2-1513e2481c45"},"source":["c_rec = c.view(np.recarray)\n","c_rec"],"execution_count":null,"outputs":[{"output_type":"execute_result","data":{"text/plain":["rec.array([('Chloe', 11, 14, 'Y'), ('Charl', 12, 15, 'Y'),\n","           ('Clara', 13, 16, 'N')],\n","          dtype=[('Name', '<U5'), ('num1', '<i4'), ('num2', '<i8'), ('True', '<U3')])"]},"metadata":{"tags":[]},"execution_count":16}]},{"cell_type":"markdown","metadata":{"id":"OjRj54k7KO-D"},"source":["原先我們是透過索引或是名稱存取元素值，但是 RecordArray 可以使用屬性的方式來取得。"]},{"cell_type":"code","metadata":{"id":"eNvNQfIaKO-E","outputId":"9668dcfc-0839-4442-d7e3-c4f8b76bb2f5"},"source":["c_rec.Name"],"execution_count":null,"outputs":[{"output_type":"execute_result","data":{"text/plain":["array(['Chloe', 'Charl', 'Clara'], dtype='<U5')"]},"metadata":{"tags":[]},"execution_count":17}]},{"cell_type":"code","metadata":{"id":"ehKiS-4xKO-G"},"source":[""],"execution_count":null,"outputs":[]}]}