190414-Manipulating DataFrames with pandas-Extracting and transforming data

1.Indexing DataFrame

• Indexing using square brackets

• Using column attribute and row labels

• Using the .loc/.iloc accessor

• selecting only some columns

2. Slicing DataFrame

• 数据中某一列的数据类型是Series, Series 是有index的，介于np arrays和dict之间
• Slicing not using accessor

• Slicing using .loc/.iloc accessor

• using list rather than slices

• series versus 1-column dataframe

• Slicing rows

.loc['b':'a':-1] -1可以表示倒序；另外.loc里，此处列值部分放空也可以

3. Filtering DataFrames

• Filtering with a Boolean Series

• Combining filters

• selecting columns with all nonzeros

• selecting columns with any nonzeros

• Selecting columns with any NaNs

• Selecting columns without NaNs

• Drop rows with any NaNs

如果换成all了name就是当该行全部都是NaN才会被drop

• Filtering a column based on another

• Modifying a column based on another

salt 大于55的，egg值加5

• Filtering using NaNs:

thresh=1000表示less than 1000， axis＝‘columns' 是因为axis默认是0，指的是rows

3. Transforming DataFrames

• DataFrame vectorized methods:

df.floordiv()

• numpy vectorized functions:

np.floor.divide(df,12)

• Plain python function(1)

def

plain python function(2)

• lambda function

• storing transformation

• working with string values

str.upper(): 小写转大写

• index没有apply method，用.map();

• 用两列的运算定义新的列：

• Using apply() to transform a column

用完def用apply，但好像不用apply也能做的。。。

• Using .map() with a dictionary

• Using vectorized functions

注意导入的方式