Why does pandas fail to join on two columns of object dtype (one of them is converted from int to object)?

huangapple go评论64阅读模式
英文:

Why does pandas fail to join on two columns of object dtype (one of them is converted from int to object)?

问题

以下是您要翻译的部分:

"为什么 pandas 在这个合并操作中失败?我可以将列转换为字符串 'dtype',然后再转回 'object',就可以成功执行:"

"现在合并是正常的,并找到了匹配项。我应该如何理解这种行为?"

英文:

The following merge strategy fails:

import pandas as pd
data1 = {'c1': ['J', 'A', 'B'],
         'key': [25, 30, 35]}
df1 = pd.DataFrame(data1)

data2 = {'c2': ['A', 'B', 'C'],
         'key': ["25","30","36"]}
df2 = pd.DataFrame(data2, dtype="O")

df1.key = df1.key.astype("O")

print(df1.merge(df2, on = "key"))

output:
Empty DataFrame
Columns: [c1, key, c2]
Index: []

Why is pandas failing in this merge? I can convert the column to string dtype as follows and then back to object and it works:

df1.key = df1.key.astype(str).astype("O")

Now the merge is okay and finds the matches. How should I understand this behavior?

答案1

得分: 2

将Series转换为对象并不会使项目变成字符串,它只是将Series的dtype设置为对象(一个对象Series可以包含任何内容:整数、浮点数、字符串、列表、类等):

df1['key'] = df1['key'].astype('O')

print(df1['key'].tolist())
# [25, 30, 35]

print(type(df1['key'].iloc[0]))
# <class 'int'>

重要的是将项目转换为相同的类型,例如字符串:

df1['key'] = df1['key'].astype(str)

print(df1['key'].tolist())
# ['25', '30', '35']
英文:

Converting the Series to object doesn't make the items strings, it just makes the Series dtype object (An object Series can contain anything: integers, floats, strings, lists, classes…):

df1[&#39;key&#39;] = df1[&#39;key&#39;].astype(&#39;O&#39;)

print(df1[&#39;key&#39;].tolist())
# [25, 30, 35]

print(type(df1[&#39;key&#39;].iloc[0]))
# &lt;class &#39;int&#39;&gt;

What is important is to convert the items to an identical type, for example strings:

df1[&#39;key&#39;] = df1[&#39;key&#39;].astype(str)

print(df1[&#39;key&#39;].tolist())
# [&#39;25&#39;, &#39;30&#39;, &#39;35&#39;]

huangapple
  • 本文由 发表于 2023年7月17日 16:43:30
  • 转载请务必保留本文链接:https://go.coder-hub.com/76702772.html
匿名

发表评论

匿名网友

:?: :razz: :sad: :evil: :!: :smile: :oops: :grin: :eek: :shock: :???: :cool: :lol: :mad: :twisted: :roll: :wink: :idea: :arrow: :neutral: :cry: :mrgreen:

确定