<p>为了提高性能,可以使用<a href="http://pandas.pydata.org/pandas-docs/stable/generated/pandas.DataFrame.dot.html" rel="nofollow noreferrer">^{<cd1>}</a>所有不带第一个的列,所有不带最后一个的列,最后一个通过<a href="http://pandas.pydata.org/pandas-docs/stable/generated/pandas.Series.str.rstrip.html" rel="nofollow noreferrer">^{<cd4>}</a>删除最后一个<code>|</code>:</p>
<pre><code>df['new'] = df.iloc[:, 1:].dot(df.columns[1:] + '|').str.rstrip('|')
print (df)
Movie Action Fantasy Vestern new
0 One 1 0 1 Action|Vestern
1 Two 0 0 1 Vestern
2 Three 1 1 0 Action|Fantasy
</code></pre>
<p>或者使用列表理解来连接所有没有空字符串的值:</p>
<pre><code>arr = df.iloc[:, 1:].values * df.columns[1:].values
df['new'] = ['|'.join(y for y in x if y) for x in arr]
print (df)
Movie Action Fantasy Vestern new
0 One 1 0 1 Action|Vestern
1 Two 0 0 1 Vestern
2 Three 1 1 0 Action|Fantasy
</code></pre>
<p><strong>性能</strong>:</p>
<pre><code>In [54]: %timeit (jez1(df.copy()))
25.2 ms ± 2.31 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)
In [55]: %timeit (jez2(df.copy()))
61.4 ms ± 769 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)
In [56]: %timeit (csm(df.copy()))
1.46 s ± 35.7 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)
df = pd.DataFrame({"Movie":['One','Two','Three'],
"Action":[1,0,1],
"Fantasy":[0,0,1],
"Vestern":[1,1,0]})
#print(df)
#30k rows
df = pd.concat([df] * 10000, ignore_index=True)
def csm(df):
cols = df.columns.tolist()[1:]
df['genres'] = df.apply(lambda x: "|".join(str(z) for z in [i for i in cols if x[i] !=0]) ,axis=1)
return df
def jez1(df):
df['new'] = df.iloc[:, 1:].dot(df.columns[1:] + '|').str.rstrip('|')
return df
def jez2(df):
arr = df.iloc[:, 1:].values * df.columns[1:].values
df['new'] = ['|'.join(y for y in x if y) for x in arr]
return df
</code></pre>