在vim中定义颜色变量(Defining a color variable in vim)

 使用pandas，避免将NaN与整数相结合 
 
 除非你有充分的理由，否则请避免这种做法。 原因是pandas只允许在连续内存块中使用向量化计算。 这仅适用于相同类型的数据，例如一系列类型为int ， float ， datetime ，但不是 object 。 
 
 NaN被认为是float 。 因此，将整数与NaN组合会强制将pandas强制转换为float整个系列。 这会增加内存使用量，但对于大多数使用情况不是问题。 
 
 如果你NaN和整数结合起来，你需要用dtype=object创建一个系列，并且让pandas使用一系列指针。 这是昂贵的计算和内存密集型。 除非你绝对必须，否则不要这样做。 
 
 但如果你只是...... 
 
 在将非NaN元素转换为整数之前，您可以将一个序列转换为object ： 
 
df['B'] = df['B'].astype(object)
 
 如上所述，您要求pandas / numpy使用系列中每个项目的指针。 您也可以开始使用列表。 

With pandas, avoid combining NaN with integers
 
Unless you have an extremely good reason, avoid this practice. The reason is pandas only allows vectorised computations with arrays in contiguous memory blocks. This is only possible with data of the same type, e.g. a series of type int, float, datetime, but not object.
 
NaN is considered float. Therefore, combining integers with NaN forces pandas to, by default, upcast the entire series to float. This increases memory usage, but for most use cases is not an issue.
 
If you wish to combine NaN with integers, you need to create a series with dtype=object, and have pandas work with a series of pointers. This is expensive computationally and memory-intensive. Do not do it unless you absolutely must.
 
But if you simply must...
 
You can convert a series to object before converting non-NaN elements to integers:
 
df['B'] = df['B'].astype(object)
 
As explained above, you are asking pandas / numpy to work with a pointer for each item in your series. You might as well start working with lists instead.