Pandas 单级多索引与普通索引

Question

One may produce single level MultiIndex in pandas like so:

index = pd.MultiIndex.from_arrays([[1,2,3]])
index
print ('Number of levels = %d' % index.nlevels)

MultiIndex(levels=[[1, 2, 3]],
           codes=[[0, 1, 2]])
Number of levels = 1

And then use it as simple index, for example in order to init DataFrame:

df = pd.DataFrame([0,0,15], index=index, columns=['d'])
df

But is there any 'special' use cases of single level multiindex instead of simple index? I mean the cases where simple index 'less comfortable' or even won't work.

Answer 1

I can only think of the memory usage , if you only have one single level you should using plain index .

df.index.memory_usage()
123
df.index=[1,2,3]
df.index.memory_usage()
24

Answer 2

I think the behavior should be the same, but there appear to be subtle differences.

For example, the single-level MultiIndex causes problems with indexing. See this question: Pandas indexing error with (1) non-unique query, and (2) single-level MultiIndex