index=False by default when exporting dataframes

[JavierLTPromofarma]

Is your feature request related to a problem?

Every single time I have exported a dataframe (usually with .to_csv), I have not needed the index.

Describe the solution you'd like

Change the default from True to False.

API breaking implications

All code that has not explicitely set index=True should make it.

Describe alternatives you've considered

Leave things as they are now.

Not using the index is something usual or am I biased by my own experience?

[TomAugspurger]

We can leave this open for a bit, but I don't think changing this is feasible. In my experience, index=True is the appropriate default, and not using a meaningful index is a bit of an anti-pattern.

This overlaps with an ongoing about an "optional index". I wasn't able to turn up the issue in a quick search.

[jreback]

agree with @TomAugspurger here
even if i agreed (i don’t) this is impossible to change because of compatibility

note this is why we had a Dataframe.from_csv method with different reading defaults, but that was just confusing and has since been removed

[jorisvandenbossche]

I personally agree with @JavierLTPromofarma that the default of index=True is not always ideal, but it's also something that is not really easy / realistic to change.

I think a non-meaningful index is not an anti-pattern. Basically whenever you don't explicitly set one of your columns as the index, you often have a meaningless index. And unless you want to take advantage of easier indexing by setting your column as the index (or from timeseries functionality), setting a column as the index is not always that important.

What we did in GeoPandas' to_file method is change the default to index=None and then "infer" wether to write it: only write the index if it is a non-integer index or if it has a name. I personally think this is a better default (it will of course do a wrong guess in some cases, but then you can explicitly set index to True or False, but I think in a majority of the cases it will be a better default).

Now, the way we can change this is probably rather through the "optional" index idea, as Tom mentioned. We have an old issue about it here: wesm/pandas2#17.

[roschly]

From a UX standpoint the default is less than ideal. The fact that a simple to_csv/read_csv doesn't produce the same dataframe by default is not intuitive.

df = pd.DataFrame(columns=["A", "B"], data=[[1, 2], [3, 4]])
df.to_csv("tmp.csv")
df2 = pd.read_csv("tmp.csv")

df:

	A	B
0	1	2
1	3	4

df2:

	Unnamed: 0	A	B
0	0	1	2
1	1	3	4

This has caused me headaches before when swapping runtime created dataframes with dataframes read from csv. The index=False is an easy fix, but also just as easy to forget.

[mroeschke]

Thanks for the suggestion, but it appears unlikely that index=False would be an easy change for pandas to make due to compatibility. Since having an optional index is covered in another issue, closing.

[DanTaranis]

I have been bothered by this for years - just not enough to do anything about it. but I think the community as a whole would much much much rather have False be the default. Only on rare occasions is the index a meaningful column.

Is there a way to ask the community - like a vote or something?

[pieterseeder]

@DanTaranis for president.