Supported pandas API¶

The following table shows the pandas APIs that implemented or non-implemented from pandas API on Spark. Some pandas API do not implement full parameters, so the third column shows missing parameters for each API.

‘Y’ in the second column means it’s implemented including its whole parameter.
‘N’ means it’s not implemented yet.
‘P’ means it’s partially implemented with the missing of some parameters.

All API in the list below computes the data with distributed execution except the ones that require the local execution by design. For example, DataFrame.to_numpy() requires to collect the data to the driver side.

If there is non-implemented pandas API or parameter you want, you can create an Apache Spark JIRA to request or to contribute by your own.

The API list is updated based on the pandas 2.0.0 pre-release.

CategoricalIndex API¶

API	Implemented	Missing parameters
`add_categories()`	Y
argsort	N
`as_ordered()`	Y
`as_unordered()`	Y
astype	N
equals	N
is_dtype_equal	N
`map()`	Y
max	N
min	N
reindex	N
`remove_categories()`	Y
`remove_unused_categories()`	Y
`rename_categories()`	Y
`reorder_categories()`	Y
searchsorted	N
`set_categories()`	Y
take_nd	N
tolist	N

DataFrame API¶

API	Implemented	Missing parameters
`add()`	P	`axis` , `fill_value` , `level`
`agg()`	P	`axis`
`aggregate()`	P	`axis`
`align()`	P	`broadcast_axis` , `fill_axis` , `fill_value` , `level` , `limit` and more. See the pandas.DataFrame.align and pyspark.pandas.DataFrame.align for detail.
`all()`	P	`level`
`any()`	P	`level` , `skipna`
`append()`	Y
`apply()`	P	`raw` , `result_type`
`applymap()`	P	`na_action`
asfreq	N
`assign()`	Y
bfill	N
`boxplot()`	P	`ax` , `backend` , `by` , `column` , `figsize` and more. See the pandas.DataFrame.boxplot and pyspark.pandas.DataFrame.boxplot for detail.
`clip()`	P	`axis` , `inplace`
combine	N
`combine_first()`	Y
compare	N
`corr()`	P	`numeric_only`
`corrwith()`	P	`numeric_only`
count	N
`cov()`	P	`numeric_only`
cummax	N
cummin	N
cumprod	N
cumsum	N
`diff()`	Y
`div()`	P	`axis` , `fill_value` , `level`
`divide()`	P	`axis` , `fill_value` , `level`
`dot()`	Y
`drop()`	P	`errors` , `inplace` , `level`
`drop_duplicates()`	Y
`dropna()`	Y
`duplicated()`	Y
`eq()`	P	`axis` , `level`
`eval()`	Y
`explode()`	Y
ffill	N
`fillna()`	P	`downcast`
`floordiv()`	P	`axis` , `fill_value` , `level`
`ge()`	P	`axis` , `level`
`groupby()`	P	`group_keys` , `level` , `observed` , `sort` , `squeeze`
`gt()`	P	`axis` , `level`
`hist()`	P	`ax` , `backend` , `by` , `column` , `data` and more. See the pandas.DataFrame.hist and pyspark.pandas.DataFrame.hist for detail.
`idxmax()`	P	`numeric_only` , `skipna`
`idxmin()`	P	`numeric_only` , `skipna`
`info()`	P	`memory_usage` , `show_counts`
`insert()`	Y
`interpolate()`	P	`axis` , `downcast` , `inplace`
isetitem	N
`isin()`	Y
`isna()`	Y
`isnull()`	Y
`items()`	Y
`iteritems()`	Y
`iterrows()`	Y
`itertuples()`	Y
`join()`	P	`other` , `sort` , `validate`
kurt	N
kurtosis	N
`le()`	P	`axis` , `level`
lookup	N
`lt()`	P	`axis` , `level`
`mad()`	P	`level` , `skipna`
`mask()`	P	`axis` , `errors` , `inplace` , `level` , `try_cast`
max	N
mean	N
median	N
`melt()`	P	`col_level` , `ignore_index`
memory_usage	N
`merge()`	P	`copy` , `indicator` , `sort` , `validate`
min	N
`mod()`	P	`axis` , `fill_value` , `level`
`mode()`	Y
`mul()`	P	`axis` , `fill_value` , `level`
`multiply()`	P	`axis` , `fill_value` , `level`
`ne()`	P	`axis` , `level`
`nlargest()`	Y
`notna()`	Y
`notnull()`	Y
`nsmallest()`	Y
`nunique()`	Y
`pivot()`	Y
`pivot_table()`	P	`dropna` , `margins` , `margins_name` , `observed` , `sort`
`pop()`	Y
`pow()`	P	`axis` , `fill_value` , `level`
prod	N
product	N
`quantile()`	P	`interpolation` , `method`
`query()`	Y
`radd()`	P	`axis` , `fill_value` , `level`
`rdiv()`	P	`axis` , `fill_value` , `level`
`reindex()`	P	`level` , `limit` , `method` , `tolerance`
`rename()`	P	`copy`
reorder_levels	N
`replace()`	Y
`resample()`	P	`axis` , `base` , `convention` , `group_keys` , `kind` and more. See the pandas.DataFrame.resample and pyspark.pandas.DataFrame.resample for detail.
`reset_index()`	P	`allow_duplicates` , `names`
`rfloordiv()`	P	`axis` , `fill_value` , `level`
`rmod()`	P	`axis` , `fill_value` , `level`
`rmul()`	P	`axis` , `fill_value` , `level`
`round()`	Y
`rpow()`	P	`axis` , `fill_value` , `level`
`rsub()`	P	`axis` , `fill_value` , `level`
`rtruediv()`	P	`axis` , `fill_value` , `level`
`select_dtypes()`	Y
sem	N
set_axis	N
`set_index()`	P	`verify_integrity`
`shift()`	P	`axis` , `freq`
skew	N
`sort_index()`	P	`key` , `sort_remaining`
`sort_values()`	P	`axis` , `key` , `kind`
`stack()`	P	`dropna` , `level`
std	N
`sub()`	P	`axis` , `fill_value` , `level`
`subtract()`	P	`axis` , `fill_value` , `level`
sum	N
`swaplevel()`	Y
`to_dict()`	Y
to_feather	N
to_gbq	N
`to_html()`	P	`encoding`
to_markdown	N
to_numpy	N
`to_orc()`	P	`engine` , `engine_kwargs` , `index`
`to_parquet()`	P	`engine` , `index` , `storage_options`
to_period	N
`to_records()`	Y
to_stata	N
`to_string()`	P	`encoding` , `max_colwidth` , `min_rows`
to_timestamp	N
to_xml	N
`transform()`	Y
`transpose()`	P	`copy`
`truediv()`	P	`axis` , `fill_value` , `level`
`unstack()`	P	`fill_value` , `level`
`update()`	P	`errors` , `filter_func`
value_counts	N
var	N
`where()`	P	`errors` , `inplace` , `level` , `try_cast`

DatetimeIndex API¶

API	Implemented	Missing parameters
`ceil()`	Y
`day_name()`	Y
`floor()`	Y
get_loc	N
`indexer_at_time()`	Y
`indexer_between_time()`	Y
isocalendar	N
`month_name()`	Y
`normalize()`	Y
`round()`	Y
slice_indexer	N
snap	N
std	N
`strftime()`	Y
to_julian_date	N
to_period	N
to_perioddelta	N
to_pydatetime	N
to_series	N
tz_convert	N
tz_localize	N
union_many	N

Index API¶

API	Implemented	Missing parameters
all	N
any	N
`append()`	Y
`argmax()`	P	`axis` , `skipna`
`argmin()`	P	`axis` , `skipna`
argsort	N
`asof()`	Y
asof_locs	N
astype	N
`copy()`	P	`dtype` , `names`
`delete()`	Y
`difference()`	Y
`drop()`	P	`errors`
`drop_duplicates()`	Y
`droplevel()`	Y
`dropna()`	Y
duplicated	N
`equals()`	Y
`fillna()`	P	`downcast`
format	N
get_indexer	N
get_indexer_for	N
get_indexer_non_unique	N
`get_level_values()`	Y
get_loc	N
get_slice_bound	N
get_value	N
groupby	N
`holds_integer()`	Y
`identical()`	Y
`insert()`	Y
`intersection()`	P	`sort`
is_	N
`is_boolean()`	Y
`is_categorical()`	Y
`is_floating()`	Y
`is_integer()`	Y
`is_interval()`	Y
is_mixed	N
`is_numeric()`	Y
`is_object()`	Y
`is_type_compatible()`	Y
isin	N
isna	N
isnull	N
join	N
`map()`	Y
`max()`	P	`axis` , `skipna`
memory_usage	N
`min()`	P	`axis` , `skipna`
notna	N
notnull	N
putmask	N
ravel	N
reindex	N
`rename()`	Y
`repeat()`	P	`axis`
`set_names()`	Y
set_value	N
shift	N
slice_indexer	N
slice_locs	N
`sort()`	Y
`sort_values()`	P	`key` , `na_position`
sortlevel	N
`symmetric_difference()`	Y
take	N
to_flat_index	N
`to_frame()`	Y
to_native_types	N
`to_series()`	P	`index`
`union()`	Y
`unique()`	Y
`view()`	Y
where	N

MultiIndex API¶

API	Implemented	Missing parameters
append	N
argsort	N
astype	N
`copy()`	P	`codes` , `dtype` , `levels` , `name` , `names`
delete	N
`drop()`	P	`errors`
`drop_duplicates()`	Y
dropna	N
duplicated	N
`equal_levels()`	Y
equals	N
fillna	N
format	N
`get_level_values()`	Y
get_loc	N
get_loc_level	N
get_locs	N
get_slice_bound	N
`insert()`	Y
is_lexsorted	N
isin	N
memory_usage	N
remove_unused_levels	N
rename	N
reorder_levels	N
repeat	N
set_codes	N
set_levels	N
set_names	N
slice_locs	N
sortlevel	N
`swaplevel()`	Y
take	N
to_flat_index	N
`to_frame()`	P	`allow_duplicates`
truncate	N
unique	N
view	N

Series API¶

API	Implemented	Missing parameters
`add()`	P	`axis` , `fill_value` , `level`
`agg()`	P	`axis`
`aggregate()`	P	`axis`
`align()`	P	`broadcast_axis` , `fill_axis` , `fill_value` , `level` , `limit` and more. See the pandas.Series.align and pyspark.pandas.Series.align for detail.
all	N
any	N
`append()`	Y
`apply()`	P	`convert_dtype`
`argsort()`	P	`axis` , `kind` , `order`
asfreq	N
`autocorr()`	Y
`between()`	Y
bfill	N
`clip()`	P	`axis`
combine	N
`combine_first()`	Y
`compare()`	P	`align_axis` , `result_names`
`corr()`	Y
count	N
`cov()`	Y
cummax	N
cummin	N
cumprod	N
cumsum	N
`diff()`	Y
`div()`	P	`axis` , `fill_value` , `level`
`divide()`	P	`axis` , `fill_value` , `level`
`divmod()`	P	`axis` , `fill_value` , `level`
`dot()`	Y
`drop()`	P	`axis` , `errors`
`drop_duplicates()`	Y
`dropna()`	P	`how`
`duplicated()`	Y
`eq()`	P	`axis` , `fill_value` , `level`
`explode()`	P	`ignore_index`
ffill	N
`fillna()`	P	`downcast`
`floordiv()`	P	`axis` , `fill_value` , `level`
`ge()`	P	`axis` , `fill_value` , `level`
`groupby()`	P	`group_keys` , `level` , `observed` , `sort` , `squeeze`
`gt()`	P	`axis` , `fill_value` , `level`
`hist()`	P	`ax` , `backend` , `by` , `figsize` , `grid` and more. See the pandas.Series.hist and pyspark.pandas.Series.hist for detail.
`idxmax()`	P	`axis`
`idxmin()`	P	`axis`
info	N
`interpolate()`	P	`axis` , `downcast` , `inplace`
isin	N
isna	N
isnull	N
`items()`	Y
`iteritems()`	Y
`keys()`	Y
kurt	N
kurtosis	N
`le()`	P	`axis` , `fill_value` , `level`
`lt()`	P	`axis` , `fill_value` , `level`
`mad()`	P	`axis` , `level` , `skipna`
`map()`	Y
`mask()`	P	`axis` , `errors` , `inplace` , `level` , `try_cast`
max	N
mean	N
median	N
memory_usage	N
min	N
`mod()`	P	`axis` , `fill_value` , `level`
`mode()`	Y
`mul()`	P	`axis` , `fill_value` , `level`
`multiply()`	P	`axis` , `fill_value` , `level`
`ne()`	P	`axis` , `fill_value` , `level`
`nlargest()`	P	`keep`
notna	N
notnull	N
`nsmallest()`	P	`keep`
`pop()`	Y
`pow()`	P	`axis` , `fill_value` , `level`
prod	N
product	N
`quantile()`	P	`interpolation`
`radd()`	P	`axis` , `fill_value` , `level`
ravel	N
`rdiv()`	P	`axis` , `fill_value` , `level`
`rdivmod()`	P	`axis` , `fill_value` , `level`
`reindex()`	Y
`rename()`	P	`axis` , `copy` , `errors` , `inplace` , `level`
reorder_levels	N
`repeat()`	P	`axis`
`replace()`	P	`inplace` , `limit` , `method`
`resample()`	P	`axis` , `base` , `convention` , `group_keys` , `kind` and more. See the pandas.Series.resample and pyspark.pandas.Series.resample for detail.
`reset_index()`	P	`allow_duplicates`
`rfloordiv()`	P	`axis` , `fill_value` , `level`
`rmod()`	P	`axis` , `fill_value` , `level`
`rmul()`	P	`axis` , `fill_value` , `level`
`round()`	Y
`rpow()`	P	`axis` , `fill_value` , `level`
`rsub()`	P	`axis` , `fill_value` , `level`
`rtruediv()`	P	`axis` , `fill_value` , `level`
`searchsorted()`	P	`sorter`
sem	N
set_axis	N
shift	N
skew	N
`sort_index()`	P	`key` , `sort_remaining`
`sort_values()`	P	`axis` , `key` , `kind`
std	N
`sub()`	P	`axis` , `fill_value` , `level`
`subtract()`	P	`axis` , `fill_value` , `level`
sum	N
`swaplevel()`	Y
take	N
`to_dict()`	Y
`to_frame()`	Y
to_markdown	N
to_period	N
`to_string()`	P	`min_rows`
to_timestamp	N
`transform()`	Y
`truediv()`	P	`axis` , `fill_value` , `level`
`unique()`	Y
`unstack()`	P	`fill_value`
`update()`	Y
var	N
view	N
`where()`	P	`axis` , `errors` , `inplace` , `level` , `try_cast`

TimedeltaIndex API¶

API	Implemented	Missing parameters
ceil	N
floor	N
get_loc	N
median	N
round	N
std	N
sum	N
to_pytimedelta	N
total_seconds	N

General Function API¶

API	Implemented	Missing parameters
array	N
bdate_range	N
`concat()`	P	`copy` , `keys` , `levels` , `names` , `verify_integrity`
crosstab	N
cut	N
`date_range()`	P	`inclusive`
eval	N
factorize	N
from_dummies	N
`get_dummies()`	Y
infer_freq	N
interval_range	N
`isna()`	Y
`isnull()`	Y
json_normalize	N
lreshape	N
`melt()`	P	`col_level` , `ignore_index`
`merge()`	P	`copy` , `indicator` , `left` , `sort` , `validate`
`merge_asof()`	Y
merge_ordered	N
`notna()`	Y
`notnull()`	Y
period_range	N
pivot	N
pivot_table	N
qcut	N
`read_clipboard()`	Y
`read_csv()`	P	`cache_dates` , `chunksize` , `compression` , `converters` , `date_parser` and more. See the pandas.read_csv and pyspark.pandas.read_csv for detail.
`read_excel()`	P	`decimal` , `na_filter` , `storage_options`
read_feather	N
read_fwf	N
read_gbq	N
read_hdf	N
`read_html()`	P	`extract_links`
`read_json()`	P	`chunksize` , `compression` , `convert_axes` , `convert_dates` , `date_unit` and more. See the pandas.read_json and pyspark.pandas.read_json for detail.
`read_orc()`	Y
`read_parquet()`	P	`engine` , `storage_options` , `use_nullable_dtypes`
read_pickle	N
read_sas	N
read_spss	N
`read_sql()`	P	`chunksize` , `coerce_float` , `params` , `parse_dates`
`read_sql_query()`	P	`chunksize` , `coerce_float` , `dtype` , `params` , `parse_dates`
`read_sql_table()`	P	`chunksize` , `coerce_float` , `parse_dates`
read_stata	N
`read_table()`	P	`cache_dates` , `chunksize` , `comment` , `compression` , `converters` and more. See the pandas.read_table and pyspark.pandas.read_table for detail.
read_xml	N
set_eng_float_format	N
show_versions	N
test	N
`timedelta_range()`	Y
`to_datetime()`	P	`cache` , `dayfirst` , `exact` , `utc` , `yearfirst`
`to_numeric()`	P	`downcast`
to_pickle	N
`to_timedelta()`	Y
unique	N
value_counts	N
wide_to_long	N

Expanding API¶

API	Implemented	Missing parameters
agg	N
aggregate	N
apply	N
corr	N
`count()`	P	`numeric_only`
cov	N
`kurt()`	P	`numeric_only`
`max()`	P	`engine` , `engine_kwargs` , `numeric_only`
`mean()`	P	`engine` , `engine_kwargs` , `numeric_only`
median	N
`min()`	P	`engine` , `engine_kwargs` , `numeric_only`
`quantile()`	P	`interpolation` , `numeric_only`
rank	N
sem	N
`skew()`	P	`numeric_only`
`std()`	P	`ddof` , `engine` , `engine_kwargs` , `numeric_only`
`sum()`	P	`engine` , `engine_kwargs` , `numeric_only`
`var()`	P	`ddof` , `engine` , `engine_kwargs` , `numeric_only`

Rolling API¶

API	Implemented	Missing parameters
agg	N
aggregate	N
apply	N
corr	N
`count()`	P	`numeric_only`
cov	N
`kurt()`	P	`numeric_only`
`max()`	P	`engine` , `engine_kwargs` , `numeric_only`
`mean()`	P	`engine` , `engine_kwargs` , `numeric_only`
median	N
`min()`	P	`engine` , `engine_kwargs` , `numeric_only`
`quantile()`	P	`interpolation` , `numeric_only`
rank	N
sem	N
`skew()`	P	`numeric_only`
`std()`	P	`ddof` , `engine` , `engine_kwargs` , `numeric_only`
`sum()`	P	`engine` , `engine_kwargs` , `numeric_only`
`var()`	P	`ddof` , `engine` , `engine_kwargs` , `numeric_only`

Window API¶

API	Implemented	Missing parameters
agg	N
aggregate	N
mean	N
std	N
sum	N
var	N

DataFrameGroupBy API¶

API	Implemented	Missing parameters
agg	N
aggregate	N
boxplot	N
filter	N
idxmax	N
idxmin	N
nunique	N
transform	N
value_counts	N

GroupBy API¶

API	Implemented	Missing parameters
`all()`	Y
`any()`	P	`skipna`
`apply()`	Y
`backfill()`	Y
`bfill()`	Y
`count()`	Y
`cumcount()`	Y
`cummax()`	P	`axis` , `numeric_only`
`cummin()`	P	`axis` , `numeric_only`
`cumprod()`	P	`axis`
`cumsum()`	P	`axis`
describe	N
`diff()`	P	`axis`
`ewm()`	Y
`expanding()`	Y
`ffill()`	Y
`first()`	Y
`head()`	Y
`last()`	Y
`max()`	P	`engine` , `engine_kwargs`
`mean()`	P	`engine` , `engine_kwargs`
`median()`	Y
`min()`	P	`engine` , `engine_kwargs`
ngroup	N
ohlc	N
`pad()`	Y
pct_change	N
`prod()`	Y
`quantile()`	P	`interpolation` , `numeric_only`
`rank()`	P	`axis` , `na_option` , `pct`
resample	N
`rolling()`	Y
sample	N
`sem()`	P	`numeric_only`
`shift()`	P	`axis` , `freq`
`size()`	Y
`std()`	P	`engine` , `engine_kwargs` , `numeric_only`
`sum()`	P	`engine` , `engine_kwargs`
`tail()`	Y
`var()`	P	`engine` , `engine_kwargs` , `numeric_only`

SeriesGroupBy API¶

API	Implemented	Missing parameters
`agg()`	P	`engine` , `engine_kwargs` , `func`
`aggregate()`	P	`engine` , `engine_kwargs` , `func`
apply	N
describe	N
filter	N
`nlargest()`	P	`keep`
`nsmallest()`	P	`keep`
nunique	N
transform	N
`value_counts()`	P	`bins` , `normalize`

Best Practices FAQ