def na_by_column(df, **kwargs):
# Sorted pandas series with the number of nans/nulls of each column.
na = df.isna().sum().sort_values(ascending=False)
kwargs = utils.default_params(kwargs, title="# NAs by Column")
relative = kwargs.pop("relative", None)
relative = len(df) if relative is not None else relative
return _prepare_and_plot_series(na, relative=relative, **kwargs)
def unique_by_column(df, **kwargs):
s = df.nunique().sort_values(ascending=False)
n_rows = df.shape[0]
title = f"# Unique values by Column (df # rows: {n_rows})"
kwargs = utils.default_params(kwargs, title=title)
relative = kwargs.pop("relative", None)
relative = len(df) if relative is not None else relative
return _prepare_and_plot_series(s, relative=relative, **kwargs)
def value_counts(s,
dropna=False,
top=None,
relative=False,
others="(others)",
nulls="(nulls)",
return_df=False,
**kwargs):
c = s.value_counts(dropna=dropna)
c = c.rename(index={np.nan: nulls})
if top is not None:
c = top_others(c, top, others=others)
if relative:
c /= c.sum()
if return_df:
return c
kwargs = utils.default_params(kwargs, x_cmap="tab10")
p = plot.bar(c, **kwargs)
return p(title="Count by value")
def _prepare_and_plot_series(s,
top=None,
others="(others)",
relative=None,
title="Best plot ever :)",
**kwargs):
if top is not None:
s = top_others(s, top, others=others)
if relative is not None:
relative = "sum" if relative is True else relative
if type(relative) == str:
total = s.agg(relative)
elif type(relative) in [float, int]:
total = relative
else:
raise ValueError(f"Unknown value for relative: f{relative}")
s /= total
kwargs = utils.default_params(kwargs, x_cmap="tab10")
p = plot.bar(s, **kwargs)
return p(title=title)
def memory_by_column(df, deep=True, **kwargs):
# Sorted pandas series with the MB used by each column.
s = (df.memory_usage(deep=deep) / 1024**2).sort_values(ascending=False)
kwargs = utils.default_params(kwargs, title="Memory Usage (MB) by Column")
return _prepare_and_plot_series(s, **kwargs)