import copy
import pandas as pd
from dtale.utils import (ChartBuildingError, classify_type, find_dtype,
find_dtype_formatter, flatten_lists, get_dtypes,
grid_columns, grid_formatter, json_int, make_list,
run_query)
YAXIS_CHARTS = ['line', 'bar', 'scatter']
ZAXIS_CHARTS = ['heatmap', '3d_scatter', 'surface']
MAX_GROUPS = 30
[docs]def valid_chart(chart_type=None, x=None, y=None, z=None, **inputs):
"""
Helper function to determine based on inputs (chart_type, x, y, z...) whether a chart can be build or not.
For example, charts must have an x & y value and for 3-dimensional charts they must have a Z-Axis as well.
:param chart_type: type of chart to build (line, bar, scatter, pie, heatmap, wordcloud, 3dscatter, surface)
:type chart_type: str, optional
:param x: column to use for the X-Axis
:type x: str, optional
:param y: columns to use for the Y-Axes
:type y: list of str, optional
:param z: column to use for the Z-Axis
:type z: str, optional
:param inputs: keyword arguments containing
:return: `True` if executed from test, `False` otherwise
:rtype: bool
"""
if chart_type == 'maps':
map_type = inputs.get('map_type')
if map_type == 'choropleth' and all(inputs.get(p) is not None for p in ['loc_mode', 'loc', 'map_val']):
return True
elif map_type == 'scattergeo' and all(inputs.get(p) is not None for p in ['lat', 'lon']):
return True
return False
if x is None or not len(y or []):
return False
if chart_type in ZAXIS_CHARTS and z is None:
return False
if inputs.get('agg') == 'rolling' and (inputs.get('window') is None or inputs.get('rolling_comp') is None):
return False
return True
[docs]def date_freq_handler(df):
"""
This returns a column definition handler which returns a series based on the specs from the front-end.
Column definitions can be a column name 'Col1' or a column name with a frequency 'Col1|M' for
columns which are of type datetime.
:Example:
Col1 -> returns series for Col1
Col1|M -> returns series for Col1 in monthly format with name 'Col1|M'
:param df: dataframe whose data needs to be checked
:type df: :class:`pandas:pandas.DataFrame`
:return: handler function
:rtype: func
"""
dtypes = get_dtypes(df)
orig_idx = df.index
def _handler(col_def):
col_def_segs = col_def.split('|')
if len(col_def_segs) > 1 and classify_type(dtypes[col_def_segs[0]]) == 'D':
col, freq = col_def_segs
if freq == 'WD':
code = "df.set_index('{col}').index.dayofweek.values"
freq_grp = df.set_index(col).index.dayofweek.values
elif freq == 'H2':
code = "df.set_index('{col}').index.hour.values"
freq_grp = df.set_index(col).index.hour.values
else:
code = "df.set_index('{col}').index.to_period('{freq}').to_timestamp(how='end').values"
freq_grp = df.set_index(col).index.to_period(freq).to_timestamp(how='end').values
code = "\tpd.Series(" + code + ", index=df.index, name='{col_def}'),"
freq_grp = pd.Series(freq_grp, index=orig_idx, name=col_def)
return freq_grp, code.format(col=col, freq=freq, col_def=col_def)
else:
return df[col_def], "\tdf['{col_def}'],".format(col_def=col_def)
return _handler
[docs]def group_filter_handler(col_def, group_val, group_classifier):
col_def_segs = col_def.split('|')
if len(col_def_segs) > 1:
col, freq = col_def_segs
if group_val == 'nan':
return '{col} != {col}'.format(col=col)
if freq == 'WD':
return '{}.dt.dayofweek == {}'.format(col, group_val)
elif freq == 'H2':
return '{}.dt.hour == {}'.format(col, group_val)
elif freq == 'H':
ts_val = pd.Timestamp(group_val)
return "{col}.dt.date == '{day}' and {col}.dt.hour == {hour}".format(
col=col, day=ts_val.strftime('%Y%m%d'), hour=ts_val.hour
)
elif freq == 'D':
ts_val = pd.Timestamp(group_val)
return "{col}.dt.date == '{day}'".format(col=col, day=ts_val.strftime('%Y%m%d'))
elif freq == 'W':
ts_val = pd.Timestamp(group_val)
return "{col}.dt.year == {year} and {col}.dt.week == {week}".format(
col=col, year=ts_val.year, week=ts_val.week
)
elif freq == 'M':
ts_val = pd.Timestamp(group_val)
return "{col}.dt.year == {year} and {col}.dt.month == {month}".format(
col=col, year=ts_val.year, month=ts_val.month
)
elif freq == 'Q':
ts_val = pd.Timestamp(group_val)
return "{col}.dt.year == {year} and {col}.dt.quarter == {quarter}".format(
col=col, year=ts_val.year, quarter=ts_val.quarter
)
elif freq == 'Y':
ts_val = pd.Timestamp(group_val)
return "{col}.dt.year == {year}".format(col=col, year=ts_val.year)
if group_val == 'nan':
return '{col} != {col}'.format(col=col_def)
if group_classifier in ['I', 'F']:
return '{col} == {val}'.format(col=col_def, val=group_val)
return "{col} == '{val}'".format(col=col_def, val=group_val)
[docs]def retrieve_chart_data(df, *args, **kwargs):
"""
Retrieves data from a dataframe for x, y, z & group inputs complete with date frequency
formatting (:meth:`dtale.charts.utils.date_freq_handler`) if specified
:param df: dataframe that contains data for chart
:type df: :class:`pandas:pandas.DataFrame`
:param args: columns to use
:type args: iterable of str
:return: dataframe of data required for chart construction
:rtype: :class:`pandas:pandas.DataFrame`
"""
freq_handler = date_freq_handler(df)
cols = flatten_lists([make_list(a) for a in args])
all_code = []
all_data = []
for col in cols:
if col is not None:
s, code = freq_handler(col)
all_data.append(s)
if code is not None:
all_code.append(code)
all_data = pd.concat(all_data, axis=1)
all_code = ["chart_data = pd.concat(["] + all_code + ["], axis=1)"]
if len(make_list(kwargs.get('group_val'))):
filters = build_group_inputs_filter(all_data, kwargs['group_val'])
all_data = run_query(all_data, filters)
all_code.append('chart_data = chart_data.query({})'.format(filters))
return all_data, all_code
[docs]def check_all_nan(df, cols=None):
"""
Checker function to test whether all data within a column of a dataframe is :attr:`numpy:numpy.nan`
:param df: dataframe whose data needs to be checked
:type df: :class:`pandas:pandas.DataFrame`
:param cols: columns to test
:type cols: list of str
:raises Exception: if all data within a column is :attr:`numpy:numpy.nan`
"""
for col in cols or df.columns:
if df[col].isnull().all():
raise Exception('All data for column "{}" is NaN!'.format(col))
LIMIT_MSG = 'Dataset exceeds {} records, cannot render. Please apply filter...'
[docs]def check_exceptions(df, allow_duplicates, unlimited_data=False, data_limit=15000, limit_msg=LIMIT_MSG):
"""
Checker function to test the output of any chart aggregations to see if it is one of the following:
- too large to be rendered by web client
- contains duplicate data points which can't be rendered (ex: multiple points for a single point on the x-axis
of a bar chart within the same series)
:param df: dataframe whose data needs to be checked
:type df: :class:`pandas:pandas.DataFrame`
:param allow_duplicates: flag to allow duplicates to be ignored (usually for scatter plots)
:type allow_duplicates: bool
:param data_limit: maximum rows allowed for chart rendering (default: 15,000)
:type data_limit: int, optional
:param limit_msg: error message template
:type limit_msg: str, optional
:raises Exception: if any failure condition is met
"""
if not allow_duplicates and any(df.duplicated()):
raise ChartBuildingError((
"{} contains duplicates, please specify group or additional filtering or select 'No Aggregation' from"
' Aggregation drop-down.'
).format(', '.join(df.columns)))
if not unlimited_data and len(df) > data_limit:
raise ChartBuildingError(limit_msg.format(data_limit))
[docs]def build_agg_data(df, x, y, inputs, agg, z=None, group_col=None, animate_by=None):
"""
Builds aggregated data when an aggregation (sum, mean, max, min...) is selected from the front-end.
:param df: dataframe that contains data for chart
:type df: :class:`pandas:pandas.DataFrame`
:param x: column to use for the X-Axis
:type x: str
:param y: columns to use for the Y-Axes
:type y: list of str
:param inputs: additional chart configurations (chart_type, group, rolling_win, rolling_comp...)
:type inputs: dict
:param agg: points to a specific function that can be applied to
:func: pandas.core.groupby.DataFrameGroupBy. Possible values are: count, first, last mean,
median, min, max, std, var, mad, prod, sum
:type agg: str
:param z: column to use for the Z-Axis
:type z: str, optional
:return: dataframe of aggregated data
:rtype: :class:`pandas:pandas.DataFrame`
"""
if agg == 'raw':
return df, []
z_exists = len(make_list(z))
if agg == 'corr':
if not z_exists:
raise NotImplementedError('Correlation aggregation is only available for 3-dimensional charts!')
if agg == 'rolling':
if z_exists:
raise NotImplementedError('Rolling computations have not been implemented for 3-dimensional charts!')
window, comp = map(inputs.get, ['rolling_win', 'rolling_comp'])
agg_df = df.set_index(x).rolling(window=window)
agg_df = pd.DataFrame({c: getattr(agg_df[c], comp)() for c in y})
agg_df = agg_df.reset_index()
code = [
"chart_data = chart_data.set_index('{x}').rolling(window={window})".format(x=x, window=window),
"chart_data = pd.DataFrame({'" + ', '.join(
["'{c}': chart_data['{c}'].{comp}()".format(c=c, comp=comp) for c in y]
) + '})',
"chart_data = chart_data.reset_index()"
]
return agg_df, code
idx_cols = make_list(animate_by) + make_list(group_col) + [x]
agg_cols = y
if z_exists:
idx_cols += make_list(y)
agg_cols = make_list(z)
groups = df.groupby(idx_cols)
if agg in ['pctsum', 'pctct']:
func = 'sum' if agg == 'pctsum' else 'size'
subidx_cols = [c for c in idx_cols if c not in make_list(group_col)]
groups = getattr(groups[agg_cols], func)()
groups = groups / getattr(df.groupby(subidx_cols)[agg_cols], func)() * 100
if len(agg_cols) > 1:
groups.columns = agg_cols
elif len(agg_cols) == 1:
groups.name = agg_cols[0]
code = (
"chart_data = chart_data.groupby(['{cols}'])[['{agg_cols}']].{agg}()\n"
"chart_data = chart_data / chart_data.groupby(['{subidx_cols}']).{agg}()\n"
"chart_data = chart_data.reset_index()"
)
code = code.format(cols="', '".join(idx_cols), subidx_cols="', '".join(subidx_cols),
agg_cols="', '".join(agg_cols), agg=func)
code = [code]
else:
groups = getattr(groups[agg_cols], agg)()
code = [
"chart_data = chart_data.groupby(['{cols}'])[['{agg_cols}']].{agg}().reset_index()".format(
cols="', '".join(idx_cols), agg_cols="', '".join(agg_cols), agg=agg
)
]
if animate_by is not None:
full_idx = pd.MultiIndex.from_product([df[c].unique() for c in idx_cols], names=idx_cols)
groups = groups.reindex(full_idx).fillna(0)
code += [
"idx_cols = ['{cols}']".format(cols="', '".join(idx_cols)),
'full_idx = pd.MultiIndex.from_product([df[c].unique() for c in idx_cols], names=idx_cols)'
'chart_data = chart_data.reindex(full_idx).fillna(0)'
]
return groups.reset_index(), code
[docs]def build_base_chart(raw_data, x, y, group_col=None, group_val=None, agg=None, allow_duplicates=False, return_raw=False,
unlimited_data=False, animate_by=None, **kwargs):
"""
Helper function to return data for 'chart-data' & 'correlations-ts' endpoints. Will return a dictionary of
dictionaries (one for each series) which contain the data for the x & y axes of the chart as well as the minimum &
maximum of all the series for the y-axis. If there is only one series (no group_col specified) the only key in the
dictionary of series data will be 'all' otherwise the keys will be the values of the groups.
:param raw_data: dataframe to be used for chart
:type raw_data: :class:`pandas:pandas.DataFrame`
:param x: column to be used as x-axis of chart
:type x: str
:param y: column to be used as y-axis of chart
:type y: list of strings
:param group: comma-separated string of columns to group chart data by
:type group: str, optional
:param agg: points to a specific function that can be applied to
:func: pandas.core.groupby.DataFrameGroupBy. Possible values are: count, first, last mean,
median, min, max, std, var, mad, prod, sum
:type agg: str, optional
:param allow_duplicates: flag to allow duplicates to be ignored (usually for scatter plots)
:type allow_duplicates: bool, optional
:return: dict
"""
group_fmt_overrides = {'I': lambda v, as_string: json_int(v, as_string=as_string, fmt='{}')}
data, code = retrieve_chart_data(raw_data, x, y, kwargs.get('z'), group_col, animate_by, group_val=group_val)
x_col = str('x')
y_cols = make_list(y)
z_col = kwargs.get('z')
z_cols = make_list(z_col)
sort_cols = (y_cols if len(z_cols) else [])
if group_col is not None and len(group_col):
main_group = group_col
if animate_by is not None:
main_group = [animate_by] + main_group
sort_cols = main_group + [x] + sort_cols
data = data.sort_values(sort_cols)
code.append("chart_data = chart_data.sort_values(['{cols}'])".format(cols="', '".join(sort_cols)))
check_all_nan(data)
data = data.rename(columns={x: x_col})
code.append("chart_data = chart_data.rename(columns={'" + x + "': '" + x_col + "'})")
if agg is not None:
data, agg_code = build_agg_data(data, x_col, y_cols, kwargs, agg, z=z_col, group_col=group_col,
animate_by=animate_by)
code += agg_code
MAX_GROUPS = 30
group_vals = data[group_col].drop_duplicates()
if len(group_vals) > MAX_GROUPS:
dtypes = get_dtypes(group_vals)
group_fmts = {c: find_dtype_formatter(dtypes[c], overrides=group_fmt_overrides) for c in group_col}
group_f, _ = build_formatters(group_vals)
group_vals = group_f.format_lists(group_vals)
group_vals = pd.DataFrame(group_vals, columns=group_col)
msg = (
'Group ({}) contains more than {} unique values, more groups than that will make the chart unreadable. '
'You can choose specific groups to display from then "Group(s)" dropdown above. The available group(s) '
'are listed below:\n\n{}'
).format(', '.join(group_col), MAX_GROUPS, group_vals.to_string(index=False))
raise ChartBuildingError(msg, group_vals.to_string(index=False))
data = data.dropna()
if return_raw:
return data.rename(columns={x_col: x})
code.append("chart_data = chart_data.dropna()")
data_f, range_f = build_formatters(data)
ret_data = dict(
data={},
min={col: fmt(data[col].min(), None) for _, col, fmt in range_f.fmts if col in [x_col] + y_cols + z_cols},
max={col: fmt(data[col].max(), None) for _, col, fmt in range_f.fmts if col in [x_col] + y_cols + z_cols},
)
dtypes = get_dtypes(data)
group_fmts = {c: find_dtype_formatter(dtypes[c], overrides=group_fmt_overrides) for c in group_col}
def _load_groups(df):
for group_val, grp in df.groupby(group_col):
def _group_filter():
for gv, gc in zip(make_list(group_val), group_col):
classifier = classify_type(dtypes[gc])
yield group_filter_handler(gc, group_fmts[gc](gv, as_string=True), classifier)
group_filter = ' and '.join(list(_group_filter()))
yield group_filter, data_f.format_lists(grp)
if animate_by is not None:
frame_fmt = find_dtype_formatter(dtypes[animate_by], overrides=group_fmt_overrides)
ret_data['frames'] = []
for frame_key, frame in data.sort_values(animate_by).groupby(animate_by):
ret_data['frames'].append(
dict(data=dict(_load_groups(frame)), name=frame_fmt(frame_key, as_string=True))
)
ret_data['data'] = copy.deepcopy(ret_data['frames'][-1]['data'])
else:
ret_data['data'] = dict(_load_groups(data))
return ret_data, code
main_group = [x]
if animate_by is not None:
main_group = [animate_by] + main_group
sort_cols = main_group + sort_cols
data = data.sort_values(sort_cols)
code.append("chart_data = chart_data.sort_values(['{cols}'])".format(cols="', '".join(sort_cols)))
check_all_nan(data)
y_cols = [str(y_col) for y_col in y_cols]
data = data[main_group + y_cols + z_cols]
data = data.rename(columns={x: x_col})
main_group = [x_col if c == x else c for c in main_group]
code.append("chart_data = chart_data.rename(columns={'" + x + "': '" + x_col + "'})")
if agg is not None:
data, agg_code = build_agg_data(data, x_col, y_cols, kwargs, agg, z=z_col, animate_by=animate_by)
code += agg_code
data = data.dropna()
if return_raw:
return data.rename(columns={x_col: x})
code.append("chart_data = chart_data.dropna()")
dupe_cols = main_group + (y_cols if len(z_cols) else [])
check_exceptions(
data[dupe_cols].rename(columns={x_col: x}),
allow_duplicates or agg == 'raw',
unlimited_data=unlimited_data,
data_limit=40000 if len(z_cols) or animate_by is not None else 15000
)
data_f, range_f = build_formatters(data)
ret_data = dict(
min={col: fmt(data[col].min(), None) for _, col, fmt in range_f.fmts if col in [x_col] + y_cols + z_cols},
max={col: fmt(data[col].max(), None) for _, col, fmt in range_f.fmts if col in [x_col] + y_cols + z_cols}
)
if animate_by is not None:
frame_fmt = find_dtype_formatter(find_dtype(data[animate_by]), overrides=group_fmt_overrides)
ret_data['frames'] = []
for frame_key, frame in data.sort_values(animate_by).groupby(animate_by):
ret_data['frames'].append(
dict(data={str('all'): data_f.format_lists(frame)}, name=frame_fmt(frame_key, as_string=True))
)
ret_data['data'] = copy.deepcopy(ret_data['frames'][-1]['data'])
else:
ret_data['data'] = {str('all'): data_f.format_lists(data)}
return ret_data, code
WEEKDAY_MAP = {idx: day for idx, day in enumerate(['Mon', 'Tues', 'Wed', 'Thur', 'Fri', 'Sat', 'Sun'])}
[docs]def weekday_tick_handler(col_data, col):
"""
Output handler for datetime data which needs to be returned as weekdays. If the column definition ends with '|WD'
then the integer values in 'data' will be mapped to their standard weekday test (Mon, Tues, Wed, Thur, Fri, Sat,
Sun)
:param col_data: iterable of values within column
:type col_data: list
:param col: column definition
:type col: str
:return: formatted column data
:rtype: list
"""
if col.endswith('|WD'):
return [WEEKDAY_MAP[d] for d in col_data]
return col_data
[docs]def find_group_vals(df, group_cols):
group_vals, _ = retrieve_chart_data(df, group_cols)
group_vals = group_vals.drop_duplicates().sort_values(group_cols)
group_f, _ = build_formatters(group_vals)
return group_f.format_dicts(group_vals.itertuples())