def _function_by_group2(function,
table=None,
model=None,
columns=None,
group_by=None,
**params):
if isinstance(model, dict) and '_grouped_data' not in model:
raise Exception('Unsupported model. model requires _grouped_data.')
if isinstance(model, dict):
groups = model['_grouped_data']['groups']
group_by = model['_grouped_data']['group_by']
if isinstance(table, pd.DataFrame):
table, groups = _group(
table, params,
group_by) # use group keys from table even there is a model.
sample_result = _sample_result(function, table, model, params, groups)
res_keys, df_keys, model_keys_containing_repr = _info_from_sample_result(
sample_result, group_by, groups)
res_dict, success_keys = _function_by_group_key(function, table, model,
params, groups, res_keys,
group_by)
for repr_key in model_keys_containing_repr:
rb = BrtcReprBuilder()
for group in success_keys:
rb.addMD(
'--- \n\n ### Group by {group_by} : {tmp_group}\n\n---'.format(
group_by=group_by, tmp_group=group))
rb.merge(res_dict[repr_key]['_grouped_data']['data'][tuple(group)]
['_repr_brtc_'])
res_dict[repr_key]['_repr_brtc_'] = rb.get()
for df_key in df_keys:
res_dict[df_key] = _flatten(res_dict[df_key], groups, group_by,
columns)
return res_dict
def _function_by_group(function,
table=None,
model=None,
group_by=None,
**params):
if table is None and model is None:
raise Exception(
'This function requires at least one of a table or a model')
if isinstance(table, pd.DataFrame) and model is None and group_by is None:
raise Exception('This function requires group_by.')
if isinstance(model, dict) and '_grouped_data' not in model:
raise Exception('Unsupported model. model requires _grouped_data.')
if isinstance(model, dict):
group_key_dict = model['_grouped_data']['group_key_dict']
group_by = model['_grouped_data']['group_by']
if isinstance(table, pd.DataFrame):
table, group_key_dict = _group(
table,
group_by) # use group keys from table even there is a model.
print('Number of groups: {}'.format(len(group_key_dict)))
print('group_by: {}'.format(group_by))
print('group_key_dict: {}'.format(group_key_dict))
sample_result = _sample_result(function, table, model, params,
group_key_dict)
res_keys, df_keys, model_keys_containing_repr = _info_from_sample_result(
sample_result, group_by, group_key_dict)
res_dict, success_keys = _function_by_group_key(function, table, model,
params, group_key_dict,
res_keys, group_by)
print(res_dict)
print(success_keys)
for repr_key in model_keys_containing_repr:
rb = BrtcReprBuilder()
for group in success_keys:
rb.addMD(
'--- \n\n ### Group by {group_by} : {tmp_group}\n\n---'.format(
group_by=group_by, tmp_group=group_key_dict[group]))
rb.merge(res_dict[repr_key]['_grouped_data']['data'][group]
['_repr_brtc_'])
res_dict[repr_key]['_repr_brtc_'] = rb.get()
for df_key in df_keys:
res_dict[df_key] = _flatten(res_dict[df_key])
return res_dict
def _function_by_group(function, table, model=None, group_by=None, **params):
if isinstance(table, pd.DataFrame) and group_by is not None:
table, group_keys = _group(table, group_by)
else:
group_keys = table['_grouped_data'].keys()
sample_group = group_keys[0]
if model is None:
sample_result = function(table=table['_grouped_data'][sample_group],
**params)
else:
sample_result = function(table=table['_grouped_data'][sample_group],
model=model['_grouped_data'][sample_group],
**params)
res_keys = sample_result.keys()
df_keys = [
k for k, v in sample_result.items() if isinstance(v, pd.DataFrame)
]
model_keys_containing_repr = [
k for k, v in sample_result.items()
if isinstance(v, dict) and '_repr_brtc_' in v
]
res_dict = dict()
for res_key in res_keys:
res_dict[res_key] = {'_grouped_data': dict()}
for group in group_keys:
if model is None:
res_group = function(table=table['_grouped_data'][group], **params)
else:
res_group = function(table=table['_grouped_data'][group],
model=model['_grouped_data'][group],
**params)
for res_key in res_keys:
res_dict[res_key]['_grouped_data'][group] = res_group[res_key]
for repr_key in model_keys_containing_repr:
rb = BrtcReprBuilder()
for group in group_keys:
rb.addMD('{group}'.format(group=group))
rb.merge(res_dict[repr_key]['_grouped_data'][group]['_repr_brtc_'])
res_dict[repr_key]['_repr_brtc_'] = rb.get()
for df_key in df_keys:
res_dict[df_key] = _flatten(res_dict[df_key])
return res_dict
