def build_column_dict(data_frame, sql):
"""
Build a dictionary with key column position and value the list of
related items (predicate and objects).
:param data_frame: The Pandas dataframe.
:param sql: The sql query.
:return:The column dictionary.
"""
column_dict = dict()
for line in sql.splitlines():
left_stripped_line = line.lstrip(' ')
words = left_stripped_line.split(' ')
declare = words[0]
declare_token = '--JOIN'
if declare == declare_token:
table_and_column = words[1]
words = table_and_column.split('.')
table_name = words[0]
column_name = words[1]
column_description = get_column_description(table_name,
column_name)
index = column_position_in_dataframe(column_description,
data_frame)
column_dict[index] = dict()
column_dict[index]['table'] = table_name
column_dict[index]['column'] = column_name
col_dict = dict()
for c, col in enumerate(data_frame.columns):
if c not in column_dict:
continue
table_name = column_dict[c]['table']
column_name = column_dict[c]['column']
rows = get_metadata_on_column(table_name, column_name)
for row in rows:
t_predicate = "%s" % row[3]
t_object = "%s" % row[4]
if '#' in t_predicate:
t_predicate = t_predicate.split('#')[1]
else:
continue
if '#' in t_object:
t_object = t_object.split('#')[1]
else:
continue
item = Item(t_predicate.strip(), t_object.strip())
if c in col_dict:
col_dict[c].append(item)
else:
elements = [item]
col_dict[c] = elements
return col_dict
def reconciles_data_frame(df, sql):
"""
Reconciles data frame using url instead of descriptions.
REGARDS: Now this function works only on un-pivoted, plain data frame.
:param df: Data frame.
:param sql: The query sql code.
:return: Reconciled Data frame.
"""
st = detect_special_columns(sql)
fks_t = dict()
code_to_url_col = dict()
desc_to_code_col = dict()
for key in st.cols:
value = st.cols[key]
column = value['column']
table = value['table']
column_desc = get_column_description(table, column)
if not column_desc in df.columns:
# It is not used in the query.
continue
if not table in fks_t:
fks = build_foreign_keys(table)
fks_t[table] = fks
else:
fks = fks_t[table]
if column in fks:
fk = fks[column]
code_to_url = build_code_to_url_mapping(fk)
if len(code_to_url) != 0:
# It contains some reconciliation rows.
code_to_url_col[column_desc] = code_to_url
desc_to_code = build_desc_to_code_mapping(fk)
desc_to_code_col[column_desc] = desc_to_code
for n, col_name in enumerate(df.columns):
if col_name is None or col_name not in code_to_url_col:
continue
code_to_url = code_to_url_col[col_name]
desc_to_code = desc_to_code_col[col_name]
c_position = df.columns.get_loc(col_name)
values = df[col_name]
for v, value in enumerate(values):
if value in desc_to_code:
code = desc_to_code[value]
url = code_to_url[code]
df.iloc[v, c_position] = url
return df
