def test_upsert_individual_values1(pandabase_loaded_db, constants):
"""upsert to update rows with only 1 of 5 values (and index) from full dataframe"""
assert pb.has_table(pandabase_loaded_db, constants.TABLE_NAME)
df = pb.read_sql(constants.TABLE_NAME, con=pandabase_loaded_db)
df2 = pd.DataFrame(index=df.index, columns=df.columns)
for col in df2.columns:
df2[col] = df2[col].astype(df[col].dtype)
df2.loc[df2.index[0], 'float'] = 9.9
df2.loc[df2.index[1], 'integer'] = 999
df2.loc[df2.index[2], 'string'] = 'nah'
df2.loc[df2.index[3], 'date'] = pd.to_datetime('1968-01-01', utc=True)
pb.to_sql(df2,
table_name=constants.TABLE_NAME,
con=pandabase_loaded_db,
how='upsert')
# check against pandabase read
loaded = pb.read_sql(constants.TABLE_NAME, con=pandabase_loaded_db)
df.loc[df.index[0], 'float'] = 9.9
df.loc[df.index[1], 'integer'] = 999
df.loc[df.index[2], 'string'] = 'nah'
df.loc[df.index[3], 'date'] = pd.to_datetime('1968-01-01', utc=True)
assert companda(df, loaded)
def test_create_select_table_range_datetime_index(empty_db, simple_df,
constants):
"""add a new table with explicit index, read it back with pandabase, check equality"""
simple_df.index = simple_df.date
simple_df = simple_df.drop('date', axis=1)
table = pb.to_sql(simple_df,
table_name='sample',
con=empty_db,
how='create_only')
# print(table.columns)
assert table.columns['date'].primary_key
assert pb.has_table(empty_db, 'sample')
loaded0 = pb.read_sql('sample',
con=empty_db,
lowest=simple_df.index[-1],
highest=simple_df.index[0])
print(loaded0)
assert len(loaded0) == 0
loaded = pb.read_sql('sample',
con=empty_db,
lowest=simple_df.index[0],
highest=simple_df.index[-1])
assert pb.companda(loaded, simple_df, ignore_all_nan_columns=True)
def test_upsert_individual_values2(pandabase_loaded_db, constants):
"""upsert to update rows with only 1 of 5 values (and index) from incomplete DataFrame"""
assert pb.has_table(pandabase_loaded_db, constants.TABLE_NAME)
df = pb.read_sql(constants.TABLE_NAME, con=pandabase_loaded_db)
df2 = pd.DataFrame(index=df.index, columns=df.columns)
for col in df2.columns:
df2[col] = df2[col].astype(df[col].dtype)
df2.loc[df2.index[0], 'float'] = 9.9
df2.loc[df2.index[3], 'date'] = pd.to_datetime('1968-01-01', utc=True)
pb.to_sql(pd.DataFrame(index=df2.index[:1], columns=['float'], data=[9.9]),
table_name=constants.TABLE_NAME,
con=pandabase_loaded_db,
how='upsert')
pb.to_sql(pd.DataFrame(index=df2.index[3:4],
columns=['date'],
data=[pd.to_datetime('1968-01-01', utc=True)]),
table_name=constants.TABLE_NAME,
con=pandabase_loaded_db,
how='upsert')
# check against pandabase read
loaded = pb.read_sql(constants.TABLE_NAME, con=pandabase_loaded_db)
df.loc[df.index[0], 'float'] = 9.9
df.loc[df.index[3], 'date'] = pd.to_datetime('1968-01-01', utc=True)
assert companda(df, loaded)
def test_select_fails_multi_index(empty_db, multi_index_df, lowest):
"""add a new minimal table & read it back with pandabase - select all"""
pb.to_sql(multi_index_df,
table_name='sample_mi',
con=empty_db,
how='create_only',
)
with pytest.raises(Exception):
pb.read_sql(con=empty_db, table_name='sample_mi', highest=(1000, 1000), lowest=lowest)
def test_auto_index_add_valid_bool(minimal_df, empty_db, constants):
pb.to_sql(
minimal_df,
table_name=constants.TABLE_NAME,
con=empty_db,
how='create_only',
auto_index=True,
)
assert pb.has_table(empty_db, constants.TABLE_NAME)
df = pd.DataFrame(index=[101, 102, 103],
columns=['boolean'],
data=[True, False, None])
pb.to_sql(
df,
table_name=constants.TABLE_NAME,
con=empty_db,
how='append',
auto_index=True,
)
df = pb.read_sql(constants.TABLE_NAME, con=empty_db)
# Int64Dtype is a fine way to store nullable boolean values
# Stored in database as boolean or NULL so the data can only be 0, 1, or None
assert is_bool_dtype(df.boolean) or is_integer_dtype(df.boolean)
# assume values were loaded in order:
x = len(df)
assert df.loc[x - 2, 'boolean']
assert not df.loc[x - 1, 'boolean']
assert pd.np.isnan(df.loc[x, 'boolean'])
with pytest.raises(KeyError):
_ = df.loc[x + 1, 'boolean']
def test_upsert_new_cols(pandabase_loaded_db, constants, col_to_duplicate):
"""upsert new rows with only 1 of 5 values (and index)"""
assert pb.has_table(pandabase_loaded_db, constants.TABLE_NAME)
df = pb.read_sql(constants.TABLE_NAME, con=pandabase_loaded_db)
df['bonus_col'] = df[col_to_duplicate].copy()
pb.to_sql(df,
table_name=constants.TABLE_NAME,
con=pandabase_loaded_db,
how='upsert',
add_new_columns=True)
# check against pandabase read
loaded = pb.read_sql(constants.TABLE_NAME, con=pandabase_loaded_db)
assert companda(df, loaded)
assert 'bonus_col' in df.columns
def test_select_table_range_fails_different_index(empty_db, simple_df,
constants):
"""add a new table with explicit index, read it back with pandabase, check equality"""
simple_df.index = simple_df.date
simple_df = simple_df.drop('date', axis=1)
pb.to_sql(simple_df, table_name='sample', con=empty_db, how='create_only')
with pytest.raises(Exception):
_ = pb.read_sql('sample', con=empty_db, lowest=0, highest=12)
def test_create_table_multi_index(empty_db, multi_index_df_4, how):
"""add a new minimal table & read it back with pandabase"""
table = pb.to_sql(multi_index_df_4,
table_name='sample_mi',
con=empty_db,
how=how,
)
loaded = pb.read_sql(con=empty_db, table_name='sample_mi')
assert companda(multi_index_df_4, loaded)
def test_select_some_multi_index(empty_db, multi_index_df, lowest, length):
"""add a new minimal table & read it back with pandabase - select all"""
table = pb.to_sql(multi_index_df,
table_name='sample_mi',
con=empty_db,
how='create_only',
)
loaded = pb.read_sql(con=empty_db, table_name='sample_mi', highest=(1000, 1000), lowest=lowest)
print('\n', loaded)
assert len(loaded) == length
def test_upsert_complete_rows(pandabase_loaded_db, constants):
"""upsert, changing individual values"""
assert pb.has_table(pandabase_loaded_db, constants.TABLE_NAME)
df = pb.read_sql(constants.TABLE_NAME, con=pandabase_loaded_db)
assert df.date.dt.tz == UTC
df.loc[778, 'float'] = 9.9
df.loc[779, 'integer'] = 999
df.loc[780, 'string'] = 'nah'
df.loc[781, 'date'] = pd.to_datetime('1968-01-01', utc=True)
# check that all values still exist
assert df.loc[1, 'integer'] == 778
assert df.date.dt.tz == UTC
pb.to_sql(df,
table_name=constants.TABLE_NAME,
con=pandabase_loaded_db,
how='upsert')
loaded = pb.read_sql(constants.TABLE_NAME, con=pandabase_loaded_db)
assert companda(df, loaded)
def test_create_select_table_index(session_db, simple_df, constants):
"""add a new table with explicit index, read it back with pandabase, check equality"""
table = pb.to_sql(simple_df,
table_name='sample',
con=session_db,
how='create_only')
# print(table.columns)
assert table.columns[constants.SAMPLE_INDEX_NAME].primary_key
assert pb.has_table(session_db, 'sample')
loaded = pb.read_sql('sample', con=session_db)
assert pb.companda(loaded, simple_df, ignore_all_nan_columns=True)
def test_upsert_incomplete_rows(pandabase_loaded_db, constants):
"""upsert new rows with only 1 of 5 values (and index)"""
assert pb.has_table(pandabase_loaded_db, constants.TABLE_NAME)
df = pb.read_sql(constants.TABLE_NAME, con=pandabase_loaded_db)
df.loc[11, 'float'] = 9.9
df.loc[12, 'integer'] = 999
df.loc[13, 'string'] = 'nah'
df.loc[14, 'date'] = pd.to_datetime('1968-01-01', utc=True)
# check that these values exist
assert df.loc[1, 'integer'] == 778
assert pd.isna(df.loc[11, 'integer'])
assert df.loc[13, 'string'] == 'nah'
pb.to_sql(df,
table_name=constants.TABLE_NAME,
con=pandabase_loaded_db,
how='upsert')
# check against pandabase read
loaded = pb.read_sql(constants.TABLE_NAME, con=pandabase_loaded_db)
assert companda(df, loaded)
def test_create_read_table_no_index(empty_db, minimal_df):
"""add a new minimal table & read it back with pandabase"""
table = pb.to_sql(minimal_df,
table_name='sample',
con=empty_db,
how='create_only',
auto_index=True,
)
# print(table.columns)
assert table.columns[PANDABASE_DEFAULT_INDEX].primary_key
loaded = pb.read_sql('sample', con=empty_db)
assert pb.has_table(empty_db, 'sample')
assert pb.companda(loaded, minimal_df, ignore_index=True)
def test_add_column_to_database(pandabase_loaded_db, actually_do, constants):
"""possibly add new column to db"""
name = 'a_new_column'
col = sqa.Column(name, primary_key=False, type_=Integer, nullable=True)
if actually_do:
pb.add_columns_to_db(col,
table_name=constants.TABLE_NAME,
con=pandabase_loaded_db)
df = pb.read_sql(table_name=constants.TABLE_NAME, con=pandabase_loaded_db)
if actually_do:
assert name in df.columns
assert is_integer_dtype(df[name])
else:
assert name not in df.columns
def test_select_all_multi_index(empty_db, multi_index_df):
"""add a new minimal table & read it back with pandabase - select all"""
table = pb.to_sql(multi_index_df,
table_name='sample_mi',
con=empty_db,
how='create_only',
)
# print(table.columns)
assert table.columns['this'].primary_key
assert table.columns['that'].primary_key
loaded = pb.read_sql(con=empty_db, table_name='sample_mi', highest=(100, 100), lowest=(0, 0))
print('\n', loaded)
assert companda(multi_index_df, loaded)
def test_create_table_multi_index(empty_db, multi_index_df, how):
"""add a new minimal table & read it back with pandabase"""
table = pb.to_sql(multi_index_df,
table_name='sample_mi',
con=empty_db,
how=how,
)
# print(table.columns)
assert table.columns['this'].primary_key
assert table.columns['that'].primary_key
loaded = pb.read_sql(con=empty_db, table_name='sample_mi')
print('\n', loaded)
assert companda(multi_index_df, loaded)
def test_coerce_integer(pandabase_loaded_db, how, constants):
"""insert an integer into float column"""
assert pb.has_table(pandabase_loaded_db, constants.TABLE_NAME)
df = pd.DataFrame(index=[1], columns=['integer'], data=[[77.0]])
df.index.name = constants.SAMPLE_INDEX_NAME
types = df.dtypes
pb.to_sql(df,
table_name=constants.TABLE_NAME,
con=pandabase_loaded_db,
how='upsert')
for col in df.columns:
assert types[col] == df.dtypes[col]
loaded = pb.read_sql(constants.TABLE_NAME, con=pandabase_loaded_db)
assert loaded.loc[1, 'integer'] == 77
def test_add_new_rows(pandabase_loaded_db, simple_df, how, constants):
"""upsert or append new complete rows"""
assert pb.has_table(pandabase_loaded_db, constants.TABLE_NAME)
df = simple_df.copy()
df.index = df.index + 100
pb.to_sql(df,
table_name=constants.TABLE_NAME,
con=pandabase_loaded_db,
how=how)
loaded = pb.read_sql(constants.TABLE_NAME, con=pandabase_loaded_db)
# print('loaded post-upsert by pandabase:')
# print(loaded)
assert loaded.isna().sum().sum() == 0
assert companda(simple_df, loaded.loc[simple_df.index])
assert companda(df, loaded.loc[df.index])
def test_create_read_table_with_different_index(session_db, simple_df, table_name, index_col_name):
"""create new tables in empty db, using different col types as index, read with pandabase"""
orig_df = simple_df.copy()
orig_df.index = orig_df[index_col_name]
print(orig_df[index_col_name])
print(orig_df.index)
orig_df = orig_df.drop(index_col_name, axis=1)
table = pb.to_sql(orig_df,
table_name=table_name,
con=session_db,
how='create_only')
assert table.columns[index_col_name].primary_key
assert pb.has_table(session_db, table_name)
loaded = pb.read_sql(table_name, con=session_db)
c = pb.companda(loaded, orig_df, ignore_all_nan_columns=True)
if not c:
raise ValueError(c.message)
def test_append_auto_index(empty_db, minimal_df):
"""add a new minimal table; add it again"""
pb.to_sql(minimal_df,
table_name='sample',
con=empty_db,
auto_index=True,
how='create_only')
table2 = pb.to_sql(minimal_df,
table_name='sample',
con=empty_db,
auto_index=True,
how='append')
assert table2.columns[PANDABASE_DEFAULT_INDEX].primary_key
loaded = pb.read_sql('sample', con=empty_db)
assert pb.has_table(empty_db, 'sample')
double_df = pd.concat([minimal_df, minimal_df], ignore_index=True)
assert pb.companda(loaded, double_df, ignore_index=True)
assert len(loaded) == len(minimal_df) * 2
def test_coerce_float_to_integer_multi(multi_index_df, empty_db, constants):
"""insert a float into integer column"""
mi = multi_index_df.copy().index
assert mi.names[0] is not None
assert mi.names[1] is not None
pb.to_sql(multi_index_df, con=empty_db, table_name=constants.TABLE_NAME)
df = pd.DataFrame(columns=['integer'], data=[[77.0]], index=mi[:1])
assert df.index.names[0] is not None
assert df.index.names[1] is not None
print('\n', df)
print()
types = df.dtypes
pb.to_sql(df, table_name=constants.TABLE_NAME, con=empty_db, how='upsert')
for col in df.columns:
assert types[col] == df.dtypes[col]
loaded = pb.read_sql(constants.TABLE_NAME, con=empty_db)
assert loaded.loc[mi[0], 'integer'] == 77
def test_select_pandas_table(pandas_loaded_db, simple_df, constants):
"""using pandabase.read_sql:
read pandas-written table containing simple_df,
this test fails because: when pandas writes the entry, it does not create
an explicit primary key. the table is treated as a multiindex"""
assert has_table(pandas_loaded_db, constants.TABLE_NAME)
df = pb.read_sql(constants.TABLE_NAME, pandas_loaded_db)
# line up pk since Pandas doesn't deal with it well
simple_df[simple_df.index.name] = simple_df.index
simple_df.index.name = None
orig_columns = make_clean_columns_dict(simple_df)
loaded_columns = make_clean_columns_dict(df)
for key in orig_columns.keys():
print(key)
if key == 'nan':
continue
assert_sqla_types_equivalent(orig_columns[key], loaded_columns[key])
assert companda(df, simple_df)
def test_upsert_valid_bool(pandabase_loaded_db, how, constants):
assert pb.has_table(pandabase_loaded_db, constants.TABLE_NAME)
df = pd.DataFrame(index=[101, 102, 103],
columns=['boolean'],
data=[True, False, None])
df.index.name = constants.SAMPLE_INDEX_NAME
pb.to_sql(df,
table_name=constants.TABLE_NAME,
con=pandabase_loaded_db,
how=how)
df = pb.read_sql(constants.TABLE_NAME, con=pandabase_loaded_db)
# Int64Dtype is a fine way to store nullable boolean values
# Stored in database as boolean or NULL so the data can only be 0, 1, or None
assert is_bool_dtype(df.boolean) or is_integer_dtype(df.boolean)
assert df.loc[101, 'boolean']
assert not df.loc[102, 'boolean']
assert pd.np.isnan(df.loc[103, 'boolean'])
with pytest.raises(KeyError):
_ = df.loc[104, 'boolean']