-
Notifications
You must be signed in to change notification settings - Fork 206
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
test(clean): add tests for clean_lat_long
- Loading branch information
1 parent
b59faa7
commit aea2602
Showing
2 changed files
with
304 additions
and
0 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,3 @@ | ||
""" | ||
dataprep.clean tests | ||
""" |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,301 @@ | ||
""" | ||
module for testing the functions clean_lat_long() and validate_lat_long() | ||
""" | ||
import logging | ||
|
||
import numpy as np | ||
import pandas as pd | ||
import pytest | ||
|
||
from ...clean import clean_lat_long, validate_lat_long | ||
|
||
LOGGER = logging.getLogger(__name__) | ||
|
||
|
||
@pytest.fixture(scope="module") # type: ignore | ||
def df_lat_long_column() -> pd.DataFrame: | ||
df = pd.DataFrame( | ||
{ | ||
"messy_lat_long": [ | ||
(41.5, -81.0), | ||
"41.5;-81.0", | ||
"41.5,-81.0", | ||
"41.5 -81.0", | ||
"41.5° N, 81.0° W", | ||
"41.5 S;81.0 E", | ||
"-41.5 S;81.0 E", | ||
"23 26m 22s N 23 27m 30s E", | ||
"23 26' 22\" N 23 27' 30\" E", | ||
"UT: N 39°20' 0'' / W 74°35' 0''", | ||
"hello", | ||
np.nan, | ||
"NULL", | ||
] | ||
} | ||
) | ||
return df | ||
|
||
|
||
@pytest.fixture(scope="module") # type: ignore | ||
def df_separate_lat_long_columns() -> pd.DataFrame: | ||
df = pd.DataFrame( | ||
{ | ||
"messy_lat": ["30° E", "41° 30′ N", "41 S", "80", "hello", "NA"], | ||
"messy_long": ["30° E", "41° 30′ N", "41 W", "80", "hello", "NA"], | ||
} | ||
) | ||
return df | ||
|
||
|
||
def test_clean_default(df_lat_long_column: pd.DataFrame) -> None: | ||
df_clean = clean_lat_long(df_lat_long_column, "messy_lat_long") | ||
df_check = df_lat_long_column.copy() | ||
df_check["messy_lat_long_clean"] = [ | ||
(41.5, -81.0), | ||
(41.5, -81.0), | ||
(41.5, -81.0), | ||
(41.5, -81.0), | ||
(41.5, -81.0), | ||
(-41.5, 81.0), | ||
np.nan, | ||
(23.4394, 23.4583), | ||
(23.4394, 23.4583), | ||
(39.3333, -74.5833), | ||
np.nan, | ||
np.nan, | ||
np.nan, | ||
] | ||
assert df_check.equals(df_clean) | ||
|
||
|
||
def test_clean_output_format(df_lat_long_column: pd.DataFrame) -> None: | ||
df_clean_ddh = clean_lat_long(df_lat_long_column, "messy_lat_long", output_format="ddh") | ||
df_clean_dms = clean_lat_long(df_lat_long_column, "messy_lat_long", output_format="dms") | ||
df_check_ddh = df_lat_long_column.copy() | ||
df_check_ddh["messy_lat_long_clean"] = [ | ||
"41.5° N, 81.0° W", | ||
"41.5° N, 81.0° W", | ||
"41.5° N, 81.0° W", | ||
"41.5° N, 81.0° W", | ||
"41.5° N, 81.0° W", | ||
"41.5° S, 81.0° E", | ||
np.nan, | ||
"23.4394° N, 23.4583° E", | ||
"23.4394° N, 23.4583° E", | ||
"39.3333° N, 74.5833° W", | ||
np.nan, | ||
np.nan, | ||
np.nan, | ||
] | ||
df_check_dms = df_lat_long_column.copy() | ||
df_check_dms["messy_lat_long_clean"] = [ | ||
"41° 30′ 0″ N, 81° 0′ 0″ W", | ||
"41° 30′ 0″ N, 81° 0′ 0″ W", | ||
"41° 30′ 0″ N, 81° 0′ 0″ W", | ||
"41° 30′ 0″ N, 81° 0′ 0″ W", | ||
"41° 30′ 0″ N, 81° 0′ 0″ W", | ||
"41° 30′ 0″ S, 81° 0′ 0″ E", | ||
np.nan, | ||
"23° 26′ 22″ N, 23° 27′ 30″ E", | ||
"23° 26′ 22″ N, 23° 27′ 30″ E", | ||
"39° 20′ 0″ N, 74° 34′ 60″ W", | ||
np.nan, | ||
np.nan, | ||
np.nan, | ||
] | ||
assert df_check_ddh.equals(df_clean_ddh) | ||
assert df_check_dms.equals(df_clean_dms) | ||
|
||
|
||
def test_clean_split(df_lat_long_column: pd.DataFrame) -> None: | ||
df_clean = clean_lat_long(df_lat_long_column, "messy_lat_long", split=True) | ||
df_check = df_lat_long_column.copy() | ||
df_check["latitude"] = [ | ||
41.5, | ||
41.5, | ||
41.5, | ||
41.5, | ||
41.5, | ||
-41.5, | ||
np.nan, | ||
23.4394, | ||
23.4394, | ||
39.3333, | ||
np.nan, | ||
np.nan, | ||
np.nan, | ||
] | ||
df_check["longitude"] = [ | ||
-81.0, | ||
-81.0, | ||
-81.0, | ||
-81.0, | ||
-81.0, | ||
81.0, | ||
np.nan, | ||
23.4583, | ||
23.4583, | ||
-74.5833, | ||
np.nan, | ||
np.nan, | ||
np.nan, | ||
] | ||
assert df_check.equals(df_clean) | ||
|
||
|
||
def test_clean_output_format_split(df_lat_long_column: pd.DataFrame) -> None: | ||
df_clean = clean_lat_long(df_lat_long_column, "messy_lat_long", output_format="dm", split=True) | ||
df_check = df_lat_long_column.copy() | ||
df_check["latitude"] = [ | ||
"41° 30′ N", | ||
"41° 30′ N", | ||
"41° 30′ N", | ||
"41° 30′ N", | ||
"41° 30′ N", | ||
"41° 30′ S", | ||
np.nan, | ||
"23° 26.3667′ N", | ||
"23° 26.3667′ N", | ||
"39° 20′ N", | ||
np.nan, | ||
np.nan, | ||
np.nan, | ||
] | ||
df_check["longitude"] = [ | ||
"81° 0′ W", | ||
"81° 0′ W", | ||
"81° 0′ W", | ||
"81° 0′ W", | ||
"81° 0′ W", | ||
"81° 0′ E", | ||
np.nan, | ||
"23° 27.5′ E", | ||
"23° 27.5′ E", | ||
"74° 35′ W", | ||
np.nan, | ||
np.nan, | ||
np.nan, | ||
] | ||
assert df_check.equals(df_clean) | ||
|
||
|
||
def test_clean_inplace(df_lat_long_column: pd.DataFrame) -> None: | ||
df_clean = clean_lat_long(df_lat_long_column, "messy_lat_long", inplace=True) | ||
df_check = pd.DataFrame( | ||
{ | ||
"messy_lat_long_clean": [ | ||
(41.5, -81.0), | ||
(41.5, -81.0), | ||
(41.5, -81.0), | ||
(41.5, -81.0), | ||
(41.5, -81.0), | ||
(-41.5, 81.0), | ||
np.nan, | ||
(23.4394, 23.4583), | ||
(23.4394, 23.4583), | ||
(39.3333, -74.5833), | ||
np.nan, | ||
np.nan, | ||
np.nan, | ||
] | ||
} | ||
) | ||
assert df_check.equals(df_clean) | ||
|
||
|
||
def test_clean_split_inplace(df_lat_long_column: pd.DataFrame) -> None: | ||
df_clean = clean_lat_long(df_lat_long_column, "messy_lat_long", split=True, inplace=True) | ||
df_check = pd.DataFrame( | ||
{ | ||
"latitude": [ | ||
41.5, | ||
41.5, | ||
41.5, | ||
41.5, | ||
41.5, | ||
-41.5, | ||
np.nan, | ||
23.4394, | ||
23.4394, | ||
39.3333, | ||
np.nan, | ||
np.nan, | ||
np.nan, | ||
], | ||
"longitude": [ | ||
-81.0, | ||
-81.0, | ||
-81.0, | ||
-81.0, | ||
-81.0, | ||
81.0, | ||
np.nan, | ||
23.4583, | ||
23.4583, | ||
-74.5833, | ||
np.nan, | ||
np.nan, | ||
np.nan, | ||
], | ||
} | ||
) | ||
assert df_check.equals(df_clean) | ||
|
||
|
||
def test_clean_lat_long_separate_columns_split(df_separate_lat_long_columns: pd.DataFrame) -> None: | ||
df_clean = clean_lat_long( | ||
df_separate_lat_long_columns, lat_col="messy_lat", long_col="messy_long", split=True | ||
) | ||
df_check = df_separate_lat_long_columns.copy() | ||
df_check["messy_lat_clean"] = [np.nan, 41.5, -41.0, 80.0, np.nan, np.nan] | ||
df_check["messy_long_clean"] = [30.0, np.nan, -41.0, 80.0, np.nan, np.nan] | ||
assert df_check.equals(df_clean) | ||
|
||
|
||
def test_validate_value() -> None: | ||
assert validate_lat_long("41° 30′ 0″ N") == False | ||
assert validate_lat_long("41.5 S;81.0 E") == True | ||
assert validate_lat_long("-41.5 S;81.0 E") == False | ||
assert validate_lat_long((41.5, 81)) == True | ||
assert validate_lat_long(41.5, lat_long=False, lat=True) == True | ||
|
||
|
||
def test_validate_series_lat_long(df_lat_long_column: pd.DataFrame) -> None: | ||
srs_valid = validate_lat_long(df_lat_long_column["messy_lat_long"]) | ||
srs_check = pd.Series( | ||
[ | ||
True, | ||
True, | ||
True, | ||
True, | ||
True, | ||
True, | ||
False, | ||
True, | ||
True, | ||
True, | ||
False, | ||
False, | ||
False, | ||
], | ||
name="messy_lat_long", | ||
) | ||
assert srs_check.equals(srs_valid) | ||
|
||
|
||
def test_validate_series_lat(df_separate_lat_long_columns: pd.DataFrame) -> None: | ||
srs_valid = validate_lat_long( | ||
df_separate_lat_long_columns["messy_lat"], lat_long=False, lat=True | ||
) | ||
srs_check = pd.Series( | ||
[ | ||
False, | ||
True, | ||
True, | ||
True, | ||
False, | ||
False, | ||
], | ||
name="messy_lat_clean", | ||
) | ||
assert srs_check.equals(srs_valid) |