Skip to content

Commit

Permalink
Merge branch 'fix_0.11.0_region_mask' into dev
Browse files Browse the repository at this point in the history
  • Loading branch information
@ctuguinay
ctuguinay authored Sep 30, 2023
2 parents 8249807 + 2276b11 commit db19c39
Show file tree
Hide file tree
Showing 8 changed files with 529 additions and 271 deletions.
4 changes: 2 additions & 2 deletions echoregions/core.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -11,9 +11,9 @@ def read_evr(
----------
filepath : str, Path object
A valid path to an EVR file
min_depth : float, default ``None``
min_depth : float, default 0
Depth value in meters to set -9999.99 depth edges to.
max_depth : float, default ``None``
max_depth : float, default 1000
Depth value in meters to set 9999.99 depth edges to.
Returns
Expand Down
202 changes: 125 additions & 77 deletions echoregions/regions2d/regions2d.py
View file
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
import warnings
from pathlib import Path
from typing import Iterable, List, Union
from typing import Iterable, List, Optional, Union

import matplotlib
import matplotlib.pyplot as plt
Expand All @@ -9,8 +9,9 @@
import regionmask
import xarray as xr
from pandas import DataFrame, Series, Timestamp, isna
from xarray import DataArray
from xarray import DataArray, Dataset

from ..utils.api import convert_mask_3d_to_2d
from ..utils.io import validate_path
from ..utils.time import parse_simrad_fname_time
from .regions2d_parser import parse_regions_file
Expand All @@ -31,8 +32,8 @@ def __init__(
self.data = parse_regions_file(input_file)
self.output_file = []

self.max_depth = max_depth
self.min_depth = min_depth
self.max_depth = max_depth

def __iter__(self) -> Iterable:
return self.data.iterrows()
Expand Down Expand Up @@ -79,10 +80,10 @@ def select_region(
----------
region_id : float, int, str, list, ``None``
A region id provided as a number, a string, or list of these.
time_range: List of 2 Pandas Timestamps.
time_range : List of 2 Pandas Timestamps.
Datetime range for expected output of subselected DataFrame. 1st
index value must be later than 0th index value.
depth_range: List of 2 floats.
depth_range : List of 2 floats.
Depth range for expected output of subselected DataFrame. 1st
index value must be larger than 0th index value.
copy : bool
Expand Down Expand Up @@ -201,7 +202,7 @@ def close_region(self, region_id: List = None) -> DataFrame:
Parameters
----------
region : float, str, list, Series, DataFrame, ``None``
region_id : List, ``None``
region(s) to select raw files with
If ``None``, select all regions. Defaults to ``None``
Expand Down Expand Up @@ -322,16 +323,16 @@ def swap_val(val: Union[int, float]) -> Union[int, float]:

def plot(
self,
region: Union[str, List, DataFrame] = None,
close_region: bool = False,
region_id: List = None,
close_regions: bool = False,
**kwargs,
) -> None:
"""Plot a region from data.
Automatically convert time and range_edges.
Parameters
---------
region : float, str, list, Series, DataFrame, ``None``
region_id : List, ``None``
Region(s) to select raw files with
If ``None``, select all regions. Defaults to ``None``
close_region : bool
Expand All @@ -341,46 +342,64 @@ def plot(
"""

# Ensure that region is a DataFrame
region = self.select_region(region)
region = self.select_region(region_id)

if close_region:
if close_regions:
region = self.close_region(region)
for _, row in region.iterrows():
plt.plot(row["time"], row["depth"], **kwargs)

def mask(
self,
da_Sv: DataArray,
region_ids: List,
mask_var: str = None,
mask_labels=None,
) -> DataArray:
"""Mask data from Data Array containing Sv data based off of a Regions2D object
region_id: List = None,
mask_name: str = "mask",
mask_labels: dict = None,
collapse_to_2d: bool = False,
) -> Optional[Dataset]:
"""
Mask data from Data Array containing Sv data based off of a Regions2D object
and its regions ids.
Parameters
----------
da_Sv : Data Array
DataArray of shape (ping_time, depth) containing Sv data.
region_ids : list
region_id : list
list IDs of regions to create mask for
mask_var : str
mask_name : str
If provided, used to name the output mask array, otherwise `mask`
mask_labels:
None: assigns labels automatically 0,1,2,...
"from_ids": uses the region ids
list: uses a list of integers as labels
mask_labels : dict
If provided, used to label the region_id dimension of the output mask.
collapse_to_2d : bool
If true, then converts 3d mask to 2d mask. If not, keeps output as 3d mask.
Returns
-------
A DataArray with the data_var masked by the specified region.
mask_ds : Dataset
Either a 3D mask or a 2D mask based on the conditions below.
If collapse_to_2d is False:
A 3D mask where each layer of the mask will contain a 1s/0s mask for each
unique label in the 2D mask. The layers will be labeled via region_id
values. The slices of the 3D array will be in the form of 1s/0s: masked areas,
and non-masked areas.
If collapse_to_2d is True:
A 2D mask where each individual data points will be in the form of integers,
demarking region_id of masked regions, and nan values, demarking non-masked
areas.
Also contains a data variable (`mask_labels`) with mask labels
corresponding to region_id values.
"""
if isinstance(region_ids, list):
if len(region_ids) == 0:
raise ValueError("region_ids is empty. Cannot be empty.")
if isinstance(region_id, list):
if len(region_id) == 0:
raise ValueError("region_id is empty. Cannot be empty.")
elif region_id is None:
# Extract all region_id values
region_id = self.data.region_id.astype(int).to_list()
else:
raise TypeError(
f"region_ids must be of type list. Currently is of type {type(region_ids)}"
f"region_id must be of type list. Currently is of type {type(region_id)}"
)

if mask_labels is None:
Expand All @@ -395,75 +414,103 @@ def mask(
"mask_labels as None."
)

# Replace nan depth in regions2d.
self.replace_nan_depth(inplace=True)

# Dataframe containing region information.
region_df = self.select_region(region_ids)
region_df = self.select_region(region_id)

# Select only columns which are important.
# Select only important columns
region_df = region_df[["region_id", "time", "depth"]]

# Organize the regions in a format for region mask.
df = region_df.explode(["time", "depth"])
# Filter for rows with depth values within self min and self max depth and
# for rows that have positive depth.
region_df = region_df[
region_df["depth"].apply(
lambda x: all(
(i >= np.max(0, int(self.min_depth)) and i <= int(self.max_depth))
for i in x
)
)
]

# Convert region time to integer timestamp.
df["time"] = matplotlib.dates.date2num(df["time"])
if region_df.empty:
print(
"All rows in Regions DataFrame have NaN Depth values after filtering depth "
"between min_depth and max_depth."
)
else:
# Organize the regions in a format for region mask.
df = region_df.explode(["time", "depth"])

# Create a list of dataframes for each regions.
grouped = list(df.groupby("region_id"))
# Convert region time to integer timestamp.
df["time"] = matplotlib.dates.date2num(df["time"])

# Convert to list of numpy arrays which is an acceptable format to create region mask.
regions_np = [np.array(region[["time", "depth"]]) for id, region in grouped]
# Create a list of dataframes for each regions.
grouped = list(df.groupby("region_id"))

# Corresponding region ids converted to int.
region_ids = [int(id) for id, region in grouped]
# Convert to list of numpy arrays which is an acceptable format to create region mask.
regions_np = [np.array(region[["time", "depth"]]) for _, region in grouped]

# Convert ping_time to unix_time since the masking does not work on datetime objects.
da_Sv = da_Sv.assign_coords(
unix_time=(
"ping_time",
matplotlib.dates.date2num(da_Sv.coords["ping_time"].values),
)
)
# Convert corresponding region_id to int.
filtered_region_id = [int(id) for id, _ in grouped]

# Set up mask labels.
if mask_labels == "from_ids":
r = regionmask.Regions(outlines=regions_np, numbers=region_ids)
elif isinstance(mask_labels, list) or mask_labels is None:
r = regionmask.Regions(outlines=regions_np)
else:
raise ValueError("mask_labels must be None, 'from_ids', or a list.")
# Convert ping_time to unix_time since the masking does not work on datetime objects.
da_Sv = da_Sv.assign_coords(
unix_time=(
"ping_time",
matplotlib.dates.date2num(da_Sv.coords["ping_time"].values),
)
)

# Create mask
try:
M = r.mask(
# Create mask
r = regionmask.Regions(
outlines=regions_np, names=filtered_region_id, name=mask_name
)
mask_da = r.mask_3D(
da_Sv["unix_time"],
da_Sv["depth"],
wrap_lon=False,
).astype(
int
) # This maps False to 0 and True to 1

# Replace region coords with region_id coords
mask_da = mask_da.rename({"names": "region_id"})
mask_da = mask_da.swap_dims({"region": "region_id"})

# Remove all coords other than depth, ping_time, region_id
mask_da = mask_da.drop_vars(
mask_da.coords._names.difference({"depth", "ping_time", "region_id"})
)
except ValueError as ve:
warnings.warn(
"Most likely using deprecated regionmask version."
"Make sure to use regionmask==0.8.0 or more recent versions.",
DeprecationWarning,
stacklevel=2,

# Transpose mask_da
mask_da = mask_da.transpose("region_id", "depth", "ping_time")

# Remove attribute standard_name
mask_da.attrs.pop("standard_name")

# Rename Data Array to mask_name
mask_da = mask_da.rename(mask_name)

# Set mask_labels_da
masked_region_id = mask_da.region_id.values.tolist()
subset_mask_labels = [
mask_labels[key] for key in masked_region_id if key in mask_labels
]
mask_labels_da = xr.DataArray(
subset_mask_labels,
dims="region_id",
coords={"region_id": masked_region_id},
)
raise ve

if isinstance(mask_labels, list):
# Convert default labels to mask_labels.
S = xr.where(~M.isnull(), 0, M)
S = M
for idx, label in enumerate(mask_labels):
S = xr.where(M == idx, label, S)
M = S
# Create dataset
mask_ds = xr.Dataset()
mask_ds["mask_labels"] = mask_labels_da
mask_ds["mask_3d"] = mask_da

# Assign specific name to mask array, otherwise 'mask'.
if mask_var:
M = M.rename(mask_var)
if collapse_to_2d:
# Convert 3d mask to 2d mask
mask_ds = convert_mask_3d_to_2d(mask_ds)

return M
return mask_ds

def transect_mask(
self,
Expand Down Expand Up @@ -493,6 +540,7 @@ def transect_mask(
bbox_distance_threshold: float
The maximum value for how far apart the left and right bounding box for each transect
value region. Default is set to 1 minute.
Returns
-------
M : Data Array
Expand Down
4 changes: 2 additions & 2 deletions echoregions/test_data/transect.evr
View file
Original file line number Diff line number Diff line change
Expand Up @@ -135,11 +135,11 @@ Log
20190702 2024556635 -9999.9900000000 20190702 2024556635 9999.9900000000 20190702 2024570135 9999.9900000000 20190702 2024570135 -9999.9900000000 2
COM

13 4 18 0 4 -1 1 20190702 1811518097 9.2447583998 20190702 1952215310 758.9732173069
13 4 18 0 4 -1 1 20190702 1840518097 9.2447583998 20190702 1952215310 758.9732173069
0
0
Trawl
20190702 1811518097 9.2447583998 20190702 1811518097 758.9732173069 20190702 1952215310 758.9732173069 20190702 1952215310 9.2447583998 0
20190702 1840518097 9.2447583998 20190702 1840518097 700 20190702 1952215310 700 20190702 1952215310 9.2447583998 0
AWT20

13 4 19 0 2 -1 1 20190702 2110200695 -10011.7976284000 20190702 2110214100 10011.2963773964
Expand Down
Loading

0 comments on commit db19c39

Please sign in to comment.