Using regex package for match (#408)

* Using regex package for match

The built-in re package does not allow for global flags like "(?i)" to be anywhere but the start of a pattern string now. The package `regex` still allows this, so it is optionally used for the `match` function if available.
* test added
* updating whats new

* Update whats-new.rst

* updates to PR

* changed to import full modules due to mypy error

* oops on the previous commit. hopefully better now.

* added regex to environment.yaml

* updated custom-criteria.md

* Update doc/whats-new.rst

Co-authored-by: Deepak Cherian <[email protected]>

* Add ignore for mypy

* added link to regex package in docs and regex to doc env

* example in doc was not being shown but now is

* Update doc/custom-criteria.md

* Update doc/custom-criteria.md

Co-authored-by: Deepak Cherian <[email protected]>

* Update doc/custom-criteria.md

---------

Co-authored-by: Deepak Cherian <[email protected]>
Co-authored-by: dcherian <[email protected]>

Author: kthyng

cf_xarray/accessor.py

@@ -207,6 +207,11 @@ def _get_custom_criteria(
     List[str], Variable name(s) in parent xarray object that matches axis, coordinate, or custom `key`
     """
 
+    try:
+        from regex import match as regex_match
+    except ImportError:
+        from re import match as regex_match  # type: ignore
+
     if isinstance(obj, DataArray):
         obj = obj._to_temp_dataset()
 
@@ -223,13 +228,13 @@ def _get_custom_criteria(
     if key in criteria_map:
         for criterion, patterns in criteria_map[key].items():
             for var in obj.variables:
-                if re.match(patterns, obj[var].attrs.get(criterion, "")):
+                if regex_match(patterns, obj[var].attrs.get(criterion, "")):
                     results.update((var,))
                 # also check name specifically since not in attributes
                 elif (
                     criterion == "name"
                     and isinstance(var, str)
-                    and re.match(patterns, var)
+                    and regex_match(patterns, var)
                 ):
                     results.update((var,))
     return list(results)

cf_xarray/tests/__init__.py

@@ -67,3 +67,4 @@ def LooseVersion(vstring):
 has_scipy, requires_scipy = _importorskip("scipy")
 has_shapely, requires_shapely = _importorskip("shapely")
 has_pint, requires_pint = _importorskip("pint")
+has_regex, requires_regex = _importorskip("regex")

cf_xarray/tests/test_accessor.py

@@ -34,7 +34,13 @@
     rotds,
     vert,
 )
-from . import raise_if_dask_computes, requires_cftime, requires_pint, requires_scipy
+from . import (
+    raise_if_dask_computes,
+    requires_cftime,
+    requires_pint,
+    requires_regex,
+    requires_scipy,
+)
 
 mpl.use("Agg")
 
@@ -1585,6 +1591,17 @@ def test_custom_criteria() -> None:
         assert_identical(ds.cf["temp"], ds["temperature"])
 
 
+@requires_regex
+def test_regex_match():
+    # test that having a global regex expression flag later in the expression will work if
+    # regex is found
+    vocab = {"temp": {"name": "tem|(?i)temp"}}
+    ds = xr.Dataset()
+    ds["Tempblah"] = [0, 1, 2]
+    with cf_xarray.set_options(custom_criteria=vocab):
+        assert_identical(ds.cf["temp"], ds["Tempblah"])
+
+
 def test_cf_standard_name_table_version() -> None:
 
     url = (

ci/doc.yml

@@ -18,6 +18,7 @@ dependencies:
   - pandas
   - pooch
   - pint
+  - regex
   - furo
   - pip:
       - git+https://github.com/xarray-contrib/cf-xarray

ci/environment.yml

@@ -13,6 +13,7 @@ dependencies:
   - pandas
   - pint
   - pooch
+  - regex
   - scipy
   - shapely
   - xarray

doc/custom-criteria.md

@@ -102,3 +102,20 @@ cfxr.set_options(custom_criteria=salt_criteria)
 
 ds.cf[["salinity"]]
 ```
+
+## More complex matches with `regex`
+
+Here is an example of a more complicated custom criteria, which requires the package [`regex`](https://github.com/mrabarnett/mrab-regex) to be installed since a behavior (allowing global flags like "(?i)" for matching case insensitive) was recently deprecated in the `re` package. The custom criteria, called "vocab", matches – case insensitive – to the variable alias "sea_ice_u" a variable whose name includes "sea" and "ice" and "u" but not "qc" or "status", or "sea" and "ice" and "x" and "vel" but not "qc" or "status".
+
+```{code-cell}
+import cf_xarray as cfxr
+import xarray as xr
+
+vocab = {"sea_ice_u": {"name": "(?i)^(?!.*(qc|status))(?=.*sea)(?=.*ice)(?=.*u)|(?i)^(?!.*(qc|status))(?=.*sea)(?=.*ice)(?=.*x)(?=.*vel)"}}
+ds = xr.Dataset()
+ds["sea_ice_velocity_x"] = [0,1,2]
+
+with cfxr.set_options(custom_criteria=vocab):
+    seaiceu = ds.cf["sea_ice_u"]
+seaiceu
+```

doc/whats-new.rst

@@ -3,6 +3,11 @@
 What's New
 ----------
 
+v0.7.8 (unreleased)
+===================
+
+- Optionally use the `regex` package to continue supporting global flags in regular expressions that are not at start of pattern. (:pr:`408`). By `Kristen Thyng`_
+
 v0.7.7 (Jan 14, 2023)
 =====================
 

pyproject.toml

@@ -21,7 +21,7 @@ dependencies = [
 dynamic = ["version"]
 
 [project.optional-dependencies]
-all = ["matplotlib", "pint", "shapely"]
+all = ["matplotlib", "pint", "shapely", "regex"]
 
 [project.urls]
 homepage = "https://cf-xarray.readthedocs.io"