environment_yml_parser: Fix faulty package name/version parsing

The code for extracting the package name from a Conda dependency
specifier was overly simplistic: it assumed that any version/build
info in the string always started with a "=" character. This is
obviously incorrect (e.g. "numpy>=1.23" is a simple counter-example).

Fix the parser to look for the same characters that the upstream looks
for, when separating package name from version + build info (see
https://github.com/conda/conda/blob/aa0fb6f3ae669a5ade575d340555aa6a9f71ef5e/conda/models/match_spec.py#L716
for details).

Add a test case copied from the cartopy project to illustrate the
correct use of such specifiers, and also add a couple of tests to verify
that we properly deal with invalid package names in dependency
specifiers (both Conda and Pip) inside an environment.yml file.

fawltydeps/extract_deps/environment_yml_parser.py

@@ -1,6 +1,7 @@
 """Code for parsing dependencies from environment.yml files."""
 
 import logging
+import re
 from pathlib import Path
 from typing import Any, Callable, Dict, Iterator, List, Union
 
@@ -13,6 +14,13 @@
 YamlDependencyData = Union[List[str], Dict[str, "YamlDependencyData"], Any, None]  # type: ignore[misc]
 
 
+class InvalidCondaRequirement(ValueError):  # noqa: N818
+    """Invalid Conda dependency specifier.
+
+    Modeled on packaging.requirements.InvalidRequirement.
+    """
+
+
 def parse_one_conda_dep(req_str: str, source: Location) -> DeclaredDependency:
     """Parse a single Conda dependency string.
 
@@ -24,9 +32,13 @@ def parse_one_conda_dep(req_str: str, source: Location) -> DeclaredDependency:
         _, req_str = req_str.split("::", 1)
     if "[" in req_str:  # remove bracket stuff from back
         req_str, _ = req_str.split("[]", 1)
-    req_str, *_ = req_str.split()  # remove anything after whitespace
-    name, *_ = req_str.split("=")  # remove version/build info
-    return DeclaredDependency(name, source)
+    # package name comes before version/build info, which starts with one of "><!=~ "
+    name_match = re.match(r"[^ =<>!~]+", req_str)
+    if not name_match:
+        raise InvalidCondaRequirement(
+            f"Expected package name at the start of dependency specifier: {req_str!r}"
+        )
+    return DeclaredDependency(name_match.group(0), source)
 
 
 def parse_environment_yml_deps(
@@ -47,7 +59,11 @@ def parse_environment_yml_deps(
         return
     for dep_item in parsed_deps:
         if isinstance(dep_item, str):
-            yield parse_item(dep_item, source)
+            try:
+                yield parse_item(dep_item, source)
+            except ValueError as e:
+                # InvalidCondaRequirement or packaging.requirements.InvalidRequirement
+                logger.error(f"{error_msg} {e}")
         elif isinstance(dep_item, dict) and len(dep_item) == 1 and "pip" in dep_item:
             pip_deps = dep_item.get("pip")
             yield from parse_environment_yml_deps(

tests/test_extract_deps_environment_yml.py

@@ -155,6 +155,70 @@
             ["scikit-learn", "pip"],
             id="mixed_conda_and_zero_pip_deps",
         ),
+        pytest.param(
+            """\
+            # To set up a development environment using conda, run:
+            #
+            #   conda env create -f environment.yml
+            #   conda activate cartopy-dev
+            #   pip install -e .
+            #
+            name: cartopy-dev
+            channels:
+              - conda-forge
+            dependencies:
+              - cython>=0.29.28
+              - numpy>=1.23
+              - shapely>=2.0
+              - pyshp>=2.3
+              - pyproj>=3.3.1
+              - packaging>=21
+              # The testing label has the proper version of freetype included
+              - conda-forge/label/testing::matplotlib-base>=3.6
+
+              # OWS
+              - owslib>=0.27
+              - pillow>=9.1
+              # Plotting
+              - scipy>=1.9
+              # Testing
+              - pytest
+              - pytest-mpl
+              - pytest-xdist
+              # Documentation
+              - pydata-sphinx-theme
+              - sphinx
+              - sphinx-gallery
+              # Extras
+              - pre-commit
+              - pykdtree
+              - ruff
+              - setuptools_scm
+            """,
+            [
+                "cython",
+                "numpy",
+                "shapely",
+                "pyshp",
+                "pyproj",
+                "packaging",
+                "matplotlib-base",
+                "owslib",
+                "pillow",
+                "scipy",
+                "pytest",
+                "pytest-mpl",
+                "pytest-xdist",
+                "pydata-sphinx-theme",
+                "sphinx",
+                "sphinx-gallery",
+                "pre-commit",
+                "pykdtree",
+                "ruff",
+                "setuptools_scm",
+            ],
+            id="cartopy_example",
+        ),
     ],
 )
 def test_parse_environment_yml__wellformed_dependencies__yields_dependencies(
@@ -239,6 +303,27 @@ class CondaTestVector:
         error_msg_fragment="Failed to parse Pip dependencies in {path}: Not a string: ",
         expect=["foo", "bar"],
     ),
+    CondaTestVector(
+        id="invalid_dependencies_malformed_names",
+        data="""\
+            dependencies:
+              - ">foo<"
+              - "bar"
+            """,
+        error_msg_fragment="Failed to parse Conda dependencies in {path}: Expected package name at the start of dependency specifier",
+        expect=["bar"],
+    ),
+    CondaTestVector(
+        id="invalid_dependencies_malformed_names",
+        data="""\
+            dependencies:
+              - "pip":
+                - "~foo"
+                - "bar"
+            """,
+        error_msg_fragment="Failed to parse Pip dependencies in {path}: Expected package name at the start of dependency specifier",
+        expect=["bar"],
+    ),
 ]