Better support for pytest.param in parametrize_plus

ci_tools/requirements-pip.txt

@@ -6,7 +6,6 @@ setuptools_scm
 # -- to install
 makefun>=1.7.0
 decopatch
-wrapt
 
 # --- to generate the reports (see scripts in ci_tools, called by .travis)
 pytest-html$PYTEST_HTML_VERSION

docs/changelog.md

@@ -1,5 +1,11 @@
 # Changelog
 
+### 1.15.0 - better `parametrize_plus` and smaller dependencies
+
+ - Better support for `pytest.param` in `parametrize_plus` and also in `fixture_union` and `fixture_param[s]`. Improved corresponding ids. Fixed [#79](https://github.com/smarie/python-pytest-cases/issues/79) and [#86](https://github.com/smarie/python-pytest-cases/issues/86)
+
+ - Removed `six`, `wrapt` and `enum34` dependencies
+
 ### 1.14.0 - bugfixes and hook feature
 
  - Fixed `ids` precedence order when using `pytest.mark.parametrize` in a `fixture_plus`. Fixed [#87](https://github.com/smarie/python-pytest-cases/issues/87)

pytest_cases/common.py

@@ -4,14 +4,14 @@
     from funcsigs import signature
 
 try:
-    from typing import Union, Callable, Any
+    from typing import Union, Callable, Any, Optional
 except ImportError:
     pass
 
 from distutils.version import LooseVersion
 from warnings import warn
 
-from six import string_types
+from .mini_six import string_types
 import pytest
 
 
@@ -274,38 +274,90 @@ def combine_ids(paramid_tuples):
     return ['-'.join(pid for pid in testid) for testid in paramid_tuples]
 
 
-def get_test_ids_from_param_values(param_names,
-                                   param_values,
-                                   ):
+def make_test_ids(global_ids, id_marks, argnames=None, argvalues=None, precomputed_ids=None):
     """
-    Replicates pytest behaviour to generate the ids when there are several parameters in a single `parametrize`
+    Creates the proper id for each test based on (higher precedence first)
+
+     - any specific id mark from a `pytest.param` (`id_marks`)
+     - the global `ids` argument of pytest parametrize (`global_ids`)
+     - the name and value of parameters (`argnames`, `argvalues`) or the precomputed ids(`precomputed_ids`)
+
+    See also _pytest.python._idvalset method
+
+    :param global_ids:
+    :param id_marks:
+    :param argnames:
+    :param argvalues:
+    :param precomputed_ids:
+    :return:
+    """
+    if global_ids is not None:
+        # overridden at global pytest.mark.parametrize level - this takes precedence.
+        try:  # an explicit list of ids ?
+            p_ids = list(global_ids)
+        except TypeError:  # a callable to apply on the values
+            p_ids = list(global_ids(v) for v in argvalues)
+    else:
+        # default: values-based
+        if precomputed_ids is not None:
+            if argnames is not None or argvalues is not None:
+                raise ValueError("Only one of `precomputed_ids` or argnames/argvalues should be provided.")
+            p_ids = precomputed_ids
+        else:
+            p_ids = make_test_ids_from_param_values(argnames, argvalues)
+
+    # Finally, local pytest.param takes precedence over everything else
+    for i, _id in enumerate(id_marks):
+        if _id is not None:
+            p_ids[i] = _id
+    return p_ids
+
+
+def make_test_ids_from_param_values(param_names,
+                                    param_values,
+                                    ):
+    """
+    Replicates pytest behaviour to generate the ids when there are several parameters in a single `parametrize.
+    Note that param_values should not contain marks.
 
     :param param_names:
     :param param_values:
     :return: a list of param ids
     """
+    if isinstance(param_names, string_types):
+        raise TypeError("param_names must be an iterable. Found %r" % param_names)
+
     nb_params = len(param_names)
     if nb_params == 0:
         raise ValueError("empty list provided")
     elif nb_params == 1:
-        paramids = list(str(v) for v in param_values)
+        paramids = []
+        for _idx, v in enumerate(param_values):
+            _id = mini_idvalset(param_names, (v,), _idx)
+            paramids.append(_id)
     else:
         paramids = []
-        for vv in param_values:
+        for _idx, vv in enumerate(param_values):
             if len(vv) != nb_params:
                 raise ValueError("Inconsistent lenghts for parameter names and values: '%s' and '%s'"
                                  "" % (param_names, vv))
-            paramids.append('-'.join([str(v) for v in vv]))
+            _id = mini_idvalset(param_names, vv, _idx)
+            paramids.append(_id)
     return paramids
 
 
 # ---- ParameterSet api ---
-def analyze_parameter_set(pmark=None, argnames=None, argvalues=None, ids=None):
+def analyze_parameter_set(pmark=None, argnames=None, argvalues=None, ids=None, check_nb=True):
     """
     analyzes a parameter set passed either as a pmark or as distinct
     (argnames, argvalues, ids) to extract/construct the various ids, marks, and
     values
 
+    See also pytest.Metafunc.parametrize method, that calls in particular
+    pytest.ParameterSet._for_parametrize and _pytest.python._idvalset
+
+    :param check_nb: a bool indicating if we should raise an error if len(argnames) > 1 and any argvalue has
+         a different length than len(argnames)
     :return: ids, marks, values
     """
     if pmark is not None:
@@ -316,38 +368,31 @@ def analyze_parameter_set(pmark=None, argnames=None, argvalues=None, ids=None):
         ids = pmark.param_ids
 
     # extract all parameters that have a specific configuration (pytest.param())
-    custom_pids, p_marks, p_values = extract_parameterset_info(argnames, argvalues)
-
-    # Create the proper id for each test
-    if ids is not None:
-        # overridden at global pytest.mark.parametrize level - this takes precedence.
-        try:  # an explicit list of ids ?
-            p_ids = list(ids)
-        except TypeError:  # a callable to apply on the values
-            p_ids = list(ids(v) for v in p_values)
-    else:
-        # default: values-based
-        p_ids = get_test_ids_from_param_values(argnames, p_values)
+    custom_pids, p_marks, p_values = extract_parameterset_info(argnames, argvalues, check_nb=check_nb)
 
-    # Finally, local pytest.param takes precedence over everything else
-    for i, _id in enumerate(custom_pids):
-        if _id is not None:
-            p_ids[i] = _id
+    # get the ids by merging/creating the various possibilities
+    p_ids = make_test_ids(argnames=argnames, argvalues=p_values, global_ids=ids, id_marks=custom_pids)
 
     return p_ids, p_marks, p_values
 
 
-def extract_parameterset_info(argnames, argvalues):
+def extract_parameterset_info(argnames, argvalues, check_nb=True):
     """
 
     :param argnames: the names in this parameterset
     :param argvalues: the values in this parameterset
+    :param check_nb: a bool indicating if we should raise an error if len(argnames) > 1 and any argvalue has
+         a different length than len(argnames)
     :return:
     """
     pids = []
     pmarks = []
     pvalues = []
+    if isinstance(argnames, string_types):
+        raise TypeError("argnames must be an iterable. Found %r" % argnames)
+    nbnames = len(argnames)
     for v in argvalues:
+        # is this a pytest.param() ?
         if is_marked_parameter_value(v):
             # --id
             id = get_marked_parameter_id(v)
@@ -356,19 +401,21 @@ def extract_parameterset_info(argnames, argvalues):
             marks = get_marked_parameter_marks(v)
             pmarks.append(marks)  # note: there might be several
             # --value(a tuple if this is a tuple parameter)
-            vals = get_marked_parameter_values(v)
-            if len(vals) != len(argnames):
-                raise ValueError("Internal error - unsupported pytest parametrization+mark combination. Please "
-                                 "report this issue")
-            if len(vals) == 1:
-                pvalues.append(vals[0])
+            v = get_marked_parameter_values(v)
+            if nbnames == 1:
+                pvalues.append(v[0])
             else:
-                pvalues.append(vals)
+                pvalues.append(v)
         else:
+            # normal argvalue
             pids.append(None)
             pmarks.append(None)
             pvalues.append(v)
 
+        if check_nb and nbnames > 1 and (len(v) != nbnames):
+            raise ValueError("Inconsistent number of values in pytest parametrize: %s items found while the "
+                             "number of parameters is %s: %s." % (len(v), nbnames, v))
+
     return pids, pmarks, pvalues
 
 
@@ -490,3 +537,43 @@ def get_pytest_scopenum(scope_str):
 
 def get_pytest_function_scopenum():
     return pt_scopes.index("function")
+
+
+from _pytest.python import _idval
+
+
+if LooseVersion(pytest.__version__) >= LooseVersion('3.0.0'):
+    _idval_kwargs = dict(idfn=None,
+                         item=None,  # item is only used by idfn
+                         config=None  # if a config hook was available it would be used before this is called)
+                         )
+else:
+    _idval_kwargs = dict(idfn=None,
+                         # item=None,  # item is only used by idfn
+                         # config=None  # if a config hook was available it would be used before this is called)
+                         )
+
+
+def mini_idval(
+        val,      # type: object
+        argname,  # type: str
+        idx,      # type: int
+    ):
+    """
+    A simplified version of idval where idfn, item and config do not need to be passed.
+
+    :param val:
+    :param argname:
+    :param idx:
+    :return:
+    """
+    return _idval(val=val, argname=argname, idx=idx, **_idval_kwargs)
+
+
+def mini_idvalset(argnames, argvalues, idx):
+    """ mimic _pytest.python._idvalset """
+    this_id = [
+        _idval(val, argname, idx=idx,**_idval_kwargs)
+        for val, argname in zip(argvalues, argnames)
+    ]
+    return "-".join(this_id)