Question: Can the parametrization separator be changed?

[liiight]

Is it possible to change the parameterization separator from dash to another char? Couldn't find the answer, I apologize if it exist.

On a side note, I've been a pytest user both professionally and in open source capacity for years now, and I absolutely love it. You guys created an icon in the software development world.

[pytestbot]

GitMate.io thinks possibly related issues are #3273 (Question: How can I parameterize my tests with global variable which changes values on runtime.), #3601 ((question) Cleaning up from parameter generation), #249 (document and/or change display of parametrization values in IDs), #531 (Problems with fixture parametrization), and #3320 ([question] Multithreaded support in a function).

[RonnyPfannschmidt]

currently the dash is hardcoded, can you outline your use-case?

[liiight]

Test names are the source of several reports I use, trying to make it a little more readable. Hardly a major issue.

I could try to do this myself if you guys think there's merit to it, been wanting to contribute to pytest directly.

Maybe pytest_parameterization_seperator hook?

[RonnyPfannschmidt]

at first glance this reads like a feature that doesnt add the value that justifies the extra complexity needed

[liiight]

Well, not having a clue about the added complexity I cannot comment on that, but I believe that any ability to customize the visibility of the results can be very useful.

If you do decide to deem this issue as worthy, I'd be happy to take a stab at it.

[DanaGoyette]

I'd also like to be able to change the separator. In my framework, I'm having to monkeypatch pytest, so I can get commas instead of dashes.

Example of a test name before:

test_snmp_get[2c-master-master]
test_snmp_get[2c-subagent-slave]

Add in a pytest_make_parametrize_id that returns '<argname>=<val>':

test_snmp_get[version=2c-mode=master-zone=master]
test_snmp_get[version=2c-mode=subagent-zone=slave]

To me, that name looks like version = 2c-mode = subagent-zone = slave.
I'd say dashes/hyphens are meant for joining things, not for separating them, and now I'm reminded of an XKCD comic ("hyphen").

After monkey-patching pytest to use commas:

test_snmp_get[2c,master,master]
test_snmp_get[2c,subagent,slave]

Re-add the pytest_make_parametrize_id:

test_snmp_get[version=2c,mode=master,zone=master]
test_snmp_get[version=2c,mode=subagent,zone=slave]

Example monkey-patch, on pytest 3.6.1:

    def id(self):
        """
        Original version joined parameters with dashes, which is weird.
        Given arguments arg1=a, arg2=x:
        Before: "test_name[a-x]"
        After: "test_name[a,x]"
        """
        return ','.join(map(str, filter(None, self._idlist)))

    # Replace the property function itself.
    setattr(_pytest.python.CallSpec2, 'id', property(id))

    def _idvalset(idx, parameterset, argnames, idfn, ids, config=None):
        from _pytest.compat import ascii_escaped
        from _pytest.python import _idval
        if parameterset.id is not None:
            return parameterset.id
        if ids is None or (idx >= len(ids) or ids[idx] is None):
            this_id = [_idval(val, argname, idx, idfn, config)
                       for val, argname in zip(parameterset.values, argnames)]
            return ",".join(this_id)
        else:
            return ascii_escaped(ids[idx])

    _pytest.python._idvalset = _idvalset

[nicoddemus]

I share @RonnyPfannschmidt's worry about increasing pytest's code complexity, but seems CallSpec and _idvalset have access to the config object so it might not get so complex at the end.

We have a pytest_make_parametrize_id hook, perhaps @liiight suggestion of having another hook for the separator would be a good approach.

An ini option has two problems:

• this seems rare enough to be a another option to users (in an ever expanding list)
• this is not something you want users to fiddle with in the command-line (with -o) given that this changes test ids and this affects other plugins (like cache with --lf options).

[DanaGoyette]

I also realize that I don't remember why I monkey-patched both of those functions. I'd imagine the two are used in different places, but I don't recall precisely where for each.

[RonnyPfannschmidt]

since iths is about reporting why do we think about making the internals more complex instead allowing more information for reporting - neither reporting nor the internals are well served by a hook to make the ids dynamic in that particular way