Blitz static get query string

[will-moore]

What this PR does

See discussion at #4950 (comment)
Since BlitzGateway wrapper_getQueryString() doesn't need to be an instance method, this is now a classmethod.

Testing this PR

Check that tests are green etc.

cc @aleksandra-tarkowska

[will-moore]

Probably this PR is causing this error:

 File "/opt/hudson/workspace/WEB-DEV-integration-deploy/OMERO.py/lib/python/omero/gateway/__init__.py", line 774, in listParents
   for x in t])
TypeError: <lambda>() takes exactly 1 argument (2 given)

[chris-allan]

While I appreciate the work being done here what exactly are we intending to achieve structurally by changing this to a class method? Are we really so sure that we're never going to need anything from the instance to generate the query?

Edit: Moving things out of the bootstrap method also has some pretty serious downstream implications.

[will-moore]

I think that _getQueryString() has to be callable without an instance of the object, since we need the query in order to retrieve any objects from the server. If you needed an instance to get the query, then we'd have a chicken and egg problem.
I think the main benefit from these changes is that we wouldn't have to instantiate a dummy object just to get the query. Currently we're doing:

wrapper()._getQueryString(opts)

This PR would allow us to get the query without creating a wrapper instance (in most cases) see ab9eba9

[chris-allan]

A ha, makes sense. Also then makes complete sense to refactor away the bootstrap method. Do we have any cases where we are using it at all after the changes you've made in 1ae9fff? We also need to know a priori what the wrapper class type mapping is anyway correct? That is, wrapper() is providing us with no additional convenience or abstraction?

[will-moore]

The error above in #4993 (comment) is caused by removal of __bstrap__() in WellSampleWrapper and can be fixed by adding back the __bstrap__ like this:

-    LINK_PARENT = lambda x: x
     LINK_CHILD = 'image'
 
+    def __bstrap__(self):
+        self.LINK_PARENT = lambda x: x

Or by providing E.g. a static method:

-    LINK_PARENT = lambda x: x
     LINK_CHILD = 'image'
+    
+    @staticmethod
+    def LINK_PARENT(parent):
+        return parent

I don't know if there's any reason why self.LINK_PARENT = lambda x: x should be better than this alternative?

The only other remaining use of __bstrap__() is in LaserWrapper

    def __bstrap__(self):
        super(_LaserWrapper, self).__bstrap__()
        self._attrs += (
            '#laserMedium',
            'frequencyMultiplication',
            'tuneable'....

[chris-allan]

There's already a LINK_PARENT static method on BlitzObjectWrapper. It's not 100% clear to me exactly what the intention was overriding it this manner in the WellSampleWrapper. Certainly overriding the static method with another static method would work. We should probably put test cases in, if we don't have them already, to cover this functionality before messing around with it however; listParents() is super old code.

[will-moore]

Yes, I'm already writing more tests for listParents(), since we only test it for PDI and it should work for a lot of others.

[will-moore]

@chris-allan The BlitzObjectWrapper.LINK_PARENT(link) returns link.parent since that is how most objects are linked but there's no link between WellSample and Well, so WellSample.listParents() returns Wells rather than links.

[will-moore]

Robot tests that were failing yesterday
https://ci.openmicroscopy.org/view/Failing/job/OMERO-DEV-merge-robotframework/468/robot/
due to
#4993 (comment)
are now passing
https://ci.openmicroscopy.org/view/Failing/job/OMERO-DEV-merge-robotframework/469/robot/

[will-moore]

testListParents is passing at https://ci.openmicroscopy.org/view/Failing/job/OMERO-DEV-breaking-integration-Python27/127/testReport/OmeroPy.test.integration.gatewaytest.test_get_objects/TestListParents/

[atarkowska]

I don't understand arg klass, http://legacy.python.org/dev/peps/pep-0008/#function-and-method-arguments

[joshmoore]

This is probably a historical (i.e. pre-PEP8) Josh/Carlos-style thing which should be dropped in favor of cls.

[chris-allan]

The first argument to a class method is a reference to the class. As class is a "reserved" word in Python so we have to use something. klass is a classical style that we inherited from Java. cls is also fine and in line with the PEP8 suggestions as @joshmoore has already mentioned.

[jburel]

Reviewed with @chris-allan, merging.