Coverage for python/lsst/pex/config/wrap.py : 55%

Hot-keys on this page
r m x p toggle line displays
j k next/prev highlighted chunk
0 (zero) top of page
1 (one) first highlighted chunk
# # LSST Data Management System # Copyright 2008, 2009, 2010 LSST Corporation. # # This product includes software developed by the # LSST Project (http://www.lsst.org/). # # This program is free software: you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation, either version 3 of the License, or # (at your option) any later version. # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # You should have received a copy of the LSST License Statement and # the GNU General Public License along with this program. If not, # see <http://www.lsstcorp.org/LegalNotices/>. #
# Mapping from C++ types to Python type: assumes we can round-trip between these using # the usual pybind11 converters, but doesn't require they be binary equivalent under-the-hood # or anything. "bool": bool, "int": int, "double": float, "float": float, "std::int64_t": int, "std::string": str }
"""A function that creates a Python config class that matches a C++ control object class.
@param ctrl C++ control class to wrap. @param name Name of the new config class; defaults to the __name__ of the control class with 'Control' replaced with 'Config'. @param base Base class for the config class. @param doc Docstring for the config class. @param module Either a module object, a string specifying the name of the module, or an integer specifying how far back in the stack to look for the module to use: 0 is the immediate caller of pex.config.wrap. This will be used to set __module__ for the new config class, and the class will also be added to the module. Ignored if None or if cls is not None, but note that the default is to use the callers' module. @param cls An existing config class to use instead of creating a new one; name, base doc, and module will be ignored if this is not None.
See the 'wrap' decorator as a way to use makeConfigClass that may be more convenient.
To use makeConfigClass, in C++, write a control object, using the LSST_CONTROL_FIELD macro in lsst/pex/config.h (note that it must have sensible default constructor):
@code // myHeader.h
struct InnerControl { LSST_CONTROL_FIELD(wim, std::string, "documentation for field 'wim'"); };
struct FooControl { LSST_CONTROL_FIELD(bar, int, "documentation for field 'bar'"); LSST_CONTROL_FIELD(baz, double, "documentation for field 'baz'"); LSST_NESTED_CONTROL_FIELD(zot, myWrappedLib, InnerControl, "documentation for field 'zot'");
FooControl() : bar(0), baz(0.0) {} }; @endcode
You can use LSST_NESTED_CONTROL_FIELD to nest control objects. Now, wrap those control objects as you would any other C++ class, but make sure you include lsst/pex/config.h before including the header file where the control object class is defined:
Now, in Python, do this:
@code import myWrappedLib import lsst.pex.config InnerConfig = lsst.pex.config.makeConfigClass(myWrappedLib.InnerControl) FooConfig = lsst.pex.config.makeConfigClass(myWrappedLib.FooControl) @endcode
This will add fully-fledged "bar", "baz", and "zot" fields to FooConfig, set FooConfig.Control = FooControl, and inject makeControl and readControl methods to create a FooControl and set the FooConfig from the FooControl, respectively. In addition, if FooControl has a validate() member function, a custom validate() method will be added to FooConfig that uses it. And, of course, all of the above will be done for InnerControl/InnerConfig too.
Any field that would be injected that would clash with an existing attribute of the class will be silently ignored; this allows the user to customize fields and inherit them from wrapped control classes. However, these names will still be processed when converting between config and control classes, so they should generally be present as base class fields or other instance attributes or descriptors.
While LSST_CONTROL_FIELD will work for any C++ type, automatic Config generation only supports bool, int, std::int64_t, double, and std::string fields, along with std::list and std::vectors of those types. """ raise ValueError("Cannot guess appropriate Config class name for %s." % ctrl) # Not only does setting __module__ make Python pretty-printers more useful, # it's also necessary if we want to pickle Config objects. elif isinstance(module, str): moduleName = module moduleObj = __import__(moduleName) else: moduleObj = module moduleName = moduleObj.__name__ # loop over all class attributes, looking for the special static methods that indicate a field # defined by one of the macros in pex/config.h. else: nestedModuleObj = importlib.import_module(nestedModuleName) except AttributeError: raise AttributeError("'%s.%s.ConfigClass' does not exist" % (moduleName, ctype)) else: raise TypeError("Could not parse field type '%s'." % ctype)
# Define a number of methods to put in the new Config class. Note that these are "closures"; # they have access to local variables defined in the makeConfigClass function (like the fields dict). """Construct a C++ Control object from this Config object.
Fields set to None will be ignored, and left at the values defined by the Control object's default constructor. """ r = self.Control() for k, f in fields.items(): value = getattr(self, k) if isinstance(f, ConfigField): value = value.makeControl() if value is not None: if isinstance(value, List): setattr(r, k, value._list) else: setattr(r, k, value) return r
"""Read values from a C++ Control object and assign them to self's fields.
The __at, __label, and __reset arguments are for internal use only; they are used to remove internal calls from the history. """ if __at is None: __at = getCallStack() values = {} for k, f in fields.items(): if isinstance(f, ConfigField): getattr(self, k).readControl(getattr(control, k), __at=__at, __label=__label, __reset=__reset) else: values[k] = getattr(control, k) if __reset: self._history = {} self.update(__at=__at, __label=__label, **values)
"""Validate the config object by constructing a control object and using a C++ validate() implementation.""" super(cls, self).validate() r = self.makeControl() r.validate()
"""Initialize the config object, using the Control objects default ctor to provide defaults.""" super(cls, self).setDefaults() try: r = self.Control() # Indicate in the history that these values came from C++, even if we can't say which line self.readControl(r, __at=[(ctrl.__name__ + " C++", 0, "setDefaults", "")], __label="defaults", __reset=True) except Exception: pass # if we can't instantiate the Control, don't set defaults
cls.validate = validate
"""A decorator that adds fields from a C++ control class to a Python config class.
Used like this:
@wrap(MyControlClass) class MyConfigClass(Config): pass
See makeConfigClass for more information; this is equivalent to calling makeConfigClass with the decorated class as the 'cls' argument. """ def decorate(cls): return makeConfigClass(ctrl, cls=cls) return decorate |