Lichen

resolving.py

817:d39dbf8b26d2
2017-06-17 Paul Boddie Test mixing of \u and \x values in unprefixed string literals.
     1 #!/usr/bin/env python     2      3 """     4 Name resolution.     5      6 Copyright (C) 2016, 2017 Paul Boddie <paul@boddie.org.uk>     7      8 This program is free software; you can redistribute it and/or modify it under     9 the terms of the GNU General Public License as published by the Free Software    10 Foundation; either version 3 of the License, or (at your option) any later    11 version.    12     13 This program is distributed in the hope that it will be useful, but WITHOUT    14 ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS    15 FOR A PARTICULAR PURPOSE.  See the GNU General Public License for more    16 details.    17     18 You should have received a copy of the GNU General Public License along with    19 this program.  If not, see <http://www.gnu.org/licenses/>.    20 """    21     22 from common import init_item    23 from results import AccessRef, InstanceRef, InvocationRef, LocalNameRef, \    24                     MultipleRef, NameRef, ResolvedNameRef    25 from referencing import Reference    26 import sys    27     28 class NameResolving:    29     30     "Resolving names mix-in for inspected modules."    31     32     # Post-inspection resolution activities.    33     34     def resolve(self):    35     36         "Resolve dependencies and complete definitions."    37     38         self.resolve_class_bases()    39         self.check_special()    40         self.check_names_used()    41         self.check_invocations()    42         self.resolve_initialisers()    43         self.resolve_return_values()    44         self.resolve_literals()    45     46     def resolve_class_bases(self):    47     48         "Resolve all class bases since some of them may have been deferred."    49     50         for name, bases in self.classes.items():    51             resolved = []    52             bad = []    53     54             for base in bases:    55                 ref = self.importer.identify(base.get_origin())    56     57                 # Obtain the origin of the base class reference.    58     59                 if not ref or not ref.has_kind("<class>"):    60                     bad.append(base)    61                     break    62     63                 resolved.append(ref)    64     65             if bad:    66                 print >>sys.stderr, "Bases of class %s were not classes: %s" % (name, ", ".join(map(str, bad)))    67             else:    68                 self.importer.classes[name] = self.classes[name] = resolved    69     70     def check_special(self):    71     72         "Check special names."    73     74         for name, value in self.special.items():    75             ref, paths = value    76             self.special[name] = self.importer.identify(ref.get_origin()), paths    77     78     def check_names_used(self):    79     80         "Check the external names used by each scope."    81     82         for key, ref in self.name_references.items():    83             path, name = key.rsplit(".", 1)    84             self.resolve_accesses(path, name, ref)    85     86     def check_invocations(self):    87     88         "Find invocations amongst module data and replace their results."    89     90         # Find members and replace invocation results with values. This is    91         # effectively the same as is done for initialised names, but refers to    92         # any unchanging value after initialisation.    93     94         for key, ref in self.objects.items():    95             if ref.has_kind("<invoke>"):    96                 ref = self.convert_invocation(ref)    97                 self.importer.objects[key] = self.objects[key] = ref    98     99         # Convert name references.   100    101         for key, ref in self.name_references.items():   102             if ref.has_kind("<invoke>"):   103                 ref = self.convert_invocation(ref)   104                 self.importer.all_name_references[key] = self.name_references[key] = ref   105    106         # Convert function defaults, which are effectively extra members of the   107         # module, and function locals.   108    109         for fname, parameters in self.function_defaults.items():   110             l = []   111             for pname, ref in parameters:   112                 if ref.has_kind("<invoke>"):   113                     ref = self.convert_invocation(ref)   114                 l.append((pname, ref))   115             self.importer.function_defaults[fname] = self.function_defaults[fname] = l   116    117         # Convert function locals referencing invocations.   118    119         for fname, names in self.function_locals.items():   120             for name, ref in names.items():   121                 if ref.has_kind("<invoke>"):   122                     ref = self.convert_invocation(ref)   123                     names[name] = ref   124    125     def convert_invocation(self, ref):   126    127         "Convert the given invocation 'ref', handling instantiation."   128    129         alias = ref.get_name()   130         ref = self.importer.identify(ref.get_origin())   131         ref = ref and ref.has_kind("<class>") and ref.instance_of() or Reference("<var>")   132         return ref and ref.alias(alias) or None   133    134     def resolve_accesses(self, path, name, ref):   135    136         """   137         Resolve any unresolved accesses in the function at the given 'path'   138         for the given 'name' corresponding to the indicated 'ref'. Note that   139         this mechanism cannot resolve things like inherited methods because   140         they are not recorded as program objects in their inherited locations.   141         """   142    143         attr_accesses = self.global_attr_accesses.get(path)   144         all_attrnames = attr_accesses and attr_accesses.get(name)   145    146         if not all_attrnames:   147             return   148    149         # Insist on constant accessors.   150    151         if not ref.has_kind(["<class>", "<module>"]):   152             return   153    154         found_attrnames = set()   155    156         for attrnames in all_attrnames:   157    158             # Start with the resolved name, adding attributes.   159    160             attrs = ref.get_path()   161             remaining = attrnames.split(".")   162             last_ref = ref   163    164             # Add each component, testing for a constant object.   165    166             while remaining:   167                 attrname = remaining[0]   168                 attrs.append(attrname)   169                 del remaining[0]   170    171                 # Find any constant object reference.   172    173                 attr_ref = self.get_resolved_object(".".join(attrs))   174    175                 # Non-constant accessors terminate the traversal.   176    177                 if not attr_ref or not attr_ref.has_kind(["<class>", "<module>", "<function>"]):   178    179                     # Provide the furthest constant accessor unless the final   180                     # access can be resolved.   181    182                     if remaining:   183                         remaining.insert(0, attrs.pop())   184                     else:   185                         last_ref = attr_ref   186                     break   187    188                 # Follow any reference to a constant object.   189                 # Where the given name refers to an object in another location,   190                 # switch to the other location in order to be able to test its   191                 # attributes.   192    193                 last_ref = attr_ref   194                 attrs = attr_ref.get_path()   195    196             # Record a constant accessor only if an object was found   197             # that is different from the namespace involved.   198    199             if last_ref:   200                 objpath = ".".join(attrs)   201                 if objpath != path:   202    203                     if last_ref.has_kind("<invoke>"):   204                         last_ref = self.convert_invocation(last_ref)   205    206                     # Establish a constant access.   207    208                     init_item(self.const_accesses, path, dict)   209                     self.const_accesses[path][(name, attrnames)] = (objpath, last_ref, ".".join(remaining))   210    211                     if len(attrs) > 1:   212                         found_attrnames.add(attrs[1])   213    214         # Remove any usage records for the name.   215    216         if found_attrnames:   217    218             # NOTE: Should be only one name version.   219    220             versions = []   221             for version in self.attr_usage[path][name]:   222                 new_usage = set()   223                 for usage in version:   224                     if found_attrnames.intersection(usage):   225                         new_usage.add(tuple(set(usage).difference(found_attrnames)))   226                     else:   227                         new_usage.add(usage)   228                 versions.append(new_usage)   229    230             self.attr_usage[path][name] = versions   231    232     def resolve_initialisers(self):   233    234         "Resolve initialiser values for names."   235    236         # Get the initialisers in each namespace.   237    238         for path, name_initialisers in self.name_initialisers.items():   239    240             # Resolve values for each name in a scope.   241    242             for name, values in name_initialisers.items():   243                 self._resolve_values(path, name, values)   244    245     def resolve_return_values(self):   246    247         "Resolve return values using name references."   248    249         # Get the return values from each namespace.   250    251         for path, values in self.return_values.items():   252    253             # Resolve each return value provided by the scope.   254    255             self._resolve_values(path, "$return", values)   256    257     def _resolve_values(self, path, name, values):   258    259         """   260         Resolve in 'path' the references for 'name' involving the given   261         'values'.   262         """   263    264         initialised_names = {}   265         aliased_names = {}   266    267         for i, name_ref in enumerate(values):   268             initialised_ref, _aliased_names = self.resolve_reference(path, name_ref)   269             if initialised_ref:   270                 initialised_names[i] = initialised_ref   271             if _aliased_names:   272                 aliased_names[i] = _aliased_names   273    274         if initialised_names:   275             self.initialised_names[(path, name)] = initialised_names   276         if aliased_names:   277             self.aliased_names[(path, name)] = aliased_names   278    279     def resolve_reference(self, path, name_ref):   280    281         """   282         Within the namespace 'path', resolve the given 'name_ref', returning any   283         initialised reference, along with any aliased name information.   284         """   285    286         initialised_ref = None   287         aliased_names = None   288         no_reference = None, None   289    290         # Attempt to obtain a coherent reference from multiple outcomes.   291    292         if isinstance(name_ref, MultipleRef):   293             refs = set()   294             aliases = []   295    296             for result in name_ref.results:   297                 _initialised_ref, _aliased_names = self.resolve_reference(path, result)   298    299                 # Unsuitable references at any level cause the result to yield   300                 # no reference.   301    302                 if not _initialised_ref:   303                     refs = None   304                 elif refs is not None:   305                     refs.add(_initialised_ref)   306    307                 if not _aliased_names:   308                     aliases = None   309                 elif aliases is not None:   310                     aliases += _aliased_names   311    312             # Only unambiguous references are returned as initialising   313             # references.   314    315             if refs and len(refs) == 1:   316                 return list(refs)[0], aliases   317             else:   318                 return None, aliases   319    320         # Unwrap invocations.   321    322         if isinstance(name_ref, InvocationRef):   323             invocation = True   324             name_ref = name_ref.name_ref   325         else:   326             invocation = False   327    328         const_accesses = self.const_accesses.get(path)   329    330         # Obtain a usable reference from names or constants.   331    332         if isinstance(name_ref, ResolvedNameRef):   333             if not name_ref.reference():   334                 return no_reference   335             ref = name_ref.reference()   336    337         # Obtain a reference from instances.   338    339         elif isinstance(name_ref, InstanceRef):   340             if not name_ref.reference():   341                 return no_reference   342             ref = name_ref.reference()   343    344         # Resolve accesses that employ constants.   345    346         elif isinstance(name_ref, AccessRef):   347             ref = None   348    349             if const_accesses:   350                 resolved_access = const_accesses.get((name_ref.original_name, name_ref.attrnames))   351                 if resolved_access:   352                     objpath, ref, remaining_attrnames = resolved_access   353                     if remaining_attrnames:   354                         ref = None   355    356             # Accesses that do not employ constants cannot be resolved,   357             # but they may be resolvable later.   358    359             if not ref:   360    361                 # Record the path used for tracking purposes   362                 # alongside original name, attribute and access   363                 # number details.   364    365                 aliased_names = [(path, name_ref.original_name, name_ref.attrnames, name_ref.number)]   366    367                 return None, aliased_names   368    369         # Attempt to resolve a plain name reference.   370    371         elif isinstance(name_ref, LocalNameRef):   372             key = "%s.%s" % (path, name_ref.name)   373             ref = self.name_references.get(key)   374    375             # Accesses that do not refer to known static objects   376             # cannot be resolved, but they may be resolvable later.   377    378             if not ref:   379    380                 # Record the path used for tracking purposes   381                 # alongside original name, attribute and access   382                 # number details.   383    384                 aliased_names = [(path, name_ref.name, None, name_ref.number)]   385    386                 return None, aliased_names   387    388             ref = self.get_resolved_object(ref.get_origin())   389             if not ref:   390                 return no_reference   391    392         elif isinstance(name_ref, NameRef):   393             key = "%s.%s" % (path, name_ref.name)   394             ref = self.name_references.get(key)   395    396             ref = ref and self.get_resolved_object(ref.get_origin())   397             if not ref:   398                 return no_reference   399    400         else:   401             return no_reference   402    403         # Resolve any hidden dependencies involving external objects   404         # or unresolved names referring to globals or built-ins.   405    406         if ref.has_kind("<depends>"):   407             ref = self.importer.identify(ref.get_origin())   408    409         # Convert class invocations to instances.   410    411         if ref and (invocation or ref.has_kind("<invoke>")):   412             target_ref = ref   413             ref = self.convert_invocation(target_ref)   414    415             if not ref or ref.has_kind("<var>"):   416                 aliased_names = self.get_aliases_for_target(target_ref.get_origin())   417             else:   418                 initialised_ref = ref   419    420         elif ref and not ref.has_kind("<var>"):   421             initialised_ref = ref   422            423         return initialised_ref, aliased_names   424    425     def get_aliases_for_target(self, path):   426    427         "Return a list of return value locations for the given 'path'."   428    429         return_values = self.importer.all_return_values.get(path)   430         locations = []   431    432         if return_values:   433             for version in range(0, len(return_values)):   434                 locations.append((path, "$return", None, version))   435    436         return locations   437    438     def resolve_literals(self):   439    440         "Resolve constant value types."   441    442         # Get the constants defined in each namespace.   443    444         for path, constants in self.constants.items():   445             for constant, n in constants.items():   446                 objpath = "%s.$c%d" % (path, n)   447                 _constant, value_type, encoding = self.constant_values[objpath]   448                 self.initialised_names[(path, objpath)] = {0 : Reference("<instance>", value_type)}   449    450         # Get the literals defined in each namespace.   451    452         for path, literals in self.literals.items():   453             for n in range(0, literals):   454                 objpath = "%s.$C%d" % (path, n)   455                 value_type = self.literal_types[objpath]   456                 self.initialised_names[(path, objpath)] = {0 : Reference("<instance>", value_type)}   457    458     # Object resolution.   459    460     def get_resolved_object(self, path, defer=False):   461    462         """   463         Get the details of an object with the given 'path' within this module.   464         Where the object has not been resolved, None is returned. This differs   465         from the get_object method used elsewhere in that it does not return an   466         unresolved object reference.   467         """   468    469         if self.objects.has_key(path):   470             ref = self.objects[path]   471             if not defer and ref.has_kind("<depends>"):   472                 return None   473             else:   474                 return ref   475         else:   476             return None   477    478 # vim: tabstop=4 expandtab shiftwidth=4