Handle multiple iterations

Author: pfultz2

addons/cppcheckdata.py

@@ -1386,6 +1386,11 @@ class CppcheckData:
         suppressions      List of Suppressions
         files             Source files for elements occurred in this configuration
 
+    The configurations property loads all configurations once and keeps them,
+    which is best when the configurations are iterated more than once.
+    iterconfigurations() parses the dump file on each call but creates one
+    configuration at a time, which uses less memory for a single iteration.
+
     To iterate through all configurations use such code:
     @code
     data = cppcheckdata.parsedump(...)
@@ -1424,6 +1429,7 @@ def __init__(self, filename):
         self.platform = None
         self.suppressions = []
         self.files = [] # source files for elements occurred in this configuration
+        self._configurations = None # set by the configurations property
 
         # Parse general configuration options from <dumps> node, the raw tokens
         # and the suppressions. The configurations are parsed by iterconfigurations.
@@ -1444,14 +1450,26 @@ def __init__(self, filename):
     def configurations(self):
         """
         Return the list of all available Configuration objects.
+        The configurations are loaded on the first access and kept, so later
+        accesses and iterconfigurations() do not parse the dump file again.
+        All configurations stay in memory as long as this object exists.
         """
-        return list(self.iterconfigurations())
+        if self._configurations is None:
+            self._configurations = list(self.iterconfigurations())
+        return self._configurations
 
     def iterconfigurations(self):
         """
         Create and return iterator for the available Configuration objects.
         The iterator loops over all Configurations in the dump file tree, in document order.
+        If the configurations property has been used then the kept configurations
+        are returned. Otherwise the dump file is parsed and each configuration is
+        created when the iterator reaches it, so a configuration can be freed when
+        the caller is done with it. Each call parses the dump file again.
         """
+        if self._configurations is not None:
+            yield from self._configurations
+            return
         # Parse the dump file incrementally to avoid large memory consumption.
         # A configuration is returned as soon as its </dump> has been parsed.
         # The garbage collector is only disabled while parsing, not while the

addons/misc.py

@@ -48,7 +48,7 @@ def isStringLiteral(tokenString):
 def stringConcatInArrayInit(data):
     # Get all string macros
     stringMacros = []
-    for cfg in data.iterconfigurations():
+    for cfg in data.configurations:
         for directive in cfg.directives:
             res = re.match(r'#define[ ]+([A-Za-z0-9_]+)[ ]+".*', directive.str)
             if res:
@@ -76,7 +76,7 @@ def stringConcatInArrayInit(data):
 
 
 def implicitlyVirtual(data):
-    for cfg in data.iterconfigurations():
+    for cfg in data.configurations:
         for function in cfg.functions:
             if function.isImplicitlyVirtual is None:
                 continue
@@ -85,7 +85,7 @@ def implicitlyVirtual(data):
             reportError(function.tokenDef, 'style', 'Function \'' + function.name + '\' overrides base class function but is not marked with \'virtual\' keyword.', 'implicitlyVirtual')
 
 def ellipsisStructArg(data):
-    for cfg in data.iterconfigurations():
+    for cfg in data.configurations:
         for tok in cfg.tokenlist:
             if tok.str != '(':
                 continue

test/addon/test_cppcheckdata.py

@@ -93,6 +93,25 @@ def test_iterconfigurations(self, multi_cfg_data):
         assert not find_tokens(cfgs[0], 'foo')
         assert find_tokens(cfgs[1], 'foo')
 
+    def test_configurations_kept(self, multi_cfg_data):
+        data = cppcheckdata.parsedump(multi_cfg_data.filename)
+        cfgs = data.configurations
+        assert len(cfgs) == 2
+        # later accesses and iterconfigurations() return the same objects
+        assert data.configurations is cfgs
+        iterated = list(data.iterconfigurations())
+        assert len(iterated) == 2
+        assert iterated[0] is cfgs[0]
+        assert iterated[1] is cfgs[1]
+
+    def test_iterconfigurations_not_kept(self, multi_cfg_data):
+        data = cppcheckdata.parsedump(multi_cfg_data.filename)
+        first = next(data.iterconfigurations())
+        # without the configurations property each call parses the dump file again
+        assert next(data.iterconfigurations()) is not first
+        assert len(data.configurations) == 2
+        assert data.configurations[0] is not first
+
     def test_standards(self, sample_cfg, sample_cpp_cfg):
         assert sample_cfg.standards.c.startswith('c')
         assert sample_cpp_cfg.standards.cpp.startswith('c++')