new loplugin:inlinefields

look for fields which can be declared inline in the parent class.

start with some likely candidates in svx

Change-Id: I56cdca273272b72bb728ed2e3f5e1e976f8c7c32
Reviewed-on: https://gerrit.libreoffice.org/36262
Tested-by: Jenkins <ci@libreoffice.org>
Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>

1 files changed, 81 insertions, 0 deletions

diff --git a/compilerplugins/clang/inlinefields.py b/compilerplugins/clang/inlinefields.py
new file mode 100755
index 000000000000..2605fd7bc173
--- /dev/null
+++ b/compilerplugins/clang/inlinefields.py
@@ -0,0 +1,81 @@
+#!/usr/bin/python
+
+import sys
+import re
+import io
+
+definitionToSourceLocationMap = dict() # dict of tuple(parentClass, fieldName) to sourceLocation
+definitionSet = set()
+excludedSet = set()
+deletedInDestructorSet = set()
+newedInConstructorSet = set();
+
+# clang does not always use exactly the same numbers in the type-parameter vars it generates
+# so I need to substitute them to ensure we can match correctly.
+normalizeTypeParamsRegex = re.compile(r"type-parameter-\d+-\d+")
+def normalizeTypeParams( line ):
+    return normalizeTypeParamsRegex.sub("type-parameter-?-?", line)
+
+# reading as binary (since we known it is pure ascii) is much faster than reading as unicode
+with io.open("loplugin.inlinefields.log", "rb", buffering=1024*1024) as txt:
+    for line in txt:
+        tokens = line.strip().split("\t")
+        if tokens[0] == "definition:":
+            parentClass = normalizeTypeParams(tokens[1])
+            fieldName = normalizeTypeParams(tokens[2])
+            sourceLocation = tokens[3]
+            fieldInfo = (parentClass, fieldName)
+            definitionSet.add(fieldInfo)
+            definitionToSourceLocationMap[fieldInfo] = sourceLocation
+        elif tokens[0] == "excluded:":
+            parentClass = normalizeTypeParams(tokens[1])
+            fieldName = normalizeTypeParams(tokens[2])
+            fieldInfo = (parentClass, fieldName)
+            excludedSet.add(fieldInfo)
+        elif tokens[0] == "deletedInDestructor:":
+            parentClass = normalizeTypeParams(tokens[1])
+            fieldName = normalizeTypeParams(tokens[2])
+            fieldInfo = (parentClass, fieldName)
+            deletedInDestructorSet.add(fieldInfo)
+        elif tokens[0] == "newedInConstructor:":
+            parentClass = normalizeTypeParams(tokens[1])
+            fieldName = normalizeTypeParams(tokens[2])
+            fieldInfo = (parentClass, fieldName)
+            newedInConstructorSet.add(fieldInfo)
+        else:
+            print( "unknown line: " + line)
+
+# Invert the definitionToSourceLocationMap
+# If we see more than one method at the same sourceLocation, it's being autogenerated as part of a template
+# and we should just ignore it
+sourceLocationToDefinitionMap = {}
+for k, v in definitionToSourceLocationMap.iteritems():
+    sourceLocationToDefinitionMap[v] = sourceLocationToDefinitionMap.get(v, [])
+    sourceLocationToDefinitionMap[v].append(k)
+for k, definitions in sourceLocationToDefinitionMap.iteritems():
+    if len(definitions) > 1:
+        for d in definitions:
+            definitionSet.remove(d)
+
+tmp1list = list()
+for d in definitionSet:
+# TODO see comment in InlineFields::VisitCXXDeleteExpr
+#    if d in excludedSet or d not in deletedInDestructorSet or d not in newedInConstructorSet:
+    if d in excludedSet or d not in newedInConstructorSet:
+        continue
+    srcLoc = definitionToSourceLocationMap[d];
+    tmp1list.append((d[0] + " " + d[1], srcLoc))
+
+# sort results by filename:lineno
+def natural_sort_key(s, _nsre=re.compile('([0-9]+)')):
+    return [int(text) if text.isdigit() else text.lower()
+            for text in re.split(_nsre, s)]
+tmp1list.sort(key=lambda v: natural_sort_key(v[1]))
+
+# print out the results
+with open("loplugin.inlinefields.report", "wt") as f:
+    for v in tmp1list:
+        f.write(v[1] + "\n")
+        f.write("    " + v[0] + "\n")
+
+