Code cleanup and performance improvements - Part 2

[brainplot]

I found a few more issues after a more thorough inspection of Clazy, that was disabled by default.

Description

• Avoid allocating temporary containers
  Calling toLower() on a QString will cause a new string to be allocated. Since the function is essentially invoked to perform a case-insensitive comparison, we can invoke the contains() method with the Qt::CaseInsensitive enum value to do just that, as opposed to calling toLower() on two different strings.

• Remove unused variables
  For some reason there were completely unused variables in the code. One of them was even initialized with the result of a method, but never used nonetheless (maybe it was the source of a bug).

• Use QVariant::toUrl() instead of QVariant::value<QUrl>()
  The analyzer was suggesting me to do this. I guess the latter is more performant since it's not a template 🤔

• Fix typo in parameter name
  This is a very minor thing. There was a typo in the declaration of a function (erroMessage instead of errorMessage), which was spelt correctly in the definition of the same function, and the analyzer reported the inconsistency between the two.

• Improve type-safety
  Some zeros used to initialize a pointer were left out of my previous commit. I think they were introduced with the PR which adds more expiration presets.

How has this been tested?

Compiled with all tests enabled (GUI tests included) and -DWITH_DEV_BUILD on and all tests passed.

Types of changes

• ✅ Bug fix (non-breaking change which fixes an issue)

Checklist:

• ✅ I have read the CONTRIBUTING document. [REQUIRED]
• ✅ My code follows the code style of this project. [REQUIRED]
• ✅ All new and existing tests passed. [REQUIRED]
• ✅ I have compiled and verified my code with -DWITH_ASAN=ON. [REQUIRED]

[phoerious]

Please revert the for loop changes. Qt containers will not detach as long as you don't call non-const functions on them or the reference counter is 1.

[phoerious]

See comment.

[brainplot]

@phoerious I just read about this feature which I didn't know about. However, isn't it better to explicitly state in code the intent of not copying a container instead of relying on an implementation detail?
This is just my thought on this. If it's wrong, I can revert my commit, no problem.
Even on StackOverflow, people seem to agree that keys() is not an efficient solution for iterating over a QMap.

[phoerious]

We use asConst() whenever we are iterating over a non-const container and there is no reason for copying rvalues.
Implicit sharing is not an implementation detail but an integral part of the Qt API.

[brainplot]

Ok, I'll revert the changes to the for-loops in my commit.
Do I have to revert my change to the toLower() call? Using Qt:CaseInsensitive will avoid the allocation of the new string and seems a more natural approach to the problem to me.

[droidmonkey]

I am not a fan of using iterators in Qt, it kind of defeats the whole point.

[droidmonkey]

Keep the Qt:CaseInsensitive, that is a good change and makes the code more readable in general.

[phoerious]

I haven't reviewed the full code yet. But reverting the iterator changes is the most important point. You may use STL-style iterators for maps for efficiency, but it's extremely ugly. After all, we don't really have performance problems.

[brainplot]

@phoerious @droidmonkey I reverted my changes.

[brainplot]

I ran make format and noticed that the beautifier applied a lot of fixes to the code formatting.
I've gone through each file one by one with my difftool, and wrapped with // clang-format off and // clang-format on those lines that looked more readable with manual formatting rather than automatic formatting.

I did this because, when I made my first PR, I wanted to make sure that my code followed the project's coding style, as stated in the CONTRIBUTING document; when I ran make format, however, I ended up with lots of modified files in my working directory. Now people should be able to format their code automatically.

[droidmonkey]

Can I give like 5 hearts? Edit, 1 per comment 😁

[droidmonkey]

Can you add these changes to your list:

Function
    isNull() const : bool
Found usages  (7 usages found)
    DatabaseOpenWidget.cpp  (1 usage found)
        169 if (!pw.isNull()) {
    DatabaseTabWidget.cpp  (2 usages found)
        160 if (!i.value().dbWidget->dbHasKey() && !(pw.isNull() && keyFile.isEmpty())) {
        199 if (!pw.isNull() || !keyFile.isEmpty()) {
    KdbxXmlReader.cpp  (1 usage found)
        734 if (!ref.first.isNull() && !ref.second.isNull()) {
    TestKeePass1Reader.cpp  (1 usage found)
        277 if (!password.isNull()) {
    TestKeys.cpp  (2 usages found)
        229 if (!error.isNull()) {
        287 if (!error.isNull()) {

QString::isNull() is inappropriate usage, should be QString::isEmpty()

[brainplot]

@droidmonkey Done! I also replaced a few checks of the form of if (obj.size() > 0) with if (!obj.empty()).
It looks more readable in my opinion.

[droidmonkey]

Please rebase onto the new develop. @phoerious please update your review.

[brainplot]

@droidmonkey done!

[droidmonkey]

Going to merge in the advanced search PR, then Phoerious' upcoming databasetabwidget refactor, then we'll rebase this one and merge next.

The changes were made

[brainplot]

Good, let me know when you're done so that I can rebase.

[droidmonkey]

@brainplot major refactors are complete, please rebase and fixup where necessary.

[brainplot]

Alright, this one was a little tricky to rebase because there was a good bunch of conflicts to deal with; but I managed it to do it correctly, I believe, since I ran all the tests and they all passed.

I made one more change at the end: there were two variables that could have their scope reduced by making them have static storage.
In one case, the variable was declared in the .cpp file as file-scoped but it was only used in a single function. I moved the variable into the function it was used in and made it static so that it'll get initialized only once regardless of how many times the function is called.

-// Escape common regex symbols except for *, ?, and |
-auto regexEscape = QRegularExpression(R"re(([-[\]{}()+.,\\\/^$#]))re");
-
 QRegularExpression convertToRegex(const QString& string, bool useWildcards, bool exactMatch, bool caseSensitive)
 {
     QString pattern = string;

+    // Escape common regex symbols except for *, ?, and |
+    static auto regexEscape = QRegularExpression(R"re(([-[\]{}()+.,\\\/^$#]))re");
+
     // Wildcard support (*, ?, |)
     if (useWildcards) {
         pattern.replace(regexEscape, "\\\\1");
         pattern.replace("*", ".*");
         pattern.replace("?", ".");
     }

     // Exact modifier
     if (exactMatch) {
         pattern = "^" + pattern + "$";
     }

     auto regex = QRegularExpression(pattern);
     if (!caseSensitive) {
         regex.setPatternOptions(QRegularExpression::CaseInsensitiveOption);
     }

     return regex;
 }

In the other case, the variable was declared as file-scoped and used in two different functions so I couldn't move it anywhere. However, enclosing it in an anonymous namespace will prevent it to "leak" into other translation units.

+namespace
+{
+    MainWindow* g_MainWindow = nullptr;
+}
+
 const QString MainWindow::BaseWindowTitle = "KeePassXC";
 
-MainWindow* g_MainWindow = nullptr;
 MainWindow* getMainWindow()
 {
     return g_MainWindow;
 }
 
 MainWindow::MainWindow()
     : m_ui(new Ui::MainWindow())
     , m_trayIcon(nullptr)
     , m_appExitCalled(false)
     , m_appExiting(false)
 {
     g_MainWindow = this;
 
     m_ui->setupUi(this);
 
    /* ... */
}

[droidmonkey]

Yah that's what I meant to do 😉

[phoerious]

You could also use an anonymous namespace for the latter case. Perhaps a bit more idiomatic C++.

[brainplot]

@phoerious thanks for your suggestion. I'll amend my commit in a few hours.

[droidmonkey]

I did another rebase. I greatly simplified the use of clang-format off/on around signal/slot connections. Instead of isolating problem ones, I just wrapped the whole block with them. I fixed other areas where the clang-format comment was causing readability to go down.

[brainplot]

I thought about doing that as well, but since turning clang-format off on a line basis will prevent it from fixing formatting issues that may be present on the whole line, I just wanted to turn it off in the narrowest scope possible. I was even considering something like this:

connect(selectionModel(),
	// clang-format off
	SIGNAL(selectionChanged(QItemSelection,QItemSelection)),
	// clang-format on
	SLOT(emitEntrySelectionChanged())
	);

but arguably that looks even uglier.

Regarding to your changes, how do I integrate them into mine so that I can then rebase my branch?

[droidmonkey]

Sorry I was waiting for my local tests to finish building, its pushed now (to your branch). The idea for me was that if we are imposing clang-format off then that means it is manually "beautiful". In those blocks we can handle the formatting.