Jure Repinc posted a small review of the book The Book of Qt 4
The Art of Building Qt Applications by Daniel Molkentin. My copy arrived yesterday from amazon.de. After hacking all day in PHP – I love that language, but I deny I ever said that when asked in public – I thought I read some chapters.

It’s indeed a very nice book. It explains understandably why you should put things on the stack or the heap, it’s better to remember when you understand why something is done a certain way then just doing so. Also I never understood what was meant by implicitly sharing. All clear now. Qt is smart.

I also came to some solution for Mailody. In the KDE3-series it was optmized to show a large amount of headers in threads quickly. Of course in Qt4 this can all be thrown away. Even more, the whole system of the header list was rewritten in a model/view. But I never got the same performance of the KDE3 Mailody. This due to a combination of bad implementation of the model/view framework and the previous optimalisations now biting me.

The Book pointed me to the obvious answer to my problem though. Mailody uses a sqlite backend for caching the mails. Qt now has Sql classes, so it should be possible to use Qt for accessing the database, which alse removes the dependency on libsqlite (I think). Using those classes is not spectacular, but what is spectacular that there is a Qt model for the Sql-database. This means I can get rid of the whole model part I wrote (and that stinks) and use a standard QSqlQueryModel in combination with a slightly adjusted QTreeView.

It would have been nice if the book told me something about the performance of all those classes. In what ways are the qt-sqlite classes faster then the regular implementation? What is the performance of the model/view approach. For Qt3 I made a system that only parses the main part of the headers when the those headers had to be shown in the headerview. So as long as the user is at the bottom of the mailbox, the headers of the messages at the top are not yet filled in, only when the user scrolls up and the headers are needed, those were parsed and shown. Which makes it pretty fast, as the parsing of the messages is the time consuming part. It would have been nice to get a feeling how it’s done in Qt’s-Interview, but I can also understand that’s out of scope for the book.

When I have the model/view setup correctly, the threading part should be done in the mapToSource() and mapFromSource() within the proxy, but that can be done at a much later date. Till already told me to do it in the proxy and the book told me the same. I better listen and do it that way.

I’ve not been very active on Mailody lately, but that might change. But ideally I would finish the book first. And the one from Johan Thelin too. And finish RSIBreak.