Marty's typepad

Introducing Kasa - The personal finance tracker for the hashtag era

Kasa - The personal finance tracker for the hashtag era. Randomized testing data displayed.

It's that time of the year again when daylight is only seen from office windows and the darkness of evenings comes way too early. Getting much more home time, I wanted to finally get a better overview of my spending. Fortunately we live in the age of apps - you need something? There's an app for it. I searched for some of that do-it-all-for-you apps, but privacy was my primary concern. Pretty much all the popular personal finance apps take your financial data and upload it to some server to crunch it. I find that completely unnecessary. It's just a bunch of numbers, why does it need to live on a remote machine where I have zero control over it and zero control over who has access to what I spend my money on? All I want is to get the transactions file from my bank every week, throw it at some app, put some tags on things and see where my money goes. Simple enough to not require any fancy servers right?

So I turned my search into the open source ranks as I know there are some popular apps like KMyMoney or GnuCash. I tried all of them and I wasn't happy with any of them. KMyMoney and Skrooge, the duo from KDE Applications suite, are swiss-army knives that I personally find hard to navigate for my simple needs. Skrooge especially - after the import it generates a dashboard with 22 random sized widgets in no particular order. It felt so overwhelming and intimidating that I knew it's not for me and I had to close it immediately (sorry Skrooge). GnuCash takes 7 seconds to load for whatever reason and I couldn't figure out how to categorize transactions, which is the most important feature to me. Then there's HomeBank. This one came closest to what I was looking for but it was missing some features that I wanted, one of which is showing the actual amounts for tags. Also the categorization feels needlessly complicated. I feel like all the pieces are there, they're just not put together the way I wanted.

At the same time I was really missing coding in Qt. My last two years have been only Android Java and if you've ever worked in that, you know how frustrating it can get at times. Very, if you've never worked with it. Although it's helped a lot by Android Studio, which is just the best IDE I've ever used. Anywho, so there I was, wishing to do some C++/Qt again on long winter evenings without an app I wanted. So I just started coding - cause how hard can it be?™

Enter Kasa

It's a very simple app - you throw an OFX file at it, it stores all the transactions from it in local database, you put tags on them and then you profit. Nothing more, nothing less. For now. The functionality at this point is extremely basic, it's pretty much just a glorified spreadsheet, but already covers lots of my needs.

So what can you actually do? You can (multi-)tag and edit transactions. The tags are then put into a pie chart which shows you exactly where your money goes, tag by tag. You can explore your transactions by tags. You can also set custom date range for the data that it's showing, this way you can immediately get a report for last 7 days, last 17 days, last 7 months, fully up to you.

Here be dragons

Kasa right now is lightyears away from something I'd consider releasing into production. It's pre-pre-pre-alpha, but it's good enough to put it out there and get some extra eyes and hands on it. At this point it needs both code and design. Mainly design. Please get in touch if this sounds interesting to you and you'd like to contribute!

What's next

The main thing I'm working on right now is to provide much better tagging experience, including bulk tagging, tags auto-complete and easy tag removal. After that I'm going to add a setting where you can define your statement period and it will default the date range to that on start (and it will auto-advance each month too). Then it's in serious need of error handling. Right now it's virtually non-existent, it always assumes valid OFX file and so on. This will be crucial. Another thing I want to work on is multi-accounts support. While the database is already set up for that, the rest of the app isn't really; right now I'm mostly focusing on my single credit-card account use case. Next, I want to have some budgeting options per tags and then track how the budget is doing. Eg. set 100/month for eating out, how much was already spent and how much is left? Etc.

Get it (the code)

The app is all open source and you can get it and build it from here. All you need is Qt 5 (including QtCharts), CMake and libOfx.

Potentially FAQ:

What does "Kasa" mean?

The word means "cash register" in Czech. I started with it only as a codeword for the project but then it kinda stuck. I may still rename it somewhere down the line.

Is/Will Kasa be for me?

My ultimate end goal for this project is to be a good personal finance tracker - not manager. I simply want this app to show where the money goes each week/month/quarter based on bank provided files (with manual input later as well). If you want/need anything beyond that, stick with the existing apps like KMyMoney, Skrooge, GnuCash or HomeBank, they're much better suited for that.

Why not help improve existing apps?

I did consider it of course but there are couple reasons - my time for this is extremely limited. All the existing apps are pretty big and I wanted to spend my limited time on actually writing Qt code rather than trying to understand the existing codebases and then try to change those to what I wanted to have. I just wanted to code something and I wanted to scratch my itch. So I just put those two together.

QML/Kirigami?

Yes please! At this point I'm focusing on the data layer and business logic and I wanted the UI to remain as simple as possible for that, for which QWidgets are good enough. But at some point I definitely want to write a QML UI for this. QItemDelegates are....not so great :) Patches welcome!

When will this be ready for regular use?

I have no dates set. It's just a hobby project to refresh my C++/Qt knowledge and I work on it every now and then. Who knows, maybe it will never see an actual release. But if you wanna help, code or design, be sure to let me know!

That said, happy 2019!

Full emojis support for your KDE/Qt applications

So in the past month I've devoted my weekends to get proper emojis support ready for KDE software. Thanks to our fantastic KEmoticons framework, all the heavy lifting is already done, we just needed a theme and utilize the theme.

I've came across this awesome open source set called Emoji One, which features over 1600 emojis and is ready to be used with CC BY 4.0 license. I've created a KEmoticons theme out of the last released set, supporting both the :ascii: notation as well as the unicode symbols which were gladly contributed by David Gil Oliva. This theme of course supports the good old emoticons too :)

And today I'm happy to announce that all is merged in now and will be released with next KDE Frameworks release.

Using it is very simple, if your app is already using KEmoticons, you can just set the EmojiOne theme explicitly and profit. If your app is not yet using KEmoticons, using is as simple as these three lines of code:

KEmoticons *kemoticons = new KEmoticons();
emojiTheme = kemoticons->theme("EmojiOne");
QString textWithEmojis = emojiTheme.parseEmoticons("Your text :poop:");

Additionally, if you want your emojis to be in a specific size, you can now newly apply a size constraint by doing this:

kemoticons->setPreferredEmoticonSize(QSize(16, 16));

...before you call parseEmoticons(..) and your parsed emojis will be in size 16x16 pixels.

What's next?

Emojis parser is nice, but how about input? Right now I have set of QML-with-C++ goodness that provides an enhanced auto-height-resizing TextArea that features emojis autocompletion, emojis rendered inline and a custom plain-text converter to give you back the plain text with the emojis in :ascii: format (with some tweaking it can return the unicode character easily).

Inline emojis with autocompletion

One other thing I have ready is nicely pre-ordered QAbstractListModel-subclass that can be used in popups to allow users to input emojis with mouse.

These extras are not yet ready to be released though, still some issues left to iron out. I think I'll end up bundling all these into its own package and release it standalone. They will require only KEmoticons, Qt and QtQuick.

Qt-only?

Currently the emojis come in KEmoticons framework, which is very flexible and provides tons of features, like support for installing custom themes that you can bundle with your app. All those features unfortunately come with some additional dependencies, but I believe we can cut these down to a single tier-1 dependency. I'll probably look into slimming KEmoticons down in my free time, but if you're interested in using this in your Qt-only app, please let me know and I'll prioritize it.

 

Qt Backpropagation Neural Network

When doing some cleanup today I stumbled upon a little university project I wrote sometime ago. It's a basic implementation of backpropagation neural network written in Qt and I thought someone might find this handy, so I decided to share it.

The original design was done by Chesnokov Yuriy in his C++ implementation over at CodeProject.com. The network can contain arbitrary number of layers and each layer can have arbitrary number of neurons. However the implementation uses only feed-forward each-to-each connection between the neurons.

The example UI is probably confusing and not intuitive at all, but it was done only to demonstrate the network capabilities to the teacher, so take it as that.

How to use

All the network logic is implemented in network.cpp and neuron.cpp, mainwindow.cpp and race-client.cpp then contain examples of usage.

The code contains three example of setting and using the network. First one shows number recognition 5x5 raster, where all the network initialization is done in the code, including the training set (see MainWindow::recognizeNumber() method). Second example features loading the network data from files. The parsing is super-shitty, but...meh, good enough :) The third example was an actual assignment, which was to train a race car, that can follow supplied path. Basically we had to implement a race client that connects to a race server, where bunch of neural-net-trained cars can race against each other. Unfortunately I can't distribute the server nor the access to the public server, so the client implementation will have to do.

Setting the network is as easy as

m_network = new Network(QList<int>() << 25 << 18 << 8 << 4, 0.8f, 0.1f, 0.05f);

which will create a network with 4 layers (25 input neurons, 4 output), alpha = 0.8, learning rate = 0.1 and max allowed network error is 0.05.

Training the network is also easy:

m_network->trainSet(trainSet, outputs);

where trainSet is QList<QList<float> > (each 'inner' list is one training set) and outputs is the expected outputs for the train set, same as the trainSet - QList<QList<float> >. If the network was trained successfully, trainSet(..) returns true, otherwise false. 

Running the network on data is done by

QList<float> *netOutputs = m_network->run(runset);

where runset is one set of data, ie. QList<float>. This returns list with network outputs on the given set.

Gimme the code!

The code is hosted on github under GPLv3, so feel free to clone it, use it, share it, improve it, just keep the copyrights in place and follow the licence. 

https://github.com/mck182/QNeuralNet

Enojy :)

GSoC: Non-destructive image editing for digiKam

Published on digiKam's blog and Planet KDE    

Hello to all the great KDE people and especially digiKam fans! 

My full name is Martin Klapetek, but everyone calls me Marty and I can be found on most services like IM or IRC by the name mck182. I'm 22 and I'm from Czech republic and ever since I started with Linux, I immediately become a huge fan of KDE (3.5 back then). I'm a terrible musician (playing guitar, bass guitar, sometimes even piano [you don't want to be there]), promising art painter and hopefully-soon-to-be graduated college student with bachelor's degree in Computer Science.

I've been accepted to this year's Google Summer of Code with project to enable non-destructive image editing and image versioning for digiKam.

I'm very keen on taking photos and I've been using Google's Picasa for organizing the quantum of pictures I've taken. But the Linux version was more and more unusable lately for me, mainly because problems with multilibs (I'm running 64bit system and Picasa needs 32bit Qt libs and I'm using non-official-distro Qt packages, so there were some conflicts) and also because I hate it's wine-ish look & feel (& speed) and non-integrationess. So I've decided to use something more KDE fashioned and the sight was set on digiKam.

I like Picasa's possibility to edit your images in any way and then anytime you come back and feel like it's not good enough, you can simply reset the image to its original state and start over. This is the feature I'd like to bring to digiKam during this summer. This will go along with feature for having multiple versions with different modifications of one image and easily switching between them.

The non-destructive editing will work fully automatized. Now when you do some changes eg. use some filters or adjust levels, you need to save your image first and if you just save, you lose the original. With the non-destructive editing, it will automatically create new version from the original and it will store all the changes there, all without bothering user about saving the image and without actually touching the original. Every time you switch to the original and start new modifications, it will create another version alongside with the first one. Each version will also be manually forkable to another new version. 

That's in short my project for GSoC 2010. I'm very excited to finally do some real work for KDE and to contribute to the world of open source. I'm looking forward to work with the already-shown-to-be-awesome people from digiKam team and especially with Marcel Wiesweg, my mentor :)

I wish happy coding and lots of fun to all the accepted students. Let's make KDE proud of us :)

Marty

PS: More detailed posts will come when the real work will start ;)