Allow the listener thread to use touchpadlib interface to receive events.

The name of the initalize_touchpadlib_usage() function should be initialize_touchpadlib_usage().

At the moment our app is killing the CPU performance. On my older laptop it happens to take 100% of the CPU when working. It might be due to the while loops. We should investigate and find out how to lower the CPU usage.

It is not hard-coded that a touchpad is always event10. We have to detect and open the correct file descriptor.

At the moment sys.exit() functions don't exit the program when called - instead, they only kill the thread they were called in.

Although we might consider moving TouchpadSignal.initialize() up before any threading happens it does not solve the problem of the sys.exit() within the TouchpadSignal.fetch_touchpad_event().

If we want to use the pressure parameter from the touchpadlib we have to find out what it means really.

For example, the pressure parameter in the Hasee Q540S touchpad takes on values from range [0;255]. If I press the touchpad really hard it only takes on values like 70 however.

This is why we have to remember about it.

There are a lot of @todo's in the app's code.

It should look like this more or less:

gcc -c -fPIC touchpadlib.c -o touchpadlib.o
gcc -shared -Wl,-soname,touchpadlib.so -o touchpadlib.so  touchpadlib.o

@michal7352, could you add a file with a list of dependencies so that the app could be installed easily?

This migth be helpful: https://www.fullstackpython.com/application-dependencies.html

(PR #58)

-Wall and all the other

What do you think? If we want to use this tool we should add a configuration file and a badge.

Here is our first check I did to try out this tool: https://landscape.io/github/0mp/io-touchpad

It is not valid however since it thinks that we use Python 2.*.

I got an error like this:

Available devices:
/dev/input/event0:    AT Translated Set 2 keyboard
/dev/input/event1:    SynPS/2 Synaptics TouchPad
/dev/input/event2:    Video Bus
/dev/input/event3:    Power Button
/dev/input/event4:    Sleep Button
/dev/input/event5:    Lid Switch
/dev/input/event6:    Power Button
/dev/input/event7:    PC Speaker
/dev/input/event8:    HDA Digital PCBeep
/dev/input/event9:    HDA SIS966 Mic
/dev/input/event10:    HDA SIS966 Headphone
Select the device event number [0-10]: 1
Touchpad specification: min/max x: 1472/5472, min/max y: 1408/4448, min/max pressure 0/255.
Use your touchpad as usual. Have a nice day!
Traceback (most recent call last):
  File "./app.py", line 68, in <module>
    main()
  File "./app.py", line 66, in main
    application.application_thread(thread_queue, learning_mode, training_size)
  File "/home/matangi/Documents/io-touchpad/app/threads/application.py", line 30, in application_thread
    clsf = classifier.Classifier()
  File "/home/matangi/Documents/io-touchpad/app/classifier/classifier.py", line 19, in __init__
    self.learning_model = pickle.load(file_with_model)
  File "sklearn/neighbors/binary_tree.pxi", line 1129, in sklearn.neighbors.ball_tree.BinaryTree.__setstate__ (sklearn/neighbors/ball_tree.c:9307)
  File "sklearn/neighbors/binary_tree.pxi", line 228, in sklearn.neighbors.ball_tree.get_memview_ITYPE_1D (sklearn/neighbors/ball_tree.c:2867)
ValueError: Buffer dtype mismatch, expected 'ITYPE_t' but got 'long long'

It seems like @michal7352 saved some classifier data in some sort of a binary format which is not compatible with 32-bit systems.

Relates #58

As of the commit 77bb76c from the #34 pull request the application does not exit with 1 if an error occured.

We still don't know who dense the data we receive is comparing to an average touchpad resolution.

Let's try to visualise the data.

Use UML.

There is a lot of code that could be removed.

Just be careful when you remove those defines!

At the moment the data files are scattered around the app/ directory. We should move them to app/classifier/data according to this example:

setup.py
src/
    mypkg/
        __init__.py
        module.py
        data/
            tables.dat
            spoons.dat
            forks.dat

Reference: https://docs.python.org/2/distutils/setupscript.html#installing-package-data
Relates: pull request #58

Learning mode would run without pre-set size of traning. Applcation would do learning and stop, on some event and confirmation that user want to end the learning.

Instead of storing symbols in two files, we could do one file and use 'pickle' library to save list of list of signals.

This tool will be able to teach our application new symbols.

app.py is a little bit messy at the moment. This is what some smart people told me.

It is not clear yet whether the detection system is based on machine learning or not.

We detect end of the group when:

• time between last two signals is longer than some constant
• waiting for new signal persists
• process signal of pressure with value lower than 1

We should unify error messages.

How about something like ERROR|WARNING: <module>: <message>?

http://stackoverflow.com/questions/12666153/how-to-find-leaks-in-python-ctypes-libraries

Measuring the similarity between symbols should be independent from location, and scale of the whole figure, so we need to normalize the drawn symbol. It could mean for example, that point coordinates will be translated, so that, their centroid is the origin of the coordinate system, and the diagonal of the "bounding box" has some constant length (or the average distance from centroid to symbol point is some constant).

During the second iteration we want to add and handle only one hard-coded symbol.

Wiki page with the summary: https://github.com/0mp/io-touchpad/wiki/Summary:-The-second-iteration

• Use bullet points.
• Don't try to write a long text.
• Make sure to tell the supervisor about all the changes.

• We created a tool for visual analysis and a wiki page about it.
• Automatic detection of that the touchpad parameters can take on values range from.
• The executor module launches gedit.

Every touchpad has different resolution and can return values from different ranges. It would be nice if we can detect it automatically.

• Create n equidistant points.

• Test the symbol detection system.
• Test the application after implementing new features.

If the symbol was recognised then app.py should execute a command linked to that symbol. For example: launch a web browser.

At the moment we do not close the device file when we close the application.

Although we add only one symbol during this iteration we should think about the way to add more symbols in the future.

We need an app which will be able to receive signals from a touchpad.

Decide if the drawn symbol is similar enough to the given symbol

Relates #45

We need to implement a simple interface to receive touchpad signals. We will modify the existing evtest code for this.

The touchpad_signal_object is not freed at the moment. We should do something about it.