An input event recorder and player for automatic testing
pip install snitch-ci==0.7.2
Snitch is a Python3 program using the Qt5 framework for GUI and the pyautogui and pynput modules for automation.
A working installation of Python3 with pip package installer.
In order to record and control the mouse on Linux systems, the packages
python3-dev are required.
pyautogui module on Linux relies on the
scrot utility to perform screen
captures, it must be installed separately. See the note about XWayland at the
end of this document.
The installation of the
scikit-image module on Windows requires the
Microsoft Build Tools 2015 (C++ compiler)
to be installed.
The use of Python 3.8 is recommended for the menubar to behave properly.
Snitch-CI provides OCR features through the use of the Tesseract OCR software, which must be installed separately. The minimum recommended version is 4.1 (prior versions don't support the use of a custom target character set).
sudo add-apt-repository ppa:alex-p/tesseract-ocr sudo apt-get update sudo apt-get install tesseract-ocr
brew install tesseract
Snitch is available on PyPI under the
snitch-ci. You can install it by running the command:
pip install snitch-ci
Once installed, start the program by simply running the commamd
After having insstalled the prerequisites, use the following commands to respectively:
pip install -r requirements.txt bash build.sh python -m snitch
Click on the
Record button. From now on, all the input events from the mouse
and the keyboard are recorded. Similar events may be gouped together, e.g.
multiple letters typed consecutively are merged into a single TextEvent. All the
recorded events appear in the uper table in the Snitch inteface. The following
events are recognized:
To stop the recording, just click on the stop button. This last click is automatically removed from the list of the recorded events.
On the right hand side of the recorded events list, the Snitch UI shows the properties (such as click location, which keys are pressed, …) of the currently selected event and buttons to add, remove and reorder events in the list.
To start the playback of a recorded sequence, click on the
Play button in the
upper part of the Snitch UI. The events are played with a default time delay of
0.2s. Each event is selected while played so you can follow the progression of
the sequence. You can also stop the playback by pressing
Note: For technical reasons, keyboard events can't be recorded while events
are played back. So the only time the press on the
Esc key can be caught is
between two events. As a consequence, multiple press on the
Esc may be
required for the sequence to stop.
The lower part of the Snitch UI is dedicated to the screenshots management. To
record a region of the screen as a reference to compare to the state of the
tested application after an automatic playback sequence, click on the
Add result button. The screen darkens a bit and the mouse cursor becomes a
crosshair. Define the area of interest by drawin a rectangle area around it
(click and drag).
On mouse button release, the capture is performend and appears in the list of
screenshots. When selecting a screenshot, a preview is displayed on the right
hand side of the UI along with its properties (size, position) and buttons to
add, remove and reorder the screenshots. Adding a screenshot with the add button
is exactly the same as using
You can save the recorded sequence of events by clicking on the
or by selecting the
File > Save… menu option. The formats available are JSON
(the images are Base64 encoded) and Pickle.
Load a previously saved file by selecting the
File > Open… menu option.
The following video presents some of the various features for events recording:
This video shows how the the screen captures of the results are taken, and how the differces between the reference image and the test result are displayed.
Feature under development.
While recording keystrokes, typing too fast, especially with key modifiers, can result in skipping the modifier, or assinging it to the next key. Typing slowly can avoid the problem. You can also edit the event manually afterwards and enter the intended text.
While using context menus with submenus be sure to click on every submenu title to record a mouse position on the item an prevent the cursor to skip steps on playback.
The most recent versions of gnome-shell are based on XWayland instead of X11 as display server. One of the main features of Wayland is the application separation, i.e. each application is isolated from one another and is not allowed to access graphical properties (window size, position, …) of others. As a consequence, the applications running with the native gnome-shell framework are invisible to Snitch.
Moreover for those versions, even if the actions can be recorded for the native
X11 applications, the scrot utility (used by
pyautogui to perform screenshots)
produces only black pictures. As a workaround, it's possible to install the
Gnome native capture utility
gnome-screenshot and create a wrapper script
scrot. The most basic way to do that is putting the following script
/usr/bin/scrot file (and granting it execution permissions):
#! /bin/bash gnome-screenshot -f $@