TachyPy

TachyPy is a psychophysics engine for Python focused on precise visual timing with OpenGL rendering, a GLFW-first display/input backend, and experiment-friendly stimulus helpers.

Why "TachyPy"?

The name TachyPy comes from the tachistoscope: a classic laboratory instrument used to present visual stimuli for precisely controlled, brief durations. TachyPy aims to bring that timing discipline into Python experiments while keeping stimulus code readable and inspectable.

Timing Validation

TachyPy's GLFW backend was tested with a photodiode setup using a centrally presented white square on a uniform gray background. The white square was shown for one frame on a 60 Hz monitor. The photodiode was placed at the screen center, over the white square, and the recording captured both serial trigger events and the photodiode signal.

In the after-flip trigger condition shown here, the dashed line marks the serial flash trigger. The blue trace is the median photodiode waveform, with the shaded region showing the 5th to 95th percentile range across flashes.

Summary for this run:

• Median rise after flash trigger: 6.35 ms
• SD of rise after flash trigger: 0.20 ms
• Median photodiode pulse width: 16.60 ms
• SD of photodiode pulse width: 0.18 ms

These measurements are hardware/display dependent, but they provide a concrete validation pattern for TachyPy timing tests: combine serial trigger logging, photodiode measurements, and TachyPy flip timestamps rather than relying on software timestamps alone.

Highlights

• OpenGL stimulus rendering (Texture, Shapes, fixation, etc.).
• GLFW-first window/input handling via Screen for tighter display control.
• Backend-aware input handling through ResponseHandler.
• Multiple text paths:
  • Text (system fonts via FreeType + HarfBuzz; polished default),
  • GLText (OpenGL bitmap glyphs),
  • GLTextSDF (distance-field text),
  • GLSystemText (backward-compatible explicit name for Text).
• Psychophysics helpers (make_gabor, gratings, normalization, dithering).
• Audio playback utility (Audio) backed by tachyaudio.
• Test suite for core logic and regressions.

Installation

Install base package:

pip install tachypy

The base install includes GLFW for display/input, PyOpenGL, Pillow text support, pyserial for serial/trigger workflows, and TachyAudio for audio playback. Pygame support has been removed; GLFW is the supported display/input backend. TachyAudio is currently published as a beta release; TachyPy requires tachyaudio>=0.2.0b2, which includes the Windows wheel fix. If your pip resolver refuses pre-releases, pass --pre explicitly.

Editable install for development:

git clone https://github.com/Charestlab/tachypy.git
cd tachypy
pip install --pre -e .

Optional extras:

pip install -e ".[test]"        # pytest
pip install -e ".[text]"        # Pillow text fallback
# Audio support (tachyaudio) is included in the base install

Audio dependency

TachyPy audio now uses tachyaudio>=0.2.0b2. TachyPy no longer depends on sounddevice or requires users to install PortAudio separately. Hardware/audio device validation should still be done on the lab machine that will run the experiment.

Quick Start

from tachypy import Screen, ResponseHandler

screen = Screen(
    screen_number=0,
    fullscreen=False,
    width=1280,
    height=720,
)
responses = ResponseHandler(screen=screen)

running = True
while running:
    screen.fill((128, 128, 128))
    screen.flip()

    responses.get_events()
    if responses.should_quit() or responses.was_key_pressed("esc"):
        running = False
    if responses.was_key_pressed("space"):
        print("Space pressed")

screen.close()

To run the full demo:

python example_tachypy.py

To run the fullscreen GLFW clock/stop-timer demo:

tachypy-clock-demo

From a source checkout:

python clock_timer_demo.py

Use Esc to quit, click START/STOP/RESET, or use Space and R. For development, use tachypy-clock-demo --windowed or python clock_timer_demo.py --windowed.

Choose a font for demo text rendering with GLFW Text/GLSystemText:

TACHYPY_FONT="Avenir Next, Helvetica, Arial" python example_tachypy.py

Backend Notes

• Screen(backend="glfw"): GLFW-managed window/events, with top-left logical coordinate handling aligned to TachyPy conventions.
• Screen(...) defaults to GLFW, so most code does not need a backend argument.
• Screen(...) also presents 60 neutral warmup frames by default before experiment timing begins; set warmup_frames=0 to disable or choose another count for a specific display.
• Initialize ResponseHandler(screen=screen) for input. It calls screen.poll_events() and owns keyboard/mouse state; Screen does not track participant responses directly.
• DraggableManager reads mouse transitions through ResponseHandler(screen=screen).

Text Rendering Notes

• Text is the polished system-font renderer and is equivalent to GLSystemText.
• GLText/GLTextSDF/GLSystemText render text directly in OpenGL and are backend-independent.
• GLSystemText supports system font selection by family name, fallback list (e.g. "Avenir Next, Helvetica, Arial"), or direct font file path.
• For production instruction text, prefer Text with .[system_text].
• The old texture-backed constructor is backbenched as tachypy.text.LegacyText.

API Naming

The psychophysics module exposes modern English APIs (for example make_gabor, make_sine_grating, normalize_to_unit_interval). Legacy French names remain as compatibility wrappers and emit DeprecationWarning.

Testing

Run tests:

pip install -e ".[test]"
pytest

Current suite covers audio timing helpers, response/key state handling, backend behavior, psychophysics invariants, text layout/renderer basics, and other regression-prone utility paths.

CI uses mocked TachyAudio streams for deterministic unit tests. Keep hardware audio-device validation as a dedicated local or lab-machine test.

Documentation

Expanded docs live in /docs and include:

• getting started
• backend behavior and input routing
• text rendering options
• audio backend guidance
• examples and contribution workflow

Hosted docs (Read the Docs): https://tachypy.readthedocs.io/

If Read the Docs is not auto-updating after pushes, reconnect GitHub in RTD and re-sync project webhooks from the RTD project settings.

Main Modules

• screen.py: display/context lifecycle and backend abstraction.
• responses.py: keyboard/mouse event handling and key-state queries.
• text.py, gltext.py, gltext_sdf.py, glsystemtext.py: text rendering.
• textures.py, shapes.py, draggable.py, scrollbar.py: visual primitives.
• psychophysics.py: stimulus generation and normalization utilities.
• audio.py: sound playback and timing helpers.

Contributing

1. Fork and clone the repository.
2. Create a branch for your change.
3. Add tests for behavioral changes.
4. Run pytest.
5. Open a pull request.

License

MIT. See LICENSE.