Developing

Backend

Setting up a development environment

To work on shapeflow, you should set up a virtual environment in .venv

python -m venv .venv
(.venv) pip install -r requirements.txt

With this environment in place, the entrypoint shapeflow.py can be run either from (.venv) or with your the global Python interpreter. In the latter case, .venv will be used implicitly.

Writing plugins

Features, transforms and filters are handled by a plugin system. You can use existing plugins as examples.

  • A plugin should be a subclass of its respective interface (Feature for features, TransformInterface for transforms and FilterInterface for filters)

  • If you define configuration for the plugin, it should derive from its respective configuration class as well (FeatureConfig for features, TransformConfig for transforms and FilterConfig for filters), and should be linked to the plugin class itself as the class variable _config_class. Configuration parameters should be defined as pydantic.Field instances, and should include a default and a description.

  • Plugins should have a descriptive docstring

  • Feature subclasses should also define a _label and _unit as class variables. These values will be used to label the y-axis on the result page of the frontend.

All plugin files are automatically loaded when shapeflow.plugins is imported.

Frontend

Setup

  1. Install Install NodeJS & npm

  2. Navigate to this directory and install the dependencies

    cd ui/ && npm install

Compiling the frontend

  1. (Set up)

  2. Compile

    cd ui/ && npm run generate

    The compiled files are stored in ui/dist/.

Running the frontend in development mode

  1. (Set up)

  2. Run the backend server (default address http://localhost:7951)

    (.venv) $ python .server.py

  3. Run the frontend development server (default address http://localhost:3000)

    cd ui/ && npm run dev

    The development server hot-reloads content from the source code in ui/ and proxies API calls to the backend server.

Generating deployment scripts

Deployment scripts are generated with gitploy.

  1. Tag the release in git

  2. Create a release on Github

  3. Compile ui/dist/, compress it with tar czf dist-<tag>.tar.gz dist/ and attach it to that release as a binary

  4. Create or update your .ploy file in shapeflow‘s root directory:

    1. Start from shapeflow/setup/ploy

    2. Add the tag of your release

    3. Double check that the check / setup script paths are still correct

  5. Run python -m gitploy in shapeflow‘s root directory.

TODO

Some known problems and minor feature ideas. Not everything in this list is worth spending time on, and some ideas are serious feature creep. Open issues for important stuff.

  • Check performance on slower hardware

  • Sidebar nav icon&text should aligned vertically

  • Frame & state image can get desynchronized sometimes

  • Dragging the seek control doesn’t update streams, but clicking/arows do (sometimes, sometimes it’s ok)

  • Alt scrolls through sidebar navs for some reason

  • Select previous results of the same analyzer in the results page

  • Masks misbehave when increasing DPI, work fine when decreasing DPI (????)

  • Skipped masks should be grayed out in the state image

  • Set a max width/height for graph (depending on number of features?)

  • Add AnalyzerState assertions to test_main.py

  • Sometimes roi resize gets applied to current & opposite side, jumpy

  • Highlight masks on hover in frontend

  • Set default filter/transform configuration when adding a new analysis

  • Optimize SchemaForm rendering speed

  • Add an option to export .meta files

  • Add an option to import analyses from .meta files and .xlsx files

  • Configure sidebar should have a fade on the bottom

  • Don’t catch events outside of frame boundary (v-bind style to div?)

  • Really fix oscillating parameter override categories

  • Two of the same feature should yield two separate graphs

  • On Windows, tkinter file dialogs don’t open when debugging

  • Support Anaconda environments

  • When current analyzer page becomes disabled, route away or gray out page

Already fixed?

  • get_overlay & get_overlay_png take 5 seconds to run sometimes

  • Align/Filter page: seek event on page load doesn’t always come through, or the streamed image doesn’t get updated

  • Frontend can freeze when adding a second/third/… analyzer

  • Adding new analyzers with large Nf is slow

  • Fix ROI rotation (probably need to initialize moveable with the aspect ratio of the design)

  • ConfigModel is made in doubles

  • CtrlZ / CtrlShiftZ requests undo_config/redo_config twice

  • When switching between analyzers, ROI sometimes gets stuck; modifying moveable doesn’t cause actual ROI to jump to the wrong one

  • Config events are sometimes missed on set_filter_click

  • Issues with page rebuild after switching analyzers multiple (3+ times)

  • Shouldn’t continue on to /api/va/<id>/launch if /api/va/<id>/set_config raises HTTP500

  • Reset filters state update should disable ‘Analyze’ button