Changelog

0.4.4

• Triumphant return of the readthedocs page

  • Tutorial

  • Major step forward documenting the Python library, along with multiple tiny tweaks to the Python codebase. Mainly cleaned up & removed unused methods.

• Fix disappearing feature_parameters in new analysis dialog and on the configure page

• Add CLI commands to interact with the git repository

  • For end users that may not want to deal with git

  • Throwback to the “mandatory update system” of the olden days:

    update to new release versions if any are available, but when you

    actually want to.

• Setup ~ CLI commands

  • Deployment scripts execute shapeflow/setup/post-deploy.py

    as a setup step

  • shapeflow.cli.GetCompiledUi

  • shapeflow.cli.SetupCairo

  • shapeflow.cli.Declutter

• Make cairo an optional dependency

  • No longer depend on OnionSVG; design file rendering is now handled by shapeflow.design.onions.Peeler and shapeflow.design.render.Renderer.

  • SVG layer operations (lxml) are decoupled from rendering

  • Renderers are selected based on the system configuration: cairo is still the go-to option, but we can now fall back on Wand (ImageMagick) and the commandline interface of a Windows installation of Inkscape.

• Fix tkinter dialog windows not appearing on Windows

  • tkinter can’t handle not being in the main thread, which was the case in the “updated” version of :module:shapeflow.util.filedialog.

  • Should not have deprecated subprocess-based filedialog script; it was added in the first place to solve this issue.

  • Made subprocess-based file dialogs complain on cancel / error; this bug stayed unnoticed because they used to be silent about it.

0.4.3

• Add frontend tests

  • And also some general clean-up and fixes in the process

• Deprecate caching contexts and related functionality

  • We’re assuming that caching will never be performed in advance of an analysis. Instead, we rely on caching during an analysis to speed up any subsequent analyses.

• Separate internal routing from general Flask routing

  • API routes are organised based on shapeflow.core.Dispatcher

  • Dispatcher instances map addresses to Endpoint instances

  • Nested Dispatcher instances in their own address space

  • The top-level Dispatcher has a flat address space of all endpoints, which it uses to resolve requests

  • The Flask server delegates requests to this top-level Dispatcher for addresses starting with "/api/"

• Expose Endpoint instances with own expose() method instead of global function

• Deprecate RootInstance / BackendInstance

  • Implementation should not care about routing

  Note

  This means that methods of BackendInstance subclass instances nested in VideoAnalyzer can no longer be exposed at Endpoint instances. Only methods of objects directly associated with Dispatcher instances can be exposed.

• More sensible API structure

  • Global top-level API at shapeflow.api.api

  • Group related functionality

    • api: general stuff

    • api.fs: dealing with files and directories

    • api.cache: dealing with the cache

    • api.db: dealing with the database

    • api.va: dealing with analyzers

    • api.va.<id>: dealing with a specific analyzer

• Open analyzers are handled by new Dispatcher instances

  • Analyzer methods should be exposed with the placeholder Dispatcher at api.va.__id__

    • By themselves, methods exposed in this way can’t be invoked since they don’t have an instance yet

  • New analyzers are opened from _VideoAnalyzerManager and given an id

    • Use shorter id strings for URL readability

    • Associate newly instantiated VideoAnalyzer with a new Dispatcher instance at api.va.<id>

    • This Dispatcher, binds methods exposed in api.va.__id__ to the VideoAnalyzer instance

    • Now these methods can be invoked when requested by /api/va/<id>/<endpoint>

  • Included in top-level address space at launch to reduce address resolution overhead

• Mirror API structure in frontend api.js

0.4.2

• Subcommands to divide up the functionality of the library.

  • Implemented to make accessing backend schemas easier when testing the frontend; instead of starting the whole server, run sf.py dump <path>. The server is now a subcommand, serve.

  • Potentially useful commands to add in the future

    • analyze could run a single analysis as specified in a .json file

    • checkout could set the repository to a specific version

    • setup could replace in-repo setup scripts

  • It may also be interesting to make these commands accessible from the frontend

• Some major naming changes

  • Entry point script shapeflow.py becomes sf.py

  • Server-related stuff renamed from main to server

0.4.1

• Tutorials and high-level documentation

0.4.0

Clean-up git history

• The first year of development was at isimple, named after the technology/the team that used it for some reason.

  Because the original repository was a bit too large, its git history was rewritten after moving to shapeflow. The old repository is still up to preserve this history and to support legacy deployment scripts.

  • gitsizer and bfg are nifty tools.

• Removed…

  • Compiled JavaScript from ui/dist/

  • An accidentally huge screenshot, mysteriously named datetime

  • An accidentally huge BMP file

• All in all, the repo went from almost 30MB to about 6MB.

bfg --delete-folders dist .
bfg --delete-files datetime .
bfg --delete-files img.bmp .

git reflow expire --expire=now --all
git --prune=now --aggressive

Note

A short summary of the major changes in the older versions will be added soon.