Skip to content

Latest commit

 

History

History
79 lines (66 loc) · 6.02 KB

DEVELOPMENT.md

File metadata and controls

79 lines (66 loc) · 6.02 KB
Raw

Tips for developers

It's dangerous to go alone. Take these tips in case you need to fit Indico to your particular needs.

Initializing the database

Use indico db prepare to create your tables based on the SQLAlchemy models and set the migration status to the most recent alembic revision.

SQL Database migrations

Whenever you modify the database structure or want to perform data migrations, create an alembic revision. To do so, use indico db revision -m 'short explanation'; optionally you may specify --autogenerate to let Alembic compare your SQLAlchemy models with your database and generate migrations automatically. However, this is not 100% reliable and for example functional indexes will always show up as "new". So if you use autogeneration, always check the generated migration steps and modify them if necessary. Especially if you've already applied your change to the database manually or let SQLAlchemy create your new table you need to write the migration for it manually or DROP the table again so Alembic knows it's new.

To perform the actual migration of the database, run indico db upgrade or indico db downgrade. Migration should always be possible in both directions, so when writing a migration step make sure to test it and to implement both directions for structure and data even if that means dropping columns or tables. Losing data during a downgrade is acceptable as long as it's data that didn't exist before that revision. Please make sure to TEST migration in both directions!

When adding a new column that is not nullable, you need to add it in two steps: First create it with a server_default value set to whatever default value you want. Afterwards, use the alter_column operation to remove the default value. While keeping it would not hurt, it's better to stay in sync with the SQLAlchemy model!

Writing models

When writing/changing models or alembic revisions, run python bin/maintenance/update_backrefs.py to keep comments about relationship backrefs in sync and python bin/utils/db_diff.py to compare the models against your current database both to ensure your alembic revision is correct and that your own database is up to date.

Updating Python dependencies

We use pip-tools to manage our requirements.txt files. To update the pinned dependencies, use pip-compile -U for regular dependencies and then pip-compile -U requirements.dev.in for dev dependencies. Afterwards, check the diff for the requirements.txt files and consult each package's changelog for important changes in case of direct dependencies that aren't just patch releases. Once that's done, you MUST install Indico with pip install -e '.[dev]' and ensure nothing is broken (depending on what changed, make sure to test affected parts manually).

JS dependencies

To update deps within semver ranges, rm -rf package-lock.json node_modules followed by npm install usually does the job. Make sure to restart the build-assets.py script afterwards to ensure things like the webpack build aren't broken. It's also recommended to test both dev and prod builds, and to build plugins. For plugins, it's recommended to try both the normal and CERN plugins (since some of those such as the Burotel plugin use more 'magic' than the standard ones).

Once that's done, npm outdated -l will list the remaining packages to be updated. Often the semver-major changes are simply dropping support for old node versions and similar harmless things. But certain other packages are known to be problematic. Here's a list to save some time during the next updates:

  • @dr.pogodin/babel-plugin-react-css-modules, css-loader and mini-css-extract-plugin are very prone to problems if not all of them have the versions the other packages expect. Trying to update css-loader to the latest major version is not straightforward either. It's best to leave those packages alone until we either drop webpack or look into webpack's asset modules feature.
  • dropzone is still on a stable release, the new version is still in beta so not worth considering.
  • prettier, eslint-plugin-prettier and stylelint-prettier cannot be updated for now due to prettier 1.19 introducing a very shitty heuristic for function composition detection (compared to a simple list of function names that are doing function composition), so updating would mess up the formatting of our createSelector calls in a way that's much less readable than what we have right now. Check prettier/prettier#6921 for updates on this.
  • history, react-router and react-router-dom have some major changes e.g. regarding navigation blocking. It looks like they finally added unstable_useBlocker() to block navigation though. In the Room Booking module we are using connected-react-router though, which is dead, and not compatible with react-router 6 so it would need to be replaced with something else first.
  • husky could in principle be updated, but we use it in a very basic way (to setup and run the pre-commit git hook), and there is no need to switch to the newer and slightly more complex configuration of newer versions.
  • jquery-ui stays on ^1.12.1 until we can get rid of it one day. Newer versions break focusing input elements inside react dialogs opened from a jquery ui modal.
  • react-dnd and react-dnd-html5-backend: TODO
  • react and react-dom would be great candidates to update, but it's stuck until react-dates starts supporting React 18 (react-dates/react-dates#2199) which depends on Enzyme supporting it (enzymejs/enzyme#2524)
  • react-dropzone requires MIME types instead of just file extensions for the files it should accept in newer versions, but we have various places where we have only file extensions, so we can't easily update.
  • react-leaflet, react-leaflet-draw and react-leaflet-markercluster are stuck because react-leaflet v3 decided to adept some dumb "ethical" license (PaulLeCam/react-leaflet#698) which is not considered a proper Open Source license.