hashlash / cimto Goto Github PK

From #19.

Options for already imported problems:

• ignore
• update/fix
• raise error

• Set ALLOWED_HOSTS and CSRF_TRUSTED_ORIGINS based on the gp url [port] from gitpod.
  We can set this to an environment variable on gitpod setup via .gitpod.yml.

Use problemset grouping like GitHub/GitLab repositories. Each repository belongs to a namespace, which could be a user namespace or an organization namespace.

Gitlab references:

• https://gitlab.com/gitlab-org/database-team/team-tasks/uploads/62c7b49b943cb2396dd17c6323acd6c1/GitLab.com_ERD_-_2020-11-05.pdf
• https://dbdiagram.io/d/5fb7f8f53a78976d7b7cc6fb
• https://gitlab.com/gitlab-org/gitlab/-/blob/master/db/structure.sql
• https://azimutt.app/gallery/gitlab

From #19

For kujawab, we can show the progress of all "categories" and then the progress for importing all problemsets of each category. But we might need to think about more general format.

Use token authentication for api

https://artofproblemsolving.com/wiki/index.php/AMC_Problems_and_Solutions

Some problems might have categories

https://www.mathjax.org/

• Enable MathJax plugin in ckeditor
• Ensure MathJax is allowed by bleach (or any future sanitizer we used)

It would be preferable to render the tex as "text" (like on math.stackexchange.com) instead of an image.

We currently run collectstatic on the web process. This is inappropriate (the web process should only run the web server) and might be the cause of slow cold startup. See the gap between the successful collectstatic log with the previous app log differs around 5s:

2023-06-26T09:46:04.949 proxy[<machine-id>] sin [info] Starting machine
2023-06-26T09:46:05.139 app[<machine-id>] sin [info] INFO Starting init (commit: 0b28cec)...
2023-06-26T09:46:05.153 app[<machine-id>] sin [info] INFO Preparing to run: `/cnb/process/web` as 1000
2023-06-26T09:46:05.157 app[<machine-id>] sin [info] INFO [fly api proxy] listening at /.fly/api
2023-06-26T09:46:05.167 app[<machine-id>] sin [info] 2023/06/26 09:46:05 listening on [<ipv6>]:22 (DNS: [fdaa::3]:53)
2023-06-26T09:46:05.186 proxy[<machine-id>] sin [info] machine started in 237.351792ms
2023-06-26T09:46:10.617 proxy[<machine-id>] sin [info] waiting for machine to be reachable on 0.0.0.0:8080 (waited 5.430489732s so far)
2023-06-26T09:46:10.764 app[<machine-id>] sin [info] 1421 static files copied to '/workspace/static', 3643 post-processed.
2023-06-26T09:46:11.121 app[<machine-id>] sin [info] [2023-06-26 09:46:11 +0000] [231] [INFO] Starting gunicorn 20.1.0
2023-06-26T09:46:11.121 app[<machine-id>] sin [info] [2023-06-26 09:46:11 +0000] [231] [INFO] Listening at: http://0.0.0.0:8080 (231)
2023-06-26T09:46:11.121 app[<machine-id>] sin [info] [2023-06-26 09:46:11 +0000] [231] [INFO] Using worker: sync
2023-06-26T09:46:11.123 app[<machine-id>] sin [info] [2023-06-26 09:46:11 +0000] [249] [INFO] Booting worker with pid: 249
2023-06-26T09:46:11.618 proxy[<machine-id>] sin [info] machine became reachable in 6.43199164s

Solution

A possible solution is running another process type that runs collecstatic. We could also provide other process types for other Django admin commands in general.

References:

• https://github.com/buildpacks/spec/blob/main/platform.md#launcher
• https://buildpacks.io/docs/app-developer-guide/run-an-app/
• https://github.com/buildpacks/samples/tree/main/buildpacks/hello-processes
• https://buildpacks.io/docs/reference/spec/migration/platform-api-0.3-0.4/#multicall-launcher
• https://fly.io/docs/reference/configuration/#the-processes-section

Some problemset has their own sections.

Example:

• OSP Informatika 2022: B1.1-3 (also has shared problem description)
• IPhO 2021: Q1

This should differ from the shared problem description, mainly in how the problem is numbered and aggregated score points. Both could have a shared description, but we might need to reassess if that is more suitable as a subproblem.

The test still use old problem model.

Currently, the buildpack only accepts requirements.txt on the root folder for pip install. We might need whitenoise only for fly.io deployment, not "production" as a whole.

Related issue: paketo-buildpacks/pip-install#205

Basic problem model.

There are some options for the model name:

• problem (TLX)
• task (IOI, IGeO)
• question

Problemset alternative:

• problemset (TLX, vjudge)
• worksheet
• exam (IBO)
• competition

It seems the most consistent and suitable for now is problemset & problem. Although the OSN problem description on TLX uses the term subtask (just like IOI), subproblem still sounds valid (we wouldn't need that sub- term for now though).

The current implementation gives F405 flake8 error when "child" settings try to modify a variable (ex. modify list/dict).

Alternatives: https://djangopackages.org/grids/g/configuration/

Some problems might share some descriptions (ex: OSK 2017 no. 15-17).

This could also be beneficial when editing the problem description. Instead of modifying each problem, which is prone to inconsistency, the user could modify only the shared description.

For security, we currently only allow some HTML tags (bleach is allowed-list-based), but the editor still uses the default toolbar layout, which might contain tags that are not being allowed. We need to define the allowed tags and configure both CKEditor and bleach with those tags.

References:

• https://www.javadoc.io/doc/com.googlecode.owasp-java-html-sanitizer/owasp-java-html-sanitizer/latest/org/owasp/html/Sanitizers.html
• https://github.com/OWASP/java-html-sanitizer/blob/main/src/main/java/org/owasp/html/HtmlPolicyBuilder.java#L260

From #19

The image URL on kujawab is in the form of a relative URL, which wouldn't work from outside of kujawab site.

Might as well fix other URLs (if there are any).

Automatically fill in the title of the problem form based on the problemset title and the problem number. The format could be something like ": ". It can be set by adding the title URL query param on the "+" add problem link.