Theming #624

Custom CSS for blogs
Custom themes for instance
New dark theme
UI for admin to upload new theme (or delete them)
UI to choose your theme

Instance level themes are just CSS files, stored in static/css (see Instance::list_themes).

TODO:

Federate blog themes
Escape url()/import
Add a way to avoid conflicts between blog and instance themes
UI to disable blog CSS (the option is already in the database, just need to add a checkbox in the settings)
SQlite migrations
Documentation

- Custom CSS for blogs - Custom themes for instance - New dark theme - UI for admin to upload new theme (or delete them) - UI to choose your theme ![image](https://user-images.githubusercontent.com/16254623/59910472-de7f1e00-9409-11e9-9d24-6e369f960608.png) Instance level themes are just CSS files, stored in `static/css` (see `Instance::list_themes`). TODO: - [x] Federate blog themes - [x] Escape url()/import - [x] Add a way to avoid conflicts between blog and instance themes - [x] UI to disable blog CSS (the option is already in the database, just need to add a checkbox in the settings) - [x] SQlite migrations - [x] Documentation Fixes #354, #403

Codecov Report

Merging #624 into master will decrease coverage by 0.32%.
The diff coverage is 12%.

@@           Coverage Diff            @@
##           master   #624      +/-   ##
========================================
- Coverage   35.33%    35%   -0.33%     
========================================
  Files          68     68              
  Lines        7915   7938      +23     
  Branches     1894   1888       -6     
========================================
- Hits         2797   2779      -18     
- Misses       4341   4381      +40     
- Partials      777    778       +1

# [Codecov](https://codecov.io/gh/Plume-org/Plume/pull/624?src=pr&el=h1) Report > Merging [#624](https://codecov.io/gh/Plume-org/Plume/pull/624?src=pr&el=desc) into [master](https://codecov.io/gh/Plume-org/Plume/commit/fb60236a54893d690a8ddba34f597e4820431154?src=pr&el=desc) will **decrease** coverage by `0.32%`. > The diff coverage is `12%`. ```diff @@ Coverage Diff @@ ## master #624 +/- ## ======================================== - Coverage 35.33% 35% -0.33% ======================================== Files 68 68 Lines 7915 7938 +23 Branches 1894 1888 -6 ======================================== - Hits 2797 2779 -18 - Misses 4341 4381 +40 - Partials 777 778 +1 ```

❓

is it still ready for review considering the design flaw pointed regarding arbitrary css security?

No, I'm implementing another approach, I think I will be able to push it this evening.

If you want to review this PR, the new version is there. Admins should now approve remote blog themes, but they may contain anything. Once a theme is approved, it is also available for use on this instance. Admins can also upload themes if they have SSH acces to their instance: they just need to create a folder called "blog-SOMETHING" in static/css, and add the theme files inside.

I will update the docs too.

If you want to review this PR, the new version is there. Admins should now approve remote blog themes, but they may contain anything. Once a theme is approved, it is also available for use on this instance. Admins can also upload themes if they have SSH acces to their instance: they just need to create a folder called "blog-SOMETHING" in `static/css`, and add the theme files inside. I will update the docs too.

Does it really make sense to have admins approve remote blog themes? Using remote resources on the local site seems fragile and prone to many more errors in the future. I think it makes more sense to limit themes to site-wide and local-user.

More generally, this is an issue that has to do with the way articles federate in Plume -- the HTML is delivered and cached to many different sites that may have varying CSS. Federating CSS as well is, I think, out-of-scope of ActivityPub (although there is nothing preventing this). In any case, it is much simpler to say that users should view the original URL if they wish to read the article with the original CSS.

Does it really make sense to have admins approve remote blog themes? Using remote resources on the local site seems fragile and prone to many more errors in the future. I think it makes more sense to limit themes to site-wide and local-user. More generally, this is an issue that has to do with the way articles federate in Plume -- the HTML is delivered and cached to many different sites that may have varying CSS. Federating CSS as well is, I think, out-of-scope of ActivityPub (although there is nothing preventing this). In any case, it is much simpler to say that users should view the original URL if they wish to read the article with the original CSS.

I'm not sure this is an actual concern but how does it play with caching? Where are the file stored and are they given a unique name (hash based?)?

The themes are stored in static/css so they should integrate with the current caching system I think?

The themes are stored in `static/css` so they should integrate with the current caching system I think?

I think they are cached, but not evicted when modified, so client might think the file was not updated when it was. Adding a timestamp, a hash, a random string or anything to the name (myfile-DEADBEEF1312.css) would fix it

I think they are cached, but not evicted when modified, so client might think the file was not updated when it was. Adding a timestamp, a hash, a random string or anything to the name (`myfile-DEADBEEF1312.css`) would fix it

I think it is going to be quite difficult, because in the current implementation, the main theme file should be named theme.css, and it may load other files of the theme, so we would have to find a way to detect the main theme file even if it has a hash in its name, and to edit all url()/@import inside of it to make sure other files are correctly loaded too… Maybe we can just set a default cache duration of one week or something like that, with a HTTP header? Or is it a problem not to have the latest theme update for a few days?

I think it is going to be quite difficult, because in the current implementation, the main theme file should be named `theme.css`, and it may load other files of the theme, so we would have to find a way to detect the main theme file even if it has a hash in its name, and to edit all `url()`/`@import` inside of it to make sure other files are correctly loaded too… Maybe we can just set a default cache duration of one week or something like that, with a HTTP header? Or is it a problem not to have the latest theme update for a few days?

Even 7 days is quiet long and will definitely be reported as a bug at some point. I'm at work currently, maybe we can discuss of it later on riot?

OK, feel free to ping me when you want to talk about it :)

👀

lgtm

 @ -0,0 +1,4 @@
 -- Your SQL goes here
 ALTER TABLE blogs ADD COLUMN theme VARCHAR;
 ALTER TABLE users ADD COLUMN preferred_theme VARCHAR;

 @ -0,0 +1,4 @@
 -- Your SQL goes here
 ALTER TABLE blogs ADD COLUMN theme VARCHAR;
 ALTER TABLE users ADD COLUMN preferred_theme VARCHAR;

 @ -0,0 +1,4 @@
 -- Your SQL goes here
 ALTER TABLE blogs ADD COLUMN theme VARCHAR;
 ALTER TABLE users ADD COLUMN preferred_theme VARCHAR;

 @ -0,0 +1,4 @@
 -- Your SQL goes here
 ALTER TABLE blogs ADD COLUMN theme VARCHAR;
 ALTER TABLE users ADD COLUMN preferred_theme VARCHAR;

 @ -0,0 +83,4 @@
     last_fetched_date,
     fqn,
     summary_html
 FROM users;

ラベル マイルストーン

Theming #624

Codecov Report

レビューア

ステップ 1:

ステップ 2:

ラベル マイルストーン

Theming #624

Codecov Report

レビューア

ステップ 1:

ステップ 2:

ラベルマイルストーン