coderplanets / coderplanets_server Goto Github PK

when query read=true, i can use:

read == true ->
   queryable
   |> join(:inner, [content, f, c], viewers in assoc(content, :viewers))
   |> where([content, f, c, viewers], viewers.user_id == ^user.id)

but came to read=false, this is not work as except

read == true ->
   queryable
   |> join(:inner, [content, f, c], viewers in assoc(content, :viewers))
   |> where([content, f, c, viewers], viewers.user_id != ^user.id)

because the viewers assoc only have records of viewed user, the unviewed user is not in that table
left_join is not working either, need help...

like react-gh-like-diff

https://twitter.com/i/status/1091198484617678848

handle situation like

• x-user lock/unlock conversion(comments) of a article when the problem solved or other reason
• article edit history with diff UI, just like the git system
• etc ..

a possible inner logger table for post would be like this:

post_loggers:

reason:

1. more readable
2. more specific feedback for test
3. useful in front-end jadgement

it will got 500 when reply text length > 256 chars

some code like

field :paged_posts, :paged_posts do
    arg(:filter, non_null(:paged_article_filter))
    
    middleware(M.PageSizeProof)
    resolve(&Resolvers.CMS.paged_posts/3)
    middleware(M.FormatPagination)
end

some file like
page_post_test

has lots of boilerplate code with job, video, repo ... should use macro to reduce it

see https://github.com/coderplanets/coderplanets_web/issues/354

details:

1. add field to every community table which tracks last content published timestamp
2. add json-field to login user's C11N subscribed communities which tracks every community's last read timestamp
3. the whole logic to frontend should abstract to a 'has_new_content' field on each subscribed-community
4. the whole logic should done by MIDDLEWARE

should have robot system to do the chores, like(not limit to):

1. set a cron to scan every month's posts/jobs/..., once the favorited count > threshold value, then recommend it to the home community
2. block user / post / .. 's recover
3. radar scan -- use tools like RSSHub etc to aggregate cool staff for ever communities
4. ...

TODO

to avoid massive trash / unrelated content(post, job, ... etc) publish in short time.
let's set a time limit in a row of publish articles.

details

add a table to statistics module to log the user publish behavior , then use a middleware to diff time between now and latest publish time.

1. as mentioned in Absinthe doc

Due to varying limits on the maximum size of URLs, we recommend using one of the POST options below instead, putting the query into the body of the request.

2. phoenix 1.5+ requires plug > 1.10, which treat all the GET parameter values as strings, it will break the workflow, see elixir-plug/plug#1006

article-like content will freeze after a day / week, if owner want to update after the limit date, the updated content can only go to the attachment field which will display "update/更新" on the front-end page.

when a article-like content in follow situations:

• the problem mentioned in article is solved
• has published very long time, the content has no value
• has no point, but not to the delete level
• etc

we should set a 🔒 flag on the content, and no edit or comments will be made/append

a possible prototype may look like this:

field(:lock, :string)
belongs_to(:lock_operator, User)

and if front-end want the lock history, should check the inner-logger, see: #104

currently the pin contents execs 2 query on the same contents table:

  def paged_contents(queryable, filter) do
    queryable
    |> flag_query(filter)
    |> ORM.find_all(filter)
    |> add_pin_contents_ifneed(queryable, filter)
  end

  defp flag_query(queryable, filter, flag \\ %{}) do
    flag = %{pin: false, trash: false} |> Map.merge(flag)

    # NOTE: this case judge is used for test case
    case filter |> Map.has_key?(:community) do
      true ->
        queryable
        |> join(:inner, [q], f in assoc(q, :community_flags))
        |> where([q, f], f.pin == ^flag.pin and f.trash == ^flag.trash)
        |> join(:inner, [q, f], c in assoc(f, :community))
        |> where([q, f, c], c.raw == ^filter.community)

      false ->
        queryable
    end
  end

  defp add_pin_contents_ifneed(contents, queryable, filter) do
    with {:ok, normal_contents} <- contents,
         true <- Map.has_key?(filter, :community),
         true <- 1 == Map.get(normal_contents, :page_number) do
      {:ok, pined_content} =
        queryable
        |> flag_query(filter, %{pin: true})
        |> ORM.find_all(filter)

      # TODO: add hot post pin/trash state ?
      # don't by flag_changeset, dataloader make things complex
      concat_contents(pined_content, normal_contents)
    else
      _error ->
        contents
    end
  end

when the contents table growing big, the "add_pin_contents_ifneed" function will query all the contents only for few pined contents, this's may have performance issue.

i think the better practice is: we split pined contents into a new separate table, then combine the small pined query results to all the results.

best practice

example

reference:

hexdoc typespeccs
specifications-types-elixir
elixir school
official doc

realworld example:

code-corps-api
nested_filter
tesla

Dialyzer

credo vs Dialyzer
use dialyxir in elixir
phoenix quick start

credo

reference

Currently the xxx-users loader is not using Dataloader , because the limit and offset in the query will affect the whole result set and not each association, this can be solved by a "Top N per Group" case, but the syntax is ugly ...

reference links:

https://elixirforum.com/t/preloading-top-comments-for-posts-in-ecto/1052/8
elixir-ecto/ecto#2281
https://spin.atomicobject.com/2016/03/12/select-top-n-per-group-postgresql/
https://stackoverflow.com/questions/40529699/how-to-select-id-with-max-date-group-by-category-in-ecto-query-with-phoenix
http://www.achraf-sallemi.com/select-top-n-per-group-in-postgresql/

1. user can have multi education field
2. user can have multi company field
3. ..

check wechat pay in elixir

alipay in elixir
alipay doc

1. user can create / remove categories
2. user can set / unset categories to content

thoughts

1. only advance account can set tip QR info
2. QR code(if set) should appear in user-page / article-bottom

use customize scrivener to remove FormatPagination middleware

thought

first of all: video should be links, i am poor ..

1. currently most of the videos should be youtubes
2. video should have post, duration, desc, source, source_link, published_at,video_author but NOT playtimes
3. should have category, playlist concept
4. only community editors can CURD videos

when create comment to a Article, the Article's updated_at field should be updated together, this is useful when use filter like latest-update to the articles list, which is default filter

see absinthe issue

1. 在 article 类资源上增加 attachments 字段，包含该 article 里面的 图片等连接信息
2. 删除资源的同时应该同时删除放在OSS 上的 attachments 资源
3. 目前 attachments 由前段解析提供，后期应该在后端再做一次校验
4. 同3， Elixir 里的正则表达式比较蛋疼。。

details

(like / star / comment) by sb => reputation +1
(favorite) by sb => reputation +2

currently delivery context uses cms's passport system, should make one for delivery.

thoughts

display recent messages/activity in twitter/weibo/youtube ... etc

tools

RssHub
Elixir-rss
fiet
elixir-scrape

see coderplanets/coderplanets_web#990

drop apollo schema for now

# - apollo schema:check  --endpoint="http://apiraw.coderplanets.com/graphiql"
# - apollo schema:publish --endpoint="http://apiraw.coderplanets.com/graphiql"

since i register groupher.com domain, it's time to change the project namespace to Groupher.

phoenix 1.4 came with better http/2.0 support, and it's upgrade guide seems has no breaking change.

see sikanhe/apollo-tracing-elixir#23

e.g :

1. last vist time statistic ...
2. log ...
3. send emails ..

seems rihanna is the best tool, not need redis
and most powerful pkg: exq