# 【真的假的】開發筆記

## System structure

[<img src="https://docs.google.com/drawings/d/1VfooYFnEb_mVx9Yso_u9evYD-nq-Y-mDSwsIlg7oDlw/pub?w=846&amp;h=467">](https://docs.google.com/drawings/d/1VfooYFnEb_mVx9Yso_u9evYD-nq-Y-mDSwsIlg7oDlw/edit)

[](https://sketch.cloud/s/Jmy0)

### Server structure

先通通塞在同一台裝著 docker 的 VM 裡頭。下面每一項都是一個 docker container，[docker-compose.yml 放在 rumors-deploy](https://github.com/MrOrz/rumors-deploy/blob/master/docker-compose.yml)：

* nginx，config 另外 scp 上去，預計將 rumors-api.hacktabl.org, rumors.hacktabl.org 做 reverse proxy 到不同 port 上。
* elasticsearch （in docker），[config](https://github.com/MrOrz/rumors-db/) 另外 mount。在 server 上直接跑 seed script 來塞資料。
* [rumors-api](https://github.com/MrOrz/rumors-api) (in docker)
* [rumors-site](https://github.com/MrOrz/rumors-site) (in docker)

## [rumors-db](https://github.com/MrOrz/rumors-db/)

Stores collected rumor documents (`articles`), mythbusting articles (`replies`) and suggested response.

### TODO

* [x] Schema (mapping)
* [x] Get it up and running in production

### Indexes

See [schema](https://github.com/MrOrz/rumors-db/tree/master/schema) for elastic search mappings, and [GraphQL models](https://github.com/MrOrz/rumors-api/tree/master/src/graphql/models) for how data is represented in api.

#### User

* authentication fields like google id and facebook id.

#### Article

has and belongs to many Replies.

Future plans:

* image features (or perceptual hash [pHash](http://www.phash.org/) + [locality sensitive hashing](http://stackoverflow.com/questions/32785803/similar-image-search-by-phash-distance-in-elasticsearch) for efficient matching in elastic search)
* file url (for images)

## [rumor-search.g0v.ronny.tw](https://rumor-search.g0v.ronny.tw/)

Stores mythbusting articles crawled by the crawlers.

1. Given message text, returns search results.
2. Stores parsed data from our data source (see 資料來源 below)
3. API: https://rumor-search.g0v.ronny.tw/

Will be replaced by [crawlers](https://github.com/MrOrz/rumors-crawlers).

### TODO

* [ ] Merge with rumors-db
* [ ] Get the crawlers source

### Notes

#### Stopword list
http://www.cnblogs.com/ibook360/archive/2011/11/23/2260397.html

#### Wrapper from rumor-search

Schema
link
title
rumor
answer

* title
* rumor: 謠言原文
* 澄清文
* link: 澄清 URL

#### Parser
:octocat:

#### 資料來源

請見[提案「使用資料」的部分](http://beta.hackfoldr.org/rumors/-LINE-bot-Zb4bFLB7pnv)。

## [rumors-api](https://github.com/MrOrz/rumors-api/)

Server: http://api.rumors.hacktabl.org

## [rumors-line-bot](https://github.com/MrOrz/rumors-line-bot)

Ruby LINE bot server

:octocat: https://github.com/MrOrz/rumors-line-bot
（開權限請洽 slack ID: MrOrz）

TODO: Implement this user flow, avoid showing users only 1 reply.

[](https://docs.google.com/drawings/d/1kStWPVHbzrpSTKu9-EK5TN52Xj7WY7uAiEBa00v6xK4/edit)

## [rumors-site](https://github.com/MrOrz/rumors-site)
Site: http://rumors.hacktabl.org/

Web UI that enables all users to crowd-source the internet rumors & their mythbusting articles.

### Requirements

The website in general should satisfy:

1. Search-engine friendly. Each rumor & answer should have an individual page that can be indexed by search engines.
2. Lightning fast on mobile -- AMP (accelerated mobile page) compatible -- after all, most of its users come from LINE!
3. Mobile first, Responsive Web Design - See 2.
4. Attract more editors to answer rumors
5. Coach users to write qualified answer

### Landing Page

* Search bar for pasting rumor text & search for answers
* List of rumors that is not answered yet. Sorted by number of user reports.

### Rumor listing page

(To list rumor search results or list similar rumors)

#### Rumor info to be showed

* content (truncated, might be around 100 words)
* Rumor link (if it has)
* created date
* whether answered or not
* answer (need to discuss)
* content
* editor
* reference link

### Rumor detail page (= answer submission page)

* Includes Rumor, Answer, and other meta data including time.
* Includes change history.
* Shows other answers to the same / similar rumors if applicable. Don't allow users to submit their answer if their answer is too similar to existing answer. Encourage reuse of answers for similar rumors.
* Hints the user about filling up references. Don't allow users to submit their answer before filling up references.