Ask in the Elixir forum

Currently, the flow of the map.sql for Collector has not been though in advance.

This is the new proposed flow

• Agg Server
• Agg Village
• Agg Player
• Agg Alliance

In The Red Natar, we are using urls as server identifier because Travian doesn't provide them.

Recently, we have discovered that Travian is reusing this urls for new servers, after thinking about it, it's obvious. So we have been mixing servers under the same url.

For fixing that, we have develop the archive system #17 and we have hardcoded a value of 7 days to consider a server finished.

The purpose of this task is to evaluate if this value is enough or we need to lower it. A confirmation by a Travian member via Discord should be enough.

Create an agg_players.c6bert table to store an player's historical data in an incremental way

Define a schema
Set up the test
Write the code

Test scenarios

Create the first agg_players.c6bert
Create the second agg_players.c6bert
Create a n agg_players.c6bert

Use GenStage or something similar for limiting the number of active process collecting data from the servers

Collector.AggServer is the flow for metrics related to the server, number of new players, number of population, etc

It is doing a distinct operation over the whole list of players, very taxing.

Maybe implement a Satellite.ServersTable which is autofilled?

Maybe autofilled first with during medusa process and after with incremental server

Travian create the snapshot after midnight but using the midnight server's time, not a standar one.

We are currently fetching the snapshot assuming a unique offset, which is not correct and can affect the model, and also we are in risk of fetching some snapshot with a huge delay and therefore not useful prediction.

eg: fetching snapshot at 01:00 UTC+0 of a server with UTC-3 will be 22:00 local time, and this info will be outdated at midnight with the new snapshot created by Travian

I am lazy to do it now :), this is just a reminder

Document the api with apidocs, easy and was working before

TTypes.grid_position is not well named, indeed, it is misleading as it is the identifier of the poisition(I think). So, propose a new name an modify it everywhere.

Posible names:

• map_id
• position_id

Collector.GenWorker.Metadata is a GenServer which relies on :travianmap.get_info() for getting server metadata(speed, starting date, ...).

As this is higlely ineficient and very prone to error due to the inestability of web scraping, the metadata is going to be processed in the agg_server flow instead of using :travianmap.get_info().

This issue is for deactivating/removing the code related to the old metadata flow, some relevant files are:

Collector
Collector.GenCollector
Collector.GenWorker.Metadata
Collector.ServerMetadata
Collector.Supervisor.Metadata

After updating to phoenix 1.7.10, some css has been broken

Sort the numeric columns in Medusa's table doesn't work properly.

We need to take a look at https://github.com/fiduswriter/simple-datatables in order to fix it

Create a function under Storage for listing all the servers under the root folder.

Consider to use TTypes.server_id_from_path to return server_ids

This function will be used for doing data reloads

Instead of generating medusa table rows in the html, do it in a json script and then call simple tables, maybe it has better performance

API for TRN

We can process the current tables(to decide which ones) in Imperatoris and push it to Legati via mnesia dets

Too much data that I will probably never use

Travian reuse urls from previous servers. Our storage system was design with that idea that this urls were unique.

We need to define a method for closing the servers, moving them to another folder and rename them(incremental rename) for not erasing previous data.

I propose to close it after 7 seven days without a new fresh map.sql

We need a landing page for the project