search

crunzphp / crunz

A PHP-based job scheduler
MIT License
176 stars 16 forks source link

Event Unique Key implementation #80

Open lucatacconi opened 6 months ago

lucatacconi commented 6 months ago

In the PR there is a possible solution to calculate a constant unique identifier for events.

This solution is linked to issue https://github.com/crunzphp/crunz/issues/33 and could be the first piece towards creating a launcher that allows you to launch a task based on the unique identifier.

lucatacconi commented 6 months ago

PR sent again as requested.

PabloKowalczyk commented 6 months ago

I have reviewed implementation and it is quite good, but I'm wondering is it address issues in #33? Ids are still "not-stable", maybe it will better to allow users to set their own id?

lucatacconi commented 6 months ago

Personalized Id could be a good idea but i think it could also lead to errors or confusion.

When you have several people entering tasks for the same company or group you may end up using the same custom ID for different tasks. At that point, by introducing a launcher per custom task ID, which one will you launch? At that point you would have to have a system that checks if there are tasks with the same custom ID and in my opinion it would become a mess.

In our company we use crunz to manage automatic processes. There are currently more than 200 tasks configured. Handling uniqueness on the personalized ID would be a problem.

But in the end you are the project leader... so I leave the choice up to you :) hihi..

PabloKowalczyk commented 6 months ago

When you have several people entering tasks for the same company or group you may end up using the same custom ID for different tasks. At that point, by introducing a launcher per custom task ID, which one will you launch?

Here is the thing, at "collecting tasks" part should be check that all users' ids (alnum BTW) are unique, if not throws exception. What do You think?

Also, please note that \md5($this->sourceFile . $this->description . $this->expression); can also generate same id for different tasks (Event), what to do then? :)

lucatacconi commented 6 months ago

Regarding the first point, the solution you propose could be valid. The only perplexity is that we can generate new tasks with the vendor/bin/crunz make:task function but also with a vi/nano or by moving them to the folder with an sftp. In the first case the exception could easily be inserted into the process of adding a new task. In other cases where would you put the exception? It cannot be placed in the engine that executes the tasks because you would also block tasks that are not affected by the problem and would not start.

As regards the second point, when generating the Unique ID I tried to use as much information as possible to uniquely locate the file. As you can see I used the path of the task file inside the file folder, the description and the cron expression. To have a unique key you must insert two tasks with the same description and same execution time into the same tasks.php file. If you do that you're a really evil person hihihi

In that case they would have two identical IDs and it would actually be a problem.

However, we could add in the statement \md5($this->sourceFile . $this->description . $this->expression); also the event sequence within a tasks.php file. In that case it will certainly be unique: with a tasks.php file with a single event inside the progressive number will always be 0 but the $this->sourceFile will be unique if nothing else. in case of multiple events within the Tasks.php file the progressive will guarantee uniqueness. This method could be cool.

However, how do you prefer development to proceed? Unique identifier indicated or calculated?

PabloKowalczyk commented 6 months ago

Regarding the first point, the solution you propose could be valid. The only perplexity is that we can generate new tasks with the vendor/bin/crunz make:task function but also with a vi/nano or by moving them to the folder with an sftp. In the first case the exception could easily be inserted into the process of adding a new task. In other cases where would you put the exception? It cannot be placed in the engine that executes the tasks because you would also block tasks that are not affected by the problem and would not start.

In defensive programming any problem should fail fast and loud so I don't see a problem in throwing exception when collecting tasks, Crunz can't proceed with duplicated task'id.

However, how do you prefer development to proceed? Unique identifier indicated or calculated?

Indicated, and if not then generated.