domainlayer

package
v0.11.1-dev Latest Latest
Warning

This package is not in the latest version of its module.

Go to latest
Published: Jun 17, 2022 License: Apache-2.0 Imports: 1 Imported by: 0

README

Why

We have Jira/TAPD for Project Management, we have Github/Gitlab/BitBucket for Code Hosting, that is, we have multiple Platforms for a certain type of problem. So, how can we calculate metrics across different Platforms?

For example, some users may use Jira as their Project Management platform, the others might opt for TAPD, if we were to implement a Requirement Count metrics for all users, should we implement 2 charts for Jira and TAPD independently? It's too impractical to begin with.

How

Domain Layer is designed to solve the problem by offering a set of Platform Independent Entities, Devlake divides all platforms into three categories: Project Management / Code Hosting / Devops, by abstracting common properties from different platforms, we can define a set of Domain Entities for each category.

The following rules make sure Domain Layer Entities serve its purpose

  1. Every platform specific entity can be mapped (or split) to one (or more) Domain Layer Entity
  2. Every Domain Layer Entity contains enough information for metrics calculation
  3. Domain Layer Entity should contains some sort of pointer to its origin record, and all entities should share a same schema

What

Domain Layer Entity

  • Each Domain Entity has a Id with type string describe its origin record in format <Plugin>:<Entity>:<PK0>:<PK1>, because:
    1. Different platforms might choice different types as their Primary Key, i.e. AutoIncremental Integer or uuid
    2. Platform might or might not use composite primary keys
    3. Primary key might overlay between entities, and multiple entities most likely will be combined into one table
    4. Different plugins might use same entity name, even they can not co-exists at the same time, so plugin name must be included for distinction
    5. This format is deterministic, each of every entity can be converted independently in parallel, and data could be rebuilt arbitrary time with same output, which mean you can truncate any table at any time, and data integrity will be restored on next run. (this is not possible for AutoIncremental Integer or uuid)
  • Each Domain Entity must contains enough fields needed for all metric calculations

Data Conversion

  • Read data from platform specific table, convert and store record into one(or multiple) domain table(s)
  • Generate its own Id accordingly
  • Generate foreign key accordlingly
  • Fields conversion

Sample code:


type Issue struct {
    Id       string  `gorm:"primaryKey"`
    BoardId  string  `gorm:"index"`
    ...
}

issue := Issue {
    Id:         "jira:JiraIssues:1:10",
    BoardId:    "jira:JiraBoard:1:10"
    ...
}

Documentation

Index

Constants

This section is empty.

Variables

This section is empty.

Functions

This section is empty.

Types

type DomainEntity

type DomainEntity struct {
	Id string `json:"id" gorm:"primaryKey;type:varchar(255);comment:This key is generated based on details from the original plugin"` // format: <Plugin>:<Entity>:<PK0>:<PK1>
	common.NoPKModel
}

Directories

Path Synopsis

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL