Upgrade to Pro — share decks privately, control downloads, hide ads and more …

LINE TECHPULSE 2022 - Data-Driven Testing For LINE SHOPPING

LINE TECHPULSE 2022 - Data-Driven Testing For LINE SHOPPING

Data-Driven Testing For LINE SHOPPING by Winter Hung / LINE SHOPPING @ LINE TECHPULSE 2022 https://techpulse.line.me/

LINE Developers Taiwan

January 21, 2022
Tweet

More Decks by LINE Developers Taiwan

Other Decks in Programming

Transcript

  1. Agenda › LINE SHOPPING Introduction › How to Use Real

    Data Integration Testing › What is Data-Driven Testing › Which Benefits are Data-Driven Testing › Conclusion 
  2. LINE SHOPPING Different Types of Costs Transaction fee COST PER

    SALE CPS  Number of clicks COST PER CLICK CPC Effective activities COST PER ACTION CPA Assist partners COST PER LEAD CPL
  3. › Personalized recommendation › User has favorites store › Have

    Business browsing history › Start sending schedule every day Official Account Push 
  4. Data-Driven Decision Linear Regression in Machine Learning Let’s predict Initializing

    the regression model Visualize the training and testing set Importing the dataset 
  5. Data-Driven Decision Linear Regression in Machine Learning Test Set The

    testing set should be 5% to 30% of the dataset Training Set Our training model which will implement the Linear Regression. 
  6. LINE SHOPPING SEARCH SYSTEM ARCHITECTURE  External Service 3FBM%BUB 3FBM%BUB

    3FBM%BUB %BUB 'FFET Data Operation LINE Commerce Platform MongoDB OBS 1SPEVDU  $BUFHPSZ LINE SHOPPING SEARCH NAVER 4FBSDI  3FDPNNFOEBUJPO 
  
 4JNJMBS
  7. LINE Commerce Platform  Application Layer LINE Shopping Front-APP LINE

    SHOPPING My-APP Domain Layer Product Domain Seller Domain Order Domain LCP
  8. LINE SHOPPING SEARCH MOCK SERVER  MOCK SERVER 3FBM%BUB 5FTU%BUB

    %BUB 'FFET Data Operation LINE Commerce Platform MongoDB OBS 1SPEVDU  $BUFHPSZ LINE SHOPPING SEARCH NAVER 4FBSDI  3FDPNNFOEBUJPO 
  
 4JNJMBS
  9. OBS Open Broadcast Services OBS is a media platform for

    LINE and LINE-family services, The platform comes with object storage by default Image presentation system › Creating a thumbnail › Transcoding videos and audiosArial › Analyzing media using advanced VISION-AI technology (OCR, Adult image filter, object detection) 
  10. Mock-Server Easy Mocking Any System Easily recreate all types of

    responses Start working against a service API Isolate the system-under-test to ensure tests 
  11. GORM Golang Object Relational Mapping Distributed applications of the GORM

    ORM in computer science is a programming technique for converting data between incompatible type systems in OOP languages. This creates, in effect, a “virtual object database” that can be used from within the programming language.LINE SHOPPING needs to verify data and consistent tests in MySQL 
  12. OVERVIEW GIN + GORM + MYSQL  MYSQL $MJFOU (*/

    (03. $MJFOU $MJFOU RESTFUL WEB SERVICE
  13. GIN Golang Web Framework › Excellent performance › Native net/http

    package › Fast httprouter › Well-designed middleware › Excellent data binding 
  14. OVERVIEW Code Folder Structure (03. .PEFMT $PO fi H $POUSPMMFST

    )FMQFST 3PVUFST Schema.go Model.go Database.go Controller.go Response.go Routers.go
  15. VOS Verda Object Storage VOS (Verda Object Storage) is an

    object storage service with comes with an S3 compatible Object Storage API. The S3 API is a de facto standard object storage API for AWS S3, which means we can utilize open-source software based on S3 API for VOS without changing our source code. We can use the S3 SDK to manage VOS objects. 
  16. TIME Data verify of spend time more than 4x faster

     MYSQL GORM & Jenkins SERVER MYSQL WORK BENCH SERVER 1 min 15 sec
  17. 1 2 3 4 5 6 7 8 9 10

    11 0 350 700 1050 1400 Partner A Partner B Partner C Partner D Partner E Partner F Orders Count CPS REPORT  Month FAIL ORDERS COUNT
  18. Coverage Health Check of Search AVG New Stores 102 AVG

    GMV Rate 15% AVG Fail Rate 0.16 
  19. 0 75000 150000 225000 300000 5 6 7 8 9

    10 Partner A Partner B Partner C Partner D Partner E Partner F Order Counts TOP6  Month
  20. 3% 20% 23% 5% 29% 20% Partner A Partner B

    Partner C Partner D Partner E Partner F 8% 9% 11% 12% 14% 46% Partner A Partner B Partner C Partner D Partner E Partner F 80 / 20 DDT Observe  80% 20%
  21. Conclusion › It is an ideal option to use realistic

    information › It allows multiple sets of data values during Health Check › Data-driven is a test automation methodology  › 80/20 is very suitable for quality control