An Empirical Study of How Developers Use - Results - Discussion - - PowerPoint PPT Presentation

An Empirical Study of How Developers Use Autocompletion

Sheldon Chi and Abdul Khader Naik Outline

• Introduction
• Dataset
• Results
• Discussion

Autocomplete

• Standard in modern IDEs
• MS Visual Studio’s Intellisense in our dataset

Research Questions

• RQ1: What are the common autocompletion targets for developers?
• RQ2: What are the differences in autocomplete targets for developers across

roles?

• RQ3: What is the autocomplete rate of developers?
• RQ4: How does autocomplete rate of developers compare across roles?

Dataset

• MSR 2018
• 11MM events
• Collected using FeedBag++
• Visual Studio
• 81 developers
• 43 professional developers
• 4 researchers
• 5 students
• 6 hobby programmers
• 24 unknown

Data cleaning

224054 - autocompletion events Student: 3 Software Engineer: 36 Unknown: 19 Researcher Academic: 3 Hobby Programmer: 5

Sample data

RQ1: What are the common autocompletion targets for developers

RQ2: What are the differences in autocomplete targets for developers across roles RQ3: What is the autocomplete rate of developers

Autocomplete - 7th most common event. 3.09 % of total events

RQ3: What is the autocomplete rate of developers

0.7% of all events are applied autocompletes events 23.56% of autocompletes are applied the rest are Filtered/cancelled

RQ4: How does autocomplete rate compare across roles

Conclusions

• Threats to validity

○ Small sample set for students (3) researchers (3) - might not be true representation. ○ We are not aware of the nature of projects (frontend or backend) that the developers were working on so might be difficult to generalize

• Future Work

○ Further evaluate context information to get more details about types of methods etc ○ Research can be done towards time taken for developers to apply a code-suggestion and try to optimize it.

Questions?