Search playground to improve adoption of advanced search features

Created by: sourcegraph-bot

RFC

GitHub project

Motivation

One of our top priorities this year is to make Sourcegraph awesome for individual sourcegraph.com users. Indexing and relevance work (increasing the core product value) will play a big part. However, we have observed in UX interviews that users were also blocked on discovering the value of Sourcegraph, and struggling to learn the way search works. A common denominator seems to be that users think Sourcegraph is semantic search. There are two competing approaches to solve this:

• change the search fundamentals

• make sure users understand what code search is, and help them learn it

We focus on the later as a first approach.

User Problem

We have added a lot to the search query language, but users can't autonomously discover and explore use cases driven by features such as structural search, search expressions, search predicates, search rules. The interactive search tour is a great introduction to the basic search UI flow, but doesn't scale to explaining how the features of our query language solve specific use cases. There is extensive validation in user feedback on the desire for more examples of search queries in the product. 

Proposal

We propose solving this by introducing a search playground/sandbox, containing useful examples, with comments explaining the use case they solve, running against a known set of repositories (so that they show results immediately). There is successful prior art in RegExr example patterns, or the TypeScript playground. In many ways, this amounts to moving documentation into the product.

Measuring success

We will measure success by the retention & power user curve of cohorts that were exposed to the search playground.