Hacker News Reader

Comment by th0ma5 | Hacker News Reader

I still don't know who Shiny is for... For very small organizations? You'd be a fool to run Shiny in a place over even just a couple hundred people... I've never seen any kind of seriousness around denial of service, pen testing, end to end data security, etc. If you're smart enough to be using Shiny to make something really great, you should also be smart enough to understand the risks of web applications. And I could also add with Shiny are you providing analysis or an illusion of an analysis platform with a risk of misinterpretation?

The best fit seems to be demoing work during a presentation or similar with it running locally... Where you can move some sliders and discuss some scenarios. Anything more than that and Shiny seems to promise you a turn key solution to weakly understood risks.

View on HN

In my shop regularly develop and deploy shiny apps for 'micro-tasks' that are scripted under the hood, for providing a project management GUI, control tasking and work tracking, etc.

All of our apps run locally, only. Since we go through many 'small' contracts, and each one can be pretty different. Retooling a shiny app to do another job so that the PM's can keep track and manage their swarm takes a few hours, but frankly, it keeps the nerf's herded. The overhead of spinning off a new shiny app is so low, its pretty nice.

It's a web application framework for R, not a turnkey solution or an analysis platform. It's for R users who want an easy way to build a web app in R without having to know much about web development or other languages.

If you want to productionize your web app, you could put it behind a reverse proxy, add authentication, etc. just like with any other web framework. The turnkey solution would be products that do this, like Shiny Server and ShinyProxy.

> The best fit seems to be demoing work during a presentation or similar with it running locally... Where you can move some sliders and discuss some scenarios.

Yeah, you got it. It’s basically good for prototyping, or throwing some light interaction onto a plot, etc. My audience is typically either myself, or maybe putting something light weight into a colleague’s hands.

It's popular with scientists (genomics) for making web portals to interactively share data / plots / analyses.

The number of users to such sites is (often) measured in dozens of people per day.

I'm a bioinformatician, not a web developer, and Shiny has been a great tool to share my work with something I'm already familiar with (R).

(I'd share a site I've made with Shiny but I'm pretty certain it'd explode with hundreds of concurrent visitors - but that's OK! I get dozens of visitors a day and don't expect any more)

I think there can definitely be a niche where Shiny fits.

For example, most web frameworks are not necessarily designed to share data sets and related analysis among many users. As data becomes more and more valuable you can expect Shiny to play an important role in such data sharing platforms.

Currently this is mostly seen within large institutions.

> For very small organizations?

For small parts of large organisations. I’ve only ever seen it deployed internally, and the clients/users are usually a small subset of the organisation.

A few use cases I've personally experienced:

1. Demo during presentation on a local server (as you suggested)

2. Demo during presentation on someone else's computer without R installed

3. Quick-and-dirty dashboard for non-technical colleagues, hosted on an on-prem server

4. Simple GUI interface to some kind of R functionality (as seen here)

It's definitely not some kind of universal tool. But it comes in handy and I'm grateful that it exists.

I don't quite get why someone would be a fool to run Shiny. Can you elaborate on that?

Fool is a tough word. IMHO it's terrific how quickly you can go from an idea to a local dashboard, but on the other hand it's non-trivial to get it from the one-person dashboard to a >1000 user production application with some uptime guarantees, authorization, metrics, error collection.

What about ShinyProxy for this purpose (“productionizing”)? https://www.shinyproxy.io/

This is one potential solution. Good read on scaling shiny: https://appsilon.com/alternatives-to-scaling-shiny/

Genomics. Massive pharma companies use R and use RStudio Connect for applet platforms. Python needs a version of Shiny that works with Django.

Reply on news.ycombinator.com