Skip to content

Latest commit

 

History

History
97 lines (77 loc) · 4.07 KB

sw.org

File metadata and controls

97 lines (77 loc) · 4.07 KB
Raw

Streaming All The Things

Introduction

Programming with streams is also called functional reactive programming (FRP). It is one (of many) ways of handling state in a way that is compatible with referential transparency; which is basically the isolation of side effects, or making sure changes to program state happen in a way that is not noticed by different sections of the code.

Streams make forefront the concerns of functional (immutability) and reactive (inversion of control) aspects of programming.

When working with Streams, one is able to describe time itself in the code you write. There are two ways to model time in FRP, either discretely (in steps) or continuously.

In a discrete view $(t_1, v_1) → (t_2, v_2) \dashrightarrow (t_n, v_n)$ you can think of a stream as a endless list of values that happen at arbitrary moments in time: $v_1, v_2, \ldots v_n$.

Explanation

The key insight is that you can map changes of a property though time as a list of these values as a way to regain composability around event streams.

There are then two distinct concepts that arise (we will use the BaconJS terminology from now on) Properties and EventStreams.

In a property, you can always ask: What is property's value now? or a property is expected to hold a value at any given moment in time.

An eventStreams on the other hand does not hold on to values across time, they are ephemeral and only send out signals (or events) as they are triggered.

Key combinators arise:

  • filter : Only consider (emit) values if they match the filter predicate
  • map : Transform all values applying a given function to each value individually.
  • take : Close a stream after taking the first n elements from it.
  • concat : Concatenate a second stream right at the end of a first one.
  • skipUntil : Ignore (do not emit) values until they match a given predicate; then continue unconditionally.
  • debounce : Only emit the last value from a noisy bunch.
  • concatMap : flatten / unroll internal streams via the concat function.
  • mergeMap : flatten / unroll internal streams via the merge function.
  • switchMap : flatten / unroll internal streams by canceling on every new value emission to the parent.

Examples

Starting point: https://jsfiddle.net/kryptt/y5fta1nc/31/

  1. Switch and light.
  2. Ensure the bulb shuts down after 2 seconds of inactivity.
  3. Turn the light on when the mouse enters the box and off when it leaves it.

Exercises

Pick one or more of:

  1. Write a login/logout module that holds the current user as a Property. Store logins and passwords for validation in memory.
  2. Write a drawing application over a HTML5 Canvas. Store the current color and brush width in a property.
  3. Write a stock market tracker.

You can look at http://dangorst.github.io/frp-with-bacon/ for inspiration.

Results

We held a workshop introducing Functional Reactive Programming or programming with Streams during our Face 2 Face session the third week of January 2020. The workshop was made in reference to improving the locality of business logic for User Interface code at IGT. The subject was selected unanimously amongst polled individuals and was fully booked with 18 participants.

We had a very fun session for which we received positive feedback. If you are interested in having this workshop with us please email David Terol. Thanks again Daan van der Munnik for helping us coordinate and prepare the first version of this workshop as well as for your feedback to improve it.