Pipes by Example Nick Partridge. nkpart on most things. Why Pipes - - PowerPoint PPT Presentation

▶

Oct 07, 2023 95 likes •330 views

Pipes by Example Nick Partridge. nkpart on most things. Why Pipes Because youre writing a web server, of course. Not just for streaming! produce :: m a consume :: a -> m () Pipe Types Lead, PVC, etc. Producers + Consumers Pipes Proxy

SLIDE 1

Pipes by Example

Nick Partridge. nkpart on most things.

SLIDE 2

Why Pipes

Because you’re writing a web server, of course.

SLIDE 3

Not just for streaming!

produce :: m a consume :: a -> m ()

SLIDE 4

Pipe Types

Lead, PVC, etc.

SLIDE 5

Producers Consumers

+

Pipes

Proxy a a’ b b’ m c

SLIDE 6

Pipe a b m r

a b m On exhaustion: r

Proxy () a () b m r

SLIDE 7

Producer b m r

_ b m On exhaustion: r

Proxy X () () b m r

SLIDE 8

Consumer a m r

a _ m On exhaustion: r

Proxy () a () X m r

SLIDE 9

The Primitives

Because live coding always works.

SLIDE 10

Use the Monad

Running producers sequentially is just >> When pipes/consumers/producers/finish, they return a value Control.Monad and friends are very useful.

SLIDE 11

Elimination

Finally some Category Theory.

SLIDE 12

Pipe a b m r

a b m On exhaustion: r

SLIDE 13

Effect m r

m On exhaustion: r _ _

Proxy X () () X m r

SLIDE 14

Substitution

Producers: for or ~> eliminates yield Consumer side: >~ eliminates await

SLIDE 15

Elimination - Loop Bodies

for generator body

for x yield == x

for (yield x) f == f x

SLIDE 16

Companion Libraries

Where the features are hiding.

SLIDE 17

Companion Libraries

pipes-parse pipes-safe pipes-group

SLIDE 18

1. Pipes Parse

type Parser a m r = forall x. StateT (Producer a m x) m r Provides ‘push-back’. You can modify the underlying producer to return a value again.

SLIDE 19

2. Pipes Safe

ReaderT (IORef (Finalizers m)) m r Resource management, exception safety, through a monad transformer on m that allows registering Finalizers.

SLIDE 20

3. Pipes Group

FreeT (Producer a m) m x Working on sub-streams in constant memory. a ‘linked-list’ of sub-stream producers, each one returning the next sub-stream on exhaustion

SLIDE 21

return () :: Effect Talk ()

SLIDE 22

Pipes by Example

Nick Partridge. nkpart on most things.

Why Pipes

Because you’re writing a web server, of course.

Not just for streaming!

produce :: m a consume :: a -> m ()

Pipe Types

Lead, PVC, etc.

Producers Consumers

+

Pipes

Proxy a a’ b b’ m c

Pipe a b m r

a b m On exhaustion: r

Proxy () a () b m r

Producer b m r

_ b m On exhaustion: r

Proxy X () () b m r

Consumer a m r

a _ m On exhaustion: r

Proxy () a () X m r

The Primitives

Because live coding always works.

Use the Monad

Running producers sequentially is just >> When pipes/consumers/producers/finish, they return a value Control.Monad and friends are very useful.

Elimination

Finally some Category Theory.

Pipe a b m r

a b m On exhaustion: r

Effect m r

m On exhaustion: r _ _

Proxy X () () X m r

Substitution

Elimination - Loop Bodies

for generator body

for (yield x) f == f x

Companion Libraries

Where the features are hiding.

Companion Libraries

pipes-parse pipes-safe pipes-group

type Parser a m r = forall x. StateT (Producer a m x) m r Provides ‘push-back’. You can modify the underlying producer to return a value again.

ReaderT (IORef (Finalizers m)) m r Resource management, exception safety, through a monad transformer on m that allows registering Finalizers.

FreeT (Producer a m) m x Working on sub-streams in constant memory. a ‘linked-list’ of sub-stream producers, each one returning the next sub-stream on exhaustion

return () :: Effect Talk ()

References

Pipes haddocks - Pipes, Pipes.Prelude, Pipes.Tutorial Haskell for All - Gabriel Gonzales’ blog Haskell Pipes Google Group “pipes-“ seach on hackage