dry-rb/chat

WE MOVED TO ZULIP => https://dry-rb.zulipchat.com <=

Pablo Herrero

@pabloh

When using dry-transaction, Is possible to modify the context within a step?

Nikita Shilnikov

@flash-gordon

@pabloh wdym by saying "context"? A transaction just accepts arguments and returns a result, it doesn't have any mutable internal state

Pablo Herrero

@pabloh

Doesn't it keep a context along the way where you can grab values?

Is not possible to store partial results somewhere there to use on subsequent steps?

Nikita Shilnikov

@flash-gordon

you just pass them from one step to another

I use keyword args for passing several values at once

def my_step(value_1: , **rest)
  value_1_transformed = do_something(value_1)
  Right(value_1: value_1_transformed, **rest)
end

@pabloh like this^

Pablo Herrero

@pabloh

So you need to make your result value a hash?

Nikita Shilnikov

@flash-gordon

I use hashes, but it's up to you. You also can define a custom step http://dry-rb.org/gems/dry-transaction/custom-step-adapters/, but this is advanced usage :)

Pablo Herrero

@pabloh

My question is do you have to put a hash besides the actual value in the result to have some kind of context?

Nikita Shilnikov

@flash-gordon

can be arrays also, or you can define your own "context" object

the main point is that dry-transaction doesn't have a mutable state inside

Pablo Herrero

@pabloh

I understand that, I meant "change" the context in the same way the state monad updates its "state".

Nikita Shilnikov

@flash-gordon

yeah, in this sense a new type of step will be useful, atm we don't have any special "context" object

txn = define_transaction do
  custom_step_adapter 'step_1', using: :value_1
  custom_step_adapter 'step_2', using: :value_2
  map 'step_3'
end

txn.call(value_1: 1, value_2: 2)


def my_step(value_1)
  value_1 + 1
end

def my_step(value_2)
  value_2 + 1
end

def step_3(context)
  context # { value_1: 2, value_2: 3 } 
end

@pabloh for example^

Nick Sutterer

@apotonick

@pabloh That's basically and exactly what a TRB operation provides (sorry for high-jacking)

Pablo Herrero

@pabloh

Ok, will check it out.

Nick Sutterer

@apotonick

@pabloh There a "context" object, we call it options which is basically a hash, and passed from step to step (per default) http://trailblazer.to/gems/operation/2.0/#state-and-result

Pablo Herrero

@pabloh

So you just store the actual result in :model as a convention?

Pablo Herrero

@pabloh

@apotonick, any reason yoy prefered that approach instead of storing the result separately?

Nick Sutterer

@apotonick

@pabloh The result is stored separately! The options hash scopes, and knows who added what

However, most operations don't need that, so the "human" interface is very simple and looks "global", but it isn't. It took months to design it

Pablo Herrero

@pabloh

I see, I think I could take a dive in the code in order to fully get it after I check out the docs some more.

George Millo

@georgemillo

hey, when I clone the dry-types repo and run specs, I get failures for all tests that involve a maybe type

I imagine that this is something to do with dry-monads not being loaded?

how do I make the tests pass?

Nikita Shilnikov

@flash-gordon

@georgemillo run with rake

George Millo

@georgemillo

aha

thanks

looks like rake runs two rspec suites

what's going on there?

Nikita Shilnikov

@flash-gordon

with/without monads ext

George Millo

@georgemillo

oh wait, I answered my own question by looking in the Rakefile

funny how that happens

Piotr Solnica

@solnic

@flash-gordon I’d like to revisit hash schema implementation in dry-types, and try to figure out a more flexible way for setting them up. This was brought up before by @Kukunin I believe, I was initially against it, but we start to see more use cases where a more flexible solution would be very useful. /cc @backus

ie now we need a symbolized hash which has strict semantics too, and we’re stuck :)

this is definitely something to figure out before 1.0.0

if we can come up with a flexible solution which is also fast, that would be awesome. Oh and the lib should still provide a bunch of common, built-in hash schema types

@apotonick I’ll let you know once I have something working with the new dry-v + dry-schema, and re rom, I reported an issue about reform/rom integration so we can discuss things there I guess

Sergey Kukunin

@Kukunin

@flash-gordon I have a crash on latest master of dry-struct. on the first attribute call.

equalizer.instance_variable_get('@keys').concat(new_schema.keys)

NoMethodError: undefined method `concat' for nil:NilClass
  /Users/kukunin/.rbenv/versions/2.3.1/lib/ruby/gems/2.3.0/bundler/gems/dry-struct-fbe71e6445a8/lib/dry/struct/class_interface.rb:78:in `attributes'
  /Users/kukunin/.rbenv/versions/2.3.1/lib/ruby/gems/2.3.0/bundler/gems/dry-struct-fbe71e6445a8/lib/dry/struct/class_interface.rb:50:in `attribute'
  /Users/kukunin/Sites/matic-www/gems/shift-domain/lib/shift/domain/attachment.rb:8:in `block in <module:Domain>'

Nikita Shilnikov

@flash-gordon

@Kukunin I think you need dry-core from master, check it out

Sergey Kukunin

@Kukunin

thanks. going to try

yep. that's the reason. thanks

any plans to release dry-struct 0.2.2 with minor changes?

it seems that major changes are coming (already in master)

Nikita Shilnikov

@flash-gordon

nope, I'll release 0.3, soon. Bumping dry-core will require releasing rom as well, takes some time...

Sergey Kukunin

@Kukunin

what's about releasing dry-rb/dry-struct@bfc867e commit as 0.2.2?

and next release will be 0.3

Where communities thrive

People

Repo info

Activity