May Your Data Ever Be Coherent

Daniel Spiewak discusses problems and design considerations which arise when dealing with multiple data that must work together in the same function- which is to say most functions.
Talks from NE Scala 2014: crcl.to/ne-scal...
nescala.org/

[@aryairani5017]

Great talk, but I feel bad for the cameraman. :-)

[@mav45678]

This example around 13:50 does not convince me. What's the difference if call those two functions in correct order inside some method's body, or wrap the calls as IOs and chain them together in correct order?

[@monad_tcp]

Matrix manipulation is almost never dependent on the length. Most operations in Linear Algebra never deal with the length. Instead, It's dependent on the number of dimensions, iow, the data structure.

[@valtih1978]

I do not agree that list.length is useless. Suppose you are asked to generate sequences of specific length. Suppose you add an item onto your path when descend one recursion deeper. You should stop and return the path when path length = n. I do not understand what is absurdish to ask the path.length here. Also, imperative style does not mean that file does not exist after write. If you get below the write means that the file was created. People implement continuations and do-expressions in Haskell to provide a convenient syntax for writing monadic expressions to get rid of your "dataflow" nest of shifted functions.

[@OlegYch100500]

"Suppose you are asked to generate sequences of specific length", you need to consider what you're asked about is incomplete, as it often is. This is basically what the talk is about.

[@philippederome2434]

you can classify your sequences as satisfying a certain length requirement as you produce them (presumably starting at one) without ever computing the length of any sequence and stop when your sequences associated length meets the requirement. I admit that my comment also contradicts Daniel's take on coherence, by adding a piece of data that seems unrelated (the length along the sequence or along a group of considered sequences).

[@monad_tcp]

"Suppose you are asked to generate sequences of specific length" nope, not even then you need to know the length, if you have a sequence generator, you only tell it the length once, and then you can forget that lists have length.