jwaldmann

Yes. Related: #217. In an ideal world, each mini notation operator (e.g., `[ .. ]` ) has a mini-Syntax AST node (`Stack`) that has a corresponding semantics function (`stack`) that...

This also hurts https://github.com/jwaldmann/safe-tidal-cli : I can catch most exceptions in the main loop, but example inputs from this issue here bring the system down. I guess the reason is...

@dktr0 `s "bd ."` no, this raises an exception but it does not kill the back-end.

Here's a reproducer: ``` {-# language OverloadedStrings #-} import Sound.Tidal.Context import Control.Concurrent main = do tidal

Indeed `streamReplace` has this comment: ``` -- Evaluation of pat is forced so exceptions are picked up here, before replacing th e existing pattern. streamReplace :: Show a => Stream...

Yes. I think we have three kinds of data here: 1. output of parser (AST) - can be forced (your code does it) 2. (Control)Pattern (semantics of AST) - cannot...

the code linked above seems to solve the immediate problem (in the simple case of just `d1` running) but I am not quite sure about the design and possible edge...

Did you look at/test with my code (https://github.com/jwaldmann/Tidal/tree/issue-606)? For the test case (in `i606.hs`) it prints (just once) ``` Listening for controls on 127.0.0.1:6010 Failed to evaluate. Returning to previous...

Re: new dependency. I did not think deeply. I found it convenient to write `runEvalIO (rdeepseq ms)` but it's really just one-liners. You want them expanded, to reduce the dependency...

This is about extra instances for messages. Is NFData for patterns really that useful? See my list item 2 above.