^ ^                 ^ ^
=(o.o)=  ~byakuren  =(o.o)=
  m m                 m m

things to do c: - profile=fast (unordered :p)

1. represent strings as packed arrays (scryer)
   • it does it by “inlining” the entire string in the heap, and then zero-terminating it
   • have an internal building block in the code that tells you the functorand arity of a term, and that hides the distinction of representation even to the internal code
   • https://github.com/mthom/scryer-prolog/issues/3054
2. test conformity better
   • quick test: how many strings of length N parse to valid integeres
     • https://github.com/trealla-prolog/trealla/issues/825#issue-3190642900
     • needs countall/2
     • number_chars/2 is good test because it tests the parser
     • watch out for embedded, and escape sequences
   • comprehensive test:
     • https://www.complang.tuwien.ac.at/ulrich/iso-prolog/conformity_testing
     • let’s get to 100% some day c:
     • test input is sent to read(G_0), G_0. or …
     • this way it tests reading and writing
     • can quickly do it in emacs: https://www.youtube.com/watch?v=Ko_IrwN9yAM
3. reliable timeout and space-out errors
   • (scryer) call_with_inference_limit
   • (SICStus) call_with_time_limit/3 time_out/3 https://sicstus.sics.se/sicstus/docs/3.12.7/html/sicstus/Timeout.html
4. implement attributed variables
   • would allow for parallel constraint solving
   • start with dif/2 (scryer:dif.pl)
   • then CLP(B)
   • then clpz
5. writing terms should be moved to prolog code
   • writing cannot tell any difference in internal representation
   • can potentially move most of the parser to prolog, keeping just axiomatic code in c