Commit: cfdc50df8ccd27c95c7c38c4644a93c9905653db - cammy (git)

A categorical programming language

Commit MetaInfo

• Revision: cfdc50df8ccd27c95c7c38c4644a93c9905653db (tree)
• Time: 2024-04-27 10:16:51
• Author: Corbin <cds@corb...>
• Commiter: Corbin

Log Message

Make a plan.

I wanted to add rpypkgs, but I effectively can't use it yet.

Change Summary

• modified: todo.txt

Incremental Difference

--- a/todo.txt
+++ b/todo.txt
@@ -478,6 +478,8 @@
     * Composites -> trees of types
   * Trails -> commits
     * Note: this means prims can't have trails?
+      * No, it means that git porcelain can't handle those commits, but
+        dulwich can emit them, no problem
   * Jets -> tags
     * Note: This can be used to force jets to not be refactored, e.g. for
       constant-time expressions
@@ -487,3 +489,28 @@
       * Type signatures -> trees of types
       * Submodules -> tree of type signatures?
     * Structs -> trees
+* CHICKEN -> Chez
+  * Motivation
+    * Speed: CHICKEN is not super-fast, Chez is quite fast
+    * Package reliability: CHICKEN is federated into many packages
+      * Normally this would be a problem for Chez; wherefore packages?
+      * But Chez has matchable and many SRFIs, so the main thing to find/write
+        is a miniKanren, and that's a well-studied exercise of <100 LoC
+      * Don't even have to write that:
+        https://github.com/michaelballantyne/faster-minikanren/tree/master
+      * Just spent half an hour fucking around with eggs.nix breakage, very
+        ready to be done with CHICKEN
+      * This is the last time egg2nix fucks us over
+    * Updates: Can't use rpypkgs without bumping nixpkgs, which breaks eggs
+  * Obstacles
+    * FFI: CHICKEN's selling point is FFI and C compat
+      * Might not be a problem for much longer; we're doing FFI in RPython now
+  * Plan
+    * Switch bk over to Python 3
+      * Gives us dulwich instead of libgit2
+      * Reuse 2to3's guts? Make 2to3 a subcommand of bk
+    * Switch movelist over to Chez
+      * Should hopefully be a matter of rewriting imports
+      * Need to unburden movelist of some of its responsibilities
+        * REPL should be done somewhere else
+          * Sampler's REPL is not bad, just bare-bones, no hive support