Warn people off

README.md

@@ -4,33 +4,12 @@
 
 [![](https://img.shields.io/endpoint?url=https%3A%2F%2Fswiftpackageindex.com%2Fapi%2Fpackages%2Fval-lang%2FLotsawa%2Fbadge%3Ftype%3Dswift-versions)](https://swiftpackageindex.com/hylo-lang/Lotsawa)
 
-A parsing library and tool with the essential features of
-[MARPA](https://jeffreykegler.github.io/Marpa-web-site/), encoded in pure Swift.
+An Earley/Leo parser in pure Swift.
 
-In particular, like MARPA, Lotsawa:
+**Don't use this code.**
 
-- Parses any [LR-regular](https://www.sciencedirect.com/science/article/pii/S0022000073800509) grammar in linear time.
-- Parses some non-LR-regular grammars in linear time.
-- Produces the parse forests for any CFG.
+There are two faster Earley parsers, [Marpa](https://jeffreykegler.github.io/Marpa-web-site/), which inspired this work, and [YAEP](https://github.com/vnmakarov/yaep).  Marpa is faster by about 1.5x but uses 10x more memory. YAEP is much faster than Marpa and uses much *much* less memory than either of the others. 
 
-Lotsawa owes almost everything of value to MARPA and its author, Jeffrey Kegler, for uncovering the
-[thread of progress in parsing technology](https://jeffreykegler.github.io/personal/timeline_v3),
-gathering it together into one group of algorithms, proving important properties about them, and
-contributing some key innovations.  This project exists primarily because MARPA is [missing
-functionality](https://github.com/jeffreykegler/libmarpa/issues/117) needed by the [Hylo
-language](https://github.com/hylo-lang/hylo) implementation.
+Most of the code was written in a very principled way, but in January 2025 I made a lot of commits as an experiment without writing comments or tests, and now some of the tests failed.
 
-Secondary reasons Lotsawa might be useful:
-
-- Lotsawa supports high-level usage from a safe, statically-typed language that compiles to efficient
-  native code.
-- Lotsawa has a simple build/installation process and no dependencies other than Swift.
-- Lotsawa encodes the grammar analysis and recognition algorithms with a relatively small amount of
-  high-level code; it may serve as a better reference for understanding the technology than either
-  the highly theoretical Marpa paper or from libmarpa's C implementation, which must be extracted
-  from a CWeb document.
-- Lotsawa can be used to precompile a grammar into static tables, eliminating some initial startup
-  cost.
-- Lotsawa uses pure [Mutable Value
-  Semantics](https://www.quora.com/What-is-mutable-value-semantics/answer/Dave-Abrahams), thus
-  eliminating many possible sources of bugs, including data races.
+You've been warned.