release 0.3.2.3

CHANGELOG

@@ -0,0 +1,30 @@
+0.3.2.3, 2018/11/13:
+  * Fix pretty-printing bug: printed programs did not include final 'End' line, so cbc could not parse the programs
+  * Relax the upper bound for containers dependency (contributed by @fumieval)
+  * Update test dependencies
+  * Update package description
+
+0.3.2.2, 2018/06/04:
+  * Build -Wall clean with GHC 8.4.3
+  * Relax the upper bound for base dependency
+  * Add missing semigroup instances (contributed by @fumieval)
+
+0.3.2.1, 2015/07/21:
+  * Naive implementations of two solving algorithms:
+    * Branch-and-bound algorithm for mixed programs
+    * Simplex algorithm for relaxed (real) programs
+  * Add tests to package description
+
+0.3.2.0, 2014/11/02:
+  * Error handling for simplifier
+
+0.3.1.0, 2014/09/19:
+  * Simplifier for Canon programs
+  * Pretty-printer for Canon programs
+
+0.3.0.0, 2014/09/03:
+  * Simplify Canonical representation
+  * Add evaluator for Program representation
+
+0.1.0.0, 2014/06/04:
+  * Initial release

README.md

@@ -1,19 +1,10 @@
 limp
 ====
 
-Eventually, this will become a pure Haskell library for Linear Integer/Mixed Programming.
+This package provides two representations for linear programs: "Numeric.Limp.Program", which is what I expect end-users to use, and
+"Numeric.Limp.Canon", which is simpler, but would be less nice for writing linear programs.
+You can convert programs from the Program representation to the Canon representation using "Numeric.Limp.Canon.Convert", and then pretty-print the program using "Numeric.Limp.Canon.Pretty".
 
-So far, however, I only have a program representation.
-I think the program representation has some minor advantages over existing ones: we separate Z-valued variables (Int) from R-valued variables (Double), which should make retrieving values a little nicer.
-The Linear type for linear functions is also indexed over whether it is purely Z, or mixed.
-While linear programming generally only supports less-than-or-equal comparisons, an less-than on Z can be changed to <=+1.
-I'm not yet sure whether the added complication on Linear is worth this minor syntactic sugar, though.
+There is a very simple branch-and-bound solver in "Numeric.Limp.Solve.Branch.Simple", and a simplex solver for relaxed (real only) programs in "Numeric.Limp.Solve.Simplex.Maps".
+See the limp-cbc package for a simple external solver.
 
-
-
-To do
-===
-Implement a Simplex-style algorithm for Canon. This probably won't be as fast as a real matrix-based simplex, but would be easy to implement.
-After both are implemented, we can check them against each other on random inputs.
-
-And lots more.

limp.cabal

@@ -1,9 +1,12 @@
 name:                limp
-version:             0.3.2.2
+version:             0.3.2.3
 synopsis:            representation of Integer Linear Programs
-description:         so far, this package just provides two representations for linear programs: "Numeric.Limp.Program", which is what I expect end-users to use, and
+description:         This package provides two representations for linear programs: "Numeric.Limp.Program", which is what I expect end-users to use, and
                      "Numeric.Limp.Canon", which is simpler, but would be less nice for writing linear programs.
-                     see the limp-cbc package for a simple solver.
+                     You can convert programs from the Program representation to the Canon representation using "Numeric.Limp.Canon.Convert", and then pretty-print the program using "Numeric.Limp.Canon.Pretty".
+
+                     There is a very simple branch-and-bound solver in "Numeric.Limp.Solve.Branch.Simple", and a simplex solver for relaxed (real only) programs in "Numeric.Limp.Solve.Simplex.Maps".
+                     See the limp-cbc package for a simple external solver.
 
 license:             MIT
 license-file:        LICENSE