Appveyor

.gitignore

@@ -1,4 +1,6 @@
 .sconsign.dblite
+.sconf_temp/
+config.log
 *.inc
 *.o
 *.os
@@ -9,14 +11,35 @@
 *.dylib
 *.dll
 *.gc*
+*.pyc
 *.neonx
+*.neond
 bin/
 errors.txt
+config.py
+test_grammar
 
 external/IntelRDFPMathLib20U1/
 external/utf8/
 external/libffi-3.2.1/
 external/PDCurses-3.4/
 external/easysid-version-1.0/
+external/hash-library/
+external/pcre2-10.10/
+external/curl-7.41.0/
+external/sqlite-amalgamation-3080803/
+external/zlib-1.2.8/
+external/bzip2-1.0.6/
+external/xz-5.2.1/
+external/minijson_writer-master/
+external/NaturalDocs/
+external/pyparsing-2.0.3/
+external/minizip11/
+external/SDL2-2.0.3/
+external/libsodium-1.0.5/
+external/libressl-2.2.4/
+
+external/etc
+external/include
 external/lib
 external/share

.neonpath

@@ -0,0 +1 @@
+lib/

.travis.yml

@@ -1,5 +1,11 @@
 language: cpp
+os:
+  - linux
+  - osx
 compiler:
   - gcc
   - clang
+before_install:
+  - which scons || brew install scons
 script: scons
+sudo: false

README.md

@@ -75,7 +75,7 @@ False
 
 This happens because `0.2` cannot be repesented exactly in binary floating point. 
 
-To resolve this problem, Neon uses the [decimal64](https://en.wikipedia.org/wiki/Decimal64_floating-point_format) floating point type, which matches the base 10 that humans use to read and write numbers.
+To resolve this problem, Neon uses the [decimal128](https://en.wikipedia.org/wiki/Decimal128_floating-point_format) floating point type, which matches the base 10 that humans use to read and write numbers.
 
 <a name="integer_division"></a>
 ### Writing division expressions such as `5 / 2` and not expecting integer division
@@ -92,7 +92,7 @@ Beginners rightly assume that `c` will be `2.5` as the result of the division.
 However, the C language definition states that `/` will be *integer* division if both operands are integers.
 So, the result in `c` is `2`.
 
-To resolve this problem, the only number type in Neon is decimal64 floating point (called `Number`).
+To resolve this problem, the only number type in Neon is decimal128 floating point (called `Number`).
 In contexts such as array indexing where integers are expected, values are checked for the presence of a fractional part before trying to use them.
 
 <a name="assignment_equals"></a>
@@ -111,11 +111,15 @@ In many common systems languages (eg. C, C++, Java, C#), a pointer may hold a "n
 
 To resolve this problem, Neon introduces the idea of a "valid" pointer. A valid pointer is one that has been checked against `NIL` (the null reference) using a special form of the `IF` statement. The resulting valid pointer can be dereferenced without causing a null pointer exception.
 
-    VAR node: POINTER TO Node
-
-    IF VALID p := node THEN
-        print(p.value)
-    END IF
+    TYPE Node IS RECORD
+        value: String
+    END RECORD
+
+    FUNCTION output(node: POINTER TO Node)
+        IF VALID node AS p THEN
+            print(p->value)
+        END IF
+    END FUNCTION
 
 <a name="empty_loop"></a>
 ### Unintended empty loop with `while (condition);`
@@ -126,16 +130,20 @@ In C and derived languages, sometimes a loop or conditional is mistakenly writte
 while (x < 5);
 {
     printf("%d\n", x);
+    x++;
 }
 ```
 
 The trailing `;` on the `while` statement is in fact an empty loop body and the loop is an infinite loop.
 
 To resolve this problem, Neon requires an explicitly terminated block in every compound statement:
 
-    WHILE x < 5
-        print(x)
-    END
+    VAR x: Number := 0
+
+    WHILE x < 5 DO
+        print("\(x)")
+        INC x
+    END WHILE
 
 <a name="logical_alternative"></a>
 ### Writing `if a == b or c` (in Python) to test whether `a` is equal to either `b` or `c`