Merge pull request #230 from xexyl/2015-burton

2015/burton/README.md

@@ -9,7 +9,7 @@
 make
 ```
 
-### To run:
+## To run:
 
 ```sh
 ./prog "expression"
@@ -23,7 +23,7 @@ or:
 
 Type in expressions on standard input.
 
-### Try:
+## Try:
 
 ```sh
 ./prog '?0x3343<<2'
@@ -49,7 +49,7 @@ grep '[0-9]' prog.c
 man ./calc.1
 ```
 
-## Alternate code
+## Alternate code:
 
 An alternate version of this entry, `prog.alt.c`, is provided.
 This alternate code is discussed in the "prog.c vs prog.alt.c" section of the
@@ -210,7 +210,7 @@ With -Wall -std=c99, both are quite noisy, and all can be ignored:
   * `value computed is not used [-Wunused-value]`
   * `operation on 'j' may be undefined [-Wsequence-point]`
 
-### Test Suite
+### Test Suite:
 
 If the program name begins with an "e", it echoes stdin to stdout.
 This allows for a convenient test suite:
@@ -220,9 +220,9 @@ ln -s prog eprog
 PATH=. eprog < test.in | diff - test.out
 ```
 
-### Spoilers
+### Spoilers:
 
-If you do not want to puzzle out how it works, see `spoilers.markdown`.
+If you do not want to puzzle out how it works, see [spoilers.md](spoilers.md).
 
 ## Copyright and CC BY-SA 4.0 License:
 

2015/burton/prog.c

@@ -14,8 +14,8 @@ I H char O;
 I H short Y;
 I H long long Z;
 
-				       O S[]="syntax_error!"
-				  "M@K~|JOEF\\^~_NHI]"; L*N,*K,*
+				       O S[]="syntax_error="
+				  "M@K~|JOEF\\^~_NHIA"; L*N,*K,*
 			     B,*E,*T,*A,*x,D; Q(*k)(),v; V z(P),j,_,*
 			   o,b,f,u,s,c,a,t,e,d; J l; Q _k(P){ R*K?*K++:
 			 ~-d; } V r(L a){ R a&&putchar(a); } L n(){ R*T=j=
@@ -90,21 +90,21 @@ I H long long Z;
 		    ,L**K){ V*A; q(K); W(w())**
 		  K-c*c-b||p(B),_i(!setjmp(l)?T=B
 		,M(((A=(*o=m()==c-~s&&(j=a|T[-b])&&
-	      s-m()==~s)?&o[j-a*u]:(T=B,&_)),*A=F())
+	      c-m()==-c)?&o[j-a*u]:(T=B,&_)),*A=F())
 	     ),B:(L*)S);		   } V z(P){ V
-	    s; W(m()			       &&v<=f*c
+	    s; W(m()			       &&v< f*c
 	    )R v<t				 ?(v-f)*
 	   z():(				  v-=u<<
 	   f)<f					   ?v?--v
 	  ?G()					    : i(z(
 	  )):~					     z( ):
-	  (v-=					     f) <u
+	  (v-=					     f) <f
 	  ?v?s					     = F()
 	  ,m()					     ==a>>
 	   b?s					    :G ():
 	   _:!					    (v-=
 	    t)?					  --T,_f
-	     ():				v-f?o[
+	     ():				v==b?o[
 	      (T[			      -b]|a)-
 	      a*u]:			      !z();
 		G();			       }

2015/burton/spoilers.md

@@ -0,0 +1,34 @@
+### Spoilers
+
+
+Calc is a recursive-descent parser.  It allocates from the stack deep enough
+that all expressions are exhausted before the recursion stack depth reaches the
+allocations.  The memory map and source is documented in src.doc.rot13.  Tokens
+are encoded using values 0-20, where 0 is an invalid token, 1 is minus, 2 is
+space, 3 is plus, etc.  The necessary constants are derived at the start of the program,
+using the guarantee that globals are initialized to zero; so a correct null pointer
+is obtained from an uninitialized global pointer variable, and the integers from
+uninitialized global int variables.  If x is a global int, then !x is 1, and the
+derivation proceeds from there, using some identities that are confusingly similar
+and interestingly combined (increasing obfuscation):
+
+	obc	clr
+	---	---
+	-~x	x+1
+	~-x	x-1
+	y-~x	y+x+1
+
+The derived constants are a compromise between simple expressions, maximum utility,
+and minimum number of variables used.
+
+Command line arguments are copied into the upper half of the input buffer,
+separated by a space.  This allows the code to have a single input parser by using
+a pointer-to-func to either getchar() if live, or a getchar() simulator.
+
+A small amount of obfuscation comes from void functions declared int, but missing return.
+This is necessary to allow these functions in ternary ?: expressions and paired with && and ||;
+but the requisite returns would add symbol count at no benefit to obscurity or functionality.
+Function decomposition was mostly chosen to avoid curly braces:
+if a for-loop required multiple statements, the body became a new function.
+Last, but not least: sharing state via global variables is an ancient and honored
+form of obfuscation, and remarkably effective.