[Oberon] Re. CASE statement error

[easlab at absamail.co.za]

Necati Ecevit wrote:
> The following code gives the error: "case range too large" in Oberon and
> traps in Bluebottle/Aos with PC0.Compile.
> 
> PROCEDURE TestCase;
> VAR x: INTEGER;
> BEGIN
>   x:=500; 
>  CASE x OF
>       1   : 
>   | 100   : 
>   | 514   :  (* err:209 case range too large *)
>   | -413  :  (* err:209 case range too large *) 
>  ELSE
>  END;
>  
> END TestCase;
> 
> According to OberonReport.html case labels must be integer or char type.
> When they are integer, min/max value of a label should be min/max of
> integer type.
> 
> This error does not occur in Component Pascal (Blackbox/ Oberon) and
> OO2C.
> 
> Is this a bug ?

A common way of implementing the 'case' construct is by an indexed
 jump table. So:
 CASE x OF
84: do84;
85: do85;
89: do89;

simplistically needs object code [of fixed/known piece-sizes]

jumpto case 84
jumpto case 85
jumpto invalid
jumpto invalid
jumpto invalid
jumpto case 89

Since the' range' in this example is 89 - 84 +1, we need 6 code-stretches
to implement the construct.
Using your range would need 514 + 413 + 1 jump entries to impliment
a mere 4 cases. Ie. the jump table is very sparse/uneconomical.
Hence it's a good reason to give an error - to tell you to use IF/ELSE
instead. NB it's not the max/min of the cases, but the range, since the
jump table uses an offset.    I speculate that cases: 888, 890 ...899 
would not give an error ?

== Chris Glur.

PS. Personally I think the only type of speed that matters is that
which saves humans time/effort [ie. human interface is all important]
but some people like to follow these speed competitions:-

URL: http://shootout.alioth.debian.org/benchmark.php?test=all&lang=a\
ll&sort=fullcpu>
Date: Mon, 7 Nov 2005 08:06:11 -0500 (EST)

Can you manipulate the
multipliers and weights to make your favourite language the fastest
programming language in the Shootout?

And remember, languages that implement more benchmarks will move to
the top of the list, and those with many missing benchmarks (No
Program, Error, Timeout) will stay at the bottom!

And remember, "For every complex problem, there is a solution that is
simple, neat, and wrong."

language score missing
D  Digital  Mars[5] 76.503
Forth  GForth[6] 75.813
C  Intel[7] 75.655
Lua[8] 68.091
Haskell  GHC[9] 66.762
Pascal  Free  Pascal[10] 66.124     <----- ?!?
SML  MLton[11] 65.132
OCaml[12] 63.514
C  gcc[13] 62.786
OCaml  (bytecode)[14] 55.492
Python[15] 55.143
C#  Mono[16] 54.962
Ada  95  GNAT[17] 53.694
Java JDK -server[18] 50.224
Lisp  CMUCL[19] 48.794
Dylan  Gwydion[20] 44.126
Lisp  SBCL[21] 43.415
Clean[22] 42.428
Oberon-2  OO2C[23] 38.458     <----- ?!?
Perl[24] 37.435
Tcl[25] 35.866
C++  g++[26] 34.7710
SML  SML/NJ[27] 33.519
Erlang[28] 31.959
Fortran  GNU[29] 31.219
Scheme  Chicken[30] 30.838
C++  Intel[31] 25.6011
PHP[32] 23.9210
JavaScript  SpiderMonkey[33] 23.0711
Ruby[34] 22.828
S-Lang[35] 22.1910
Fortran  Intel[36] 16.1413
Rexx Regina[37] 4.1113

!! amazing oberon slower than Java; is it running on a virtual
machine a la P-code interpreter ?