Re: Advice needed: book on interpreter construction

lhf@csg.uwaterloo.ca (Luiz Henrique de Figueiredo) writes:
>
>try Lua


Lua is a splendid piece of software, imho -- very clean language
design, powerful through its reflexivity, nice API. The documentation
is a pleasure to read. Would that all software were as neat as Lua!


It also appears to be quite portable. I don't know if it was ever
intended to be used in the MS-DOS environment, but I got it running
more or less "out of the box". Borland C 3.1 compiled all the files
fine. There was a warning for some files: "Undefined structure 'hash'"
which I corrected by including hash.h in the pertinent files. The test
program executed all but one of the sample scripts correctly. Frankly,
I was amazed and delighted.


However, the script long.lua failed spectacularly crashing the DOS
session I was running under Windoze. This script creates an array of
5000 descending integers by adding them one at a time, selection sorts
them into ascending order and prints the 512th element. Using Bounds
Checker to screen memory accesses I discovered that the fault occurred
when a vector node occupying more than 64k bytes memory was created
and initialised (in hashnodecreate). Of course, this was the old
segmented memory problem: pointers in the large memory model wrap
round at 64k. Recompiling using the huge memory model didn't work --
I think pointers have to be declared as 'huge' too.


I'd love to be able to use Lua in my DOS programs, so if anyone can
think of a simple solution to this I would be extremely grateful.


Francis


[maybe followups should be set out of comp.compilers, to
comp.os.msdos.programmer?]
--

Return to the comp.compilers page.
Search the comp.compilers archives again.