Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!cornell!rochester!rit!ritcv!dgr0093
From: dgr0093%ritcv@cs.rit.edu (340 Ok)
Newsgroups: comp.lang.pascal
Subject: Re: Pointers / Linked lists with files..
Summary: This looks moderately crazy.
Message-ID: <1119@cs.rit.edu>
Date: 14 May 89 16:43:37 GMT
References: <5708@cs.Buffalo.EDU> <1389@bucket.UUCP> <3377@westfort.UUCP>
Sender: news@cs.rit.edu
Reply-To: dgr0093%ritcv@cs.rit.edu (Michelangelo H. Jones)
Organization: Rochester Institute of Technology, Rochester, NY
Lines: 43

In article <3377@westfort.UUCP> westfort!dragon@tut.cis.ohio-state.edu writes:
>        In one of my programs, I have need for several variables which are 
>diminsioned to an array of 100, which are declared under a Record Type.. When 
>I define the record and the file name in my VARS section, I'm told the 
>structure is too large...
>
>Type
>mbrec=record
[ big record deleted]
>
>Var
>  mb:array[1..50] of mbrec;
>  fm:file of mbrec;

Well, I think I know why your compiler's complaining.. that record, under any
Pascal I can think of, would take at least 10952 bytes (that figure for IBM 
Turbo Pascal). An array of 50 of 'em would take 547,600 bytes. Ten thousand
bytes is an awful lot for any compiler to bite off as one chunk. Do you know
how long it would take to execute the loop the compiler would generate just 
for a simple assignment such as

mbrec1 := mbrec2; ?

That's a lot of stuff to be copying around. And then you have your much larger
variable mb; and you're saying you're going to be doing writes and reads of
these 534K chunks, all at once.

This is silly. You're asking the compiler to be doing work that you as the
programmer should be doing at a higher organizational level.

If that's not possible, I have 2 suggestions.

1> Look into some form of RLL encoding to shorten those strings. If you know
that the compressed length of a 100 character string will never exceed 50, an
array [1..100] of these just saved 5000 bytes.

2> If you only need one of these things at a time (which in this case I doubt),
look into using a free union variant record. That'll significantly drop the
size of each of these records.

Good luck...

Jones.