Re: generic queue packages

[mfeldman@seas.gwu.edu (Michael Feldman)]

In article <1mp0h3INNp2u@huon.itd.adelaide.edu.au> andrewd@cs.adelaide.edu.au (Andrew Dunstan) writes:
>In article <1993Feb26.190255.24811@seas.gwu.edu>, mfeldman@seas.gwu.edu (Michael Feldman) writes:
>
>A "realistic" maximum might still be quite large. Say we allow lines up
>to 255 chrs. Storage of 100,000 lines of average length 10 will still 
>waste megabytes of memory if a static allocation strategy is used by the 
>compiler.

Indeed.
>
>That is why I generally avoid this sort of use of variant records for
>this sort of processing - if I want dynamic sizing I make it explicit
>by doing it myself - that makes for portable code. It also means I
>don't have to use arbitrary limits on the size of some structure.
>
I agree. The price we pay for having powerful structures - in our case
variant records with unconstrained array objects as fields - is that
we need to consider very carefully how to use them, if at all.

If the field in question is a string, using a C-style dynamically
allocated string, with the record containing a pointer thereto,
probably allows one better to control the storage allocation.
I'd probably carry a length field, though, instead of null-terminating
the string C-style, to avoid linear searching to find the length.

Is this what you had in mind, Andrew?

Mike Feldman

PS - it's really nice to discuss some technical issues for a change :-)