just attended the struct webex

Author: ulrichmerkel@web.de (ulrich-merkel)

thanks to daniel iseli for the demo and some insight of struct manipulating code. What I have missed (but maybe it's hard to do in such a conference) is some focus on: "we have our assignment", - what do we have to do (access a group of data, move it to another location, ...) - what does that mean for our struct? - what commands to we need (remember: reference vs. copy of data,) - when do we just change pointers, when do we create new data? because of the "by reference", one may destroy data even for an :IN parameter. Looking forward to the slides and demo forms for further experiments on that matter. I wonder how structs fits in the lists, xmlload, retrieve/reconnect {EntityName} , DTD realms. I think Iain Sharp has mentioned some experiences on the struct2component area.

2 Comments

  1. with a couple of more experiments, I found that it may take some additional struct-variables to store intermediate results to succeed building up a struct "from scratch". I transfer an associative list-in-list to a struct and need to reconstruct the same list from the struct again in the old sequences. Just as in Daniel's code it takes a couple of "v_struct_temp"s to achieve the "right" structure showing a hashtable as well as the correct sequence of entries.


    Author: ulrich-merkel (ulrichmerkel@web.de)
  2. very interesting: depending on the commands you use to build up a struct, it looks like you can optimise the struct for different usages. Or to see it from the other side of the medal: constructing a struct the wrong way may drastically slow down performance. But perhaps the compiler is not clever enough to benefit from that. It does not seem to have a lot of optimisation patterns


    Author: ulrich-merkel (ulrichmerkel@web.de)