On Jul 11, 11:33 am, rusi <rustompm...@gmail.com> wrote: > A gui-builder reduces the semantic gap by showing a widget when the > programmer things 'widget.' > Banging out hundreds of lines in vi/emacs for the same purpose does a > measurably poorer job.
It is very rare to need to "bang out" hundreds of lines of code to replace a mouse click interface. If properly designed a good API can compete with a GUI. In far less time than it takes me to scroll down a list of widgets, pick the appropriate one, drag it across the screen, tinker with it's absolute position, and set some attributes, i could have typed Widget(parent, **kw).geometry(blah, blah) and been done. > Note it can reduce but not close. By choosing fidelity to the gui we > have corresponding less fidelity to the algos and data-structures [And > one may assume that someone even using a gui toolkit wants to do > something with the gui and not just paint the screen] Exactly. For this very reason i have always refused to used any "point- and-click" GUI builders. I prefer to get up-close and personal with my code bases. Of course i use high levels of API abstraction for most of the work, however i already know what is happening in the lower levels if i need to dive down one tier. -- http://mail.python.org/mailman/listinfo/python-list
