On 29/03/14 17:50, Vincent van Ravesteijn wrote:
> Tomasso's patch was not improving the situation, but was somehow hiding the
> underlying mess by building some messy function on top of it.
well, I have to some-what disagree here:
- removing from the program the possibility of an infinite loop leaving the
user with a "damn!"
- replacing two nearly identical methods with only one and a slight variation
on the params,
preserving substantially the same API and all seamless for any caller
these are, IMHO, an improvement of the situation :-).
But we can argue about whether we should refactor entirely the whole string
export machinery of LyX, and/or the whole Advanced F&R. For me, OO-wise, a
generic visitor/double-dispatching interface should have been introduced long
ago, to support generic walk-through the entire objects graph, where different
visitors would have realized various things: save, export as this, export as
that, even search or perhaps the diff function! How many methods across all
insets basically replicate a recurrent visiting structure?
Still, we can start that in the new branch for 3.0 that will go out in 1 year,
for what it matters, refactoring the LFUN dispatching as we're at it, as the
computer that formats and parses back human strings somewhat hurts my
sensibility :-), and why not plug on top the XML or git based export :-) ?...,
and still leave any 2.x further release with the infinite loop :-) ?!?
Apart from jokes, yes, it could have been fixed in other ways (removing the
replace all button, among others :-)!). Namely, I plan a final one with an
accompanying additional checkbox on the GUI allowing me to decide whether I
want to consider change tracking or not in the search, as discussed.
Bye,
T.
