Le 15/10/2014 11:16, Alfredo Braunstein a écrit :
I don't understand: I guess many inset just inherit their ::latex method.
What is the risk exactly.
Advanced Search & replace all [math:x]->[math:y] in a document containing
[disabled branch: blah blah][math:x]
the result is
[math:y]
OK, but what is the link between InsetERT not having a ::latex method
ans not outputting anything? The InsetText::latex is sufficiently
flexible to handle it. Likewise for many insets derived from
InsetCommand (label, bibitem...).
Also, you may want to take in account the following method:
/// Is the content of this inset part of the output document?
virtual bool producesOutput() const { return true; }
You could also check whether there have been some output after calling
::latex and add a special marker otherwise.
Where? *::latex gets called recursively... The advanced s&r stuff
seems pretty complicated to me (and, from a quick look, a bit of a
hack if you ask me. No offense to the authors I hope, it is actually
an amazing feature).
It is a complet ehack. But this hack was easier to write than the proper
version (especially when one wants to do regex).
Adding a force_output flag could be an easy fix
to the dataloss with no negative consequences that I can think of.
What is your concern?
My concern is that we already have two pices of code where we override
the "no output property":
* in Inset::addToToc, because we sometimes want the headings inside
notes to be available inthe outline sidebar
* in Buffer::updateStatistics, because some people want to have the
words in their notes or inactive branches counted
Before we had yet another thing for yet another use, I was wondering
whether we could have a unique mechanism with correct semantics to do that.
One poiibility would be to be able to control what Inset::producesOutput
returns (maybe via a Buffer or BufferParams parameter, I don't know).
The explicit code in InsetNote should actually rely on producesOutput
instead of looking at type of note.
JMarc
