Le 17 août 07 à 12:34, Dov Feldstern a écrit :
It's not just a matter of convenience --- I'm still not convinced
that insets are *conceptually* right in this case.
An inset conveys to me the following idea: I have the main text,
and then inside it is something which is a *break* from the normal
flow. It may be an integral part of the text, but it's still a
break in some *textual* sense. Notes, footnotes, equations --- in
fact, nearly all of the insets we currently have --- fit into this
category.
Then the question is: *should* they all fit into this category? I see
no particular reason for that constraint. Besides, does a note and an
equation really constitute comparable breaks? What about ERTs, notes
and spaces which are insets? (Do they really constitute a textual
break?) What about lists (and layouts in general) which are not
insets? (Don't they constitute textual breaks?)
I don't think that ignore-spellcheck fits into this model. It is
not a textual break --- it doesn't really have any textual
significance. It's more of a technical thing. I think this fits in
better in a slightly different concept, which is that of "marking
off" regions of the text.
It has a "technical" significance indeed (even though it may be used
for delimitating natural and technical languages). Nevertheless it
responds to a user need among all other needs (textual among others),
even if it has no consequence into the output (BTW, this is true also
for notes).
Can I clearly delineate when something should be an inset and when
a region? No. There are probably many gray areas, and whether a
region or an inset is preferred in these cases is mainly a matter
of taste.
Why do you feel the need to delineate?
But I'll try anyhow to give some guidelines which make sense to me:
An inset is the right idea when everything inside of it --- without
even knowing what's going to be inside it --- should inherit
whatever property the inset is endowing. So for example, when I
insert a Note, then I'm saying that everything inside is not part
of the text, it's notes to the author. Even if I insert a footnote,
or a graphic, or whatever, these should all not be displayed. Or,
if I insert an image in a footnote, then I want that image to
appear inside the footnote, not in the main text.
With a "NoSpellcheck" inset, everything inside it is not
spellchecked, unless you use a "Spellcheck" inset at some point
inside. This behavior is not bad, as not unclear to the user IMO.
A region should be preferred when that kind of guarantee can't be
made. For example, when I say that part of the text should be
\emph, does that mean that also a footnote from within that text
should be \emph? I don't think so.[*] Same with language: the fact
that I switch languages within a text does not mean that everything
"within" this language will also be of the same language.
Then, if you put a footnote inside a foreign language region, what
should be the language of this footnote? Perhaps you (as me) have no
answer to this. I think this is normal because it depends on the
context (the kind of document you're writing).
Of course, one could implement virtually all these situations with
either only insets or only regions. But the fact is, there is a
conceptual difference between the two, and it makes sense to
support both concepts at the implementation level.
What about merging those two things? What you call a "conceptual
difference" is only a slightly different behavior, and an internal
property in my vision. This property doesn't have to be known by the
user, despite it should be specified at implementation level. The
user shouldn't have to think of such things.
Getting back to spell-checker, IMO spell-checking belongs more to
the region concept than to the inset concept. Obviously, not
everyone agrees with this ;).
If you looked at my last patch, you can imbricate "NoSpellcheck" and
"Spellcheck" insets. So obviously, I agree with this behavior ;)
However, insets and regions could be the same thing: a delimited area
of text having some properties useful to the user. This area could be
drawn on screen in many different ways.
But portraying things as if we're just trying to do what's easiest,
without thinking of the concepts, is not really fair...
Well, you cannot make me forget that replacing one info by another
info and two processes for each of comment, note and unselected
branches, is not a good thing (due to maintainability). Perhaps,
nothing but the font and character themselves should be stored into a
font or character property.
[*] Actually, in this case in LyX we *do* currently make the text
of the footnote \emph. I think the rationale for doing this is one
of user experience: generally speaking, I think the user expects
the font attributes to remain as they are, until explicitly
changed. However, in this specific case, I imagine that users will
usually switch off \emph within the footnote.
Not sure. Is there a bug about this? Again this may depend on the
context (or personal habits).
Mael.
