more escaping fun

["Geir Magnusson Jr." <ge...@optonline.net>]

Cameron found a few more bugs in the parser regarding escaping, and I
think I fixed them.  Historically, I am very muddleheaded when thinking
about escaping (doesn't involve threads or something...), so we should
review what I did :

Since this is stuff for 'specialized' use, I think, in that the users of
this will be 'experts' doing something unusual, like generating Velocity
templates or producing VTL documentation, I am not to upset that it's
not as clean as our regular escaping rules.   If someone is, lets talk
about it.

1) The biggie : In a reference-like thing (RLT for short), when you
preceed a '!' with any number of \ characters, that RLT is no longer
considered a reference.  Contrast this to \\$ref which will render as
\<ref> if ref in the context, or \\$ref if not.  My argument is that
$\\!ref has no VTL meaning at all.

Therefore, no matter what the context holds :

$\!foo renders as $!foo
$\\!foo renders as $\!foo
$\\\!foo renders as $\\!foo

So, only the rightmost \ binds to the !, and unlike escapes preceeding $
and #, they don't start collapsing by twos (like our other escapeing
rules) because :

$\\!foo would be void of meaning if we collapsed the two slashes
together and had none left to escape the !.  I know I am explaining this
badly, but just contrast to how it works for $ or # and you will see
what I mean :

\\#if()  - renders as ->  \
\#if() - renders as ->  #if()
\\$ref - renders as >  \<ref> or \$ref depending on context

We could bind from the left into pairs as conventional escaping:

$\!foo =>  $!foo
$\\!foo => $\!foo
$\\\!foo => $\!foo

That's wierder to me.  I guess the symmetry is broken because the fact
that a \ is between $ and ! is an 'action' unto itself, whereas in
normal escaping, the \x binding has meaning.

Anyway, lets dicuss if history repeats and I muddled this up.

2) The other :

if foo is in the context and bar isn't :

\$!foo renders as $!foo 

\$!bar renders as \$!bar


geir

-- 
Geir Magnusson Jr.                               geirm@optonline.com
Velocity : it's not just a good idea. It should be the law.
http://jakarta.apache.org/velocity