Re: [FMPP] Couple newbie questions

[Daniel D. <dd...@fr...>]

Friday, October 8, 2004, 8:18:14 PM, Erich Oliphant wrote:

> Daniel Dekany wrote:
>
>>Friday, October 8, 2004, 4:31:16 PM, Erich Oliphant wrote:
>>
>>  
>>
>>>Hi,
>>>I am using FMPP w/ Ant to create a installer for our application.  
>>>Working pretty good so far but have a couple issues:
>>>
>>>1) We've using the ant XMLProperty to create Ant properties based on XML 
>>>config files.  This is great for us, as we have lots of config items and 
>>>grouping them hierarchically is convienient.  The problem, of course, is 
>>>that ant properties of the form 'parent.property' are a problem for 
>>>FMPP/Freemarker since ${parent.property} tries to resolve 'parent' then 
>>>'property' as a child.  I read the FM docs regarding the 
>>>${.vars["parent.property"]} syntax to get around the problem.  That 
>>>works but is syntactically clunky.  Is there a cleaner way to do this?
>>>    
>>>
>>
>>No... This is actually a FreeMarker syntax issue, although I don't blame
>>FreeMarker for this.
>>
> Yeah understand it's not necessarily a bad thing from an FM point of 
> view. Just like item 2, it's really only an issue when you are trying to 
> do generic text proc'ing on top of it ;)

Whatever expression language you chose, there is a high chance that it
will clash with something... OK, maybe if FreeMarker uses an extreme
separator character for separating variables from subvariables, say, ~
(foo~bar), there is far less chance for clashes, but using such extreme
syntax would drive most users crazy... :) Actually, in XML related
languages the usual subvariable separator is /, which is used as
division operator in most other languages. Maybe that's a good choice
for template languages? I don't know...

>>>2) Some of the files that we are configuring are Unix shell scripts.  In
>>>some cases the ${SHELL_VAR} form for shell variables is used, and the 
>>>FMPP processing throws an error when it hits one of these as they are 
>>>not defined.  This becomes a tricky problem as it's nice to know from an 
>>>FMPP perspective that I've hit an undefined variable reference, but in 
>>>this case I'd like to maybe log, the undefined var and keep processing.  
>>>Or is there some way to define an alternate variable resolution syntax 
>>>for FMPP/Freemarker?  This is a Java web application were building so we 
>>>are going to run into the same problem with Java Expression Language 
>>>variables that use the ${} form as well.
>>>    
>>That's an issue with FreeMarker (again), namely that it doesn't allow
>>the configuration of its syntax. I know this is a real problem for
>>general purpose file preprocessors, because so many other formats use
>>${exp}... Currently you can write ${'$'}{SHELL_VAR}, which is obviously
>>terrible if you have to do it frequently. A workaround would be to
>>replace all @{...}-s with ${...}, and all ${...} with ${'$'}{...} when
>>FMPP loads the template files, thus simulate that the FreeMarker
>>interpolation syntax is @{exp}. That slows things down a little bit,
>>also possibly distorts the column number in error messages, but I guess
>>that would be better than nothing. What do you think?
>>
> Sounds good to me :)  You have any sample code/expressions that would 
> illustrate this?

No, it's just and idea, that can be implemented (by me, mostly). Since
then I have realized that it has a little problem: FreeMarker has
built-ins for evaluating FTL expressions/fragments that are given with a
(possibly dynamically calculated) string value... in those the
interpolations would still use ${...} syntax. Well, the replacement
thing is a dirty hack, and the result is dirty as well.

-- 
Best regards,
 Daniel Dekany


____________________________________________________________________
Miert fizetsz az internetert? Korlatlan, ingyenes internet hozzaferes a FreeStarttol.
Probald ki most! http://www.freestart.hu