D
Dick Davies
I've sat down to write a data-driven webapp. Nice steep
learning curve for me (I'm on my last week here so thought
I'd waste it productively).
So far FastCGI and PostGreSQL (this is an excuse for me
to finally spend time learning SQL in detail, trolls please stfu).
Few questions at the bottom of this, but it'd help to explain
how I got into this mess first...
[ Can I also place a hearty plug for Subversion here - I've
changed designs three or four times over the weekend and
renaming files/classes and nesting libs has been a dream
with it. Definitely a refactoring-friendly VCS..]
The db connection is wrapped in a class, Cnps, which I start up
outside the fcgi mainloop, and is the only thing that is SQL
savvy. The mainloop just calls methods on that class to get data
on demand.
It all works great - thanks to Carl and Ara for their
rubygarden wiki discussions on fastcgi - but
I'm coming to the stage where I have had enough of
out=<<"EOO"
<html><body>
#{cnps.champ} is winning with a score of #{cnps.score(cnps.champ)}
</body>
</html>
EOO
and want to use some kind of template thing.
First go was erb. Works well enough, but pretty verbose for
the Wiki Generation
So plan B was to use redcloth and erb,
to generate eruby from redcloth templates, store them in another
class and eval them when the requests come in.
Problem with *that* is that I need placeholders in the Textile
for each thing Eruby can parse , so I end up with a lot of <% %>
noise in the templates.
(Doing it the other way round seemed hard to do, since I want to
actually evaluate the dynamic tags at request time, but still preload
the templates).
So, Plan C. Use the ability of RedCloth to create id tags to hook up
to Amrita.
So now we have
request comes in,
fcgi decides what to display,
asks a helper class to get the template,
executes the template using Amrita
and returns the output.
the helper class is responsible for converting the
requested template (textile, (blue|red)cloth, whatever)
to Amrita-ready html.
Problem here is that Amrita doesn't duck type at all, and
it'd really like a Hash. I was hoping to have Cnps just quack
like a Hash, but it isn't going to happen. Arse.
So a few questions:
Is there a way to delay evaluation of tags?
In a 'do the simplest thing' frame of mind, is there a way to
have ruby "" strings evaluated *after* construction?
Amrita seems a bit musty, has it got a successor? Does Kwartz
help me any?
[ This would be my personal preference ]
Is there a facility to extend RedCloth with custom tags?
I could avoid the whole 'do a template from redcloth, then
evaluate template' step that way, and create some kind of unholy
crossbreed of erb/amrita with redcloth?
Something like '>instance.method<' , then these tags get
processed first and return *more* Textile, which finally
gets converted... I'd need to do the evaluation during
to_html(), but that's kind of neat.
Obviously I don't expect anyone else to
put up with that, but if I could add my own tag, that'd be
great. I could see it having lots of uses for others too.
Am I reinventing a wheel here?
learning curve for me (I'm on my last week here so thought
I'd waste it productively).
So far FastCGI and PostGreSQL (this is an excuse for me
to finally spend time learning SQL in detail, trolls please stfu).
Few questions at the bottom of this, but it'd help to explain
how I got into this mess first...
[ Can I also place a hearty plug for Subversion here - I've
changed designs three or four times over the weekend and
renaming files/classes and nesting libs has been a dream
with it. Definitely a refactoring-friendly VCS..]
The db connection is wrapped in a class, Cnps, which I start up
outside the fcgi mainloop, and is the only thing that is SQL
savvy. The mainloop just calls methods on that class to get data
on demand.
It all works great - thanks to Carl and Ara for their
rubygarden wiki discussions on fastcgi - but
I'm coming to the stage where I have had enough of
out=<<"EOO"
<html><body>
#{cnps.champ} is winning with a score of #{cnps.score(cnps.champ)}
</body>
</html>
EOO
and want to use some kind of template thing.
First go was erb. Works well enough, but pretty verbose for
the Wiki Generation
So plan B was to use redcloth and erb,to generate eruby from redcloth templates, store them in another
class and eval them when the requests come in.
Problem with *that* is that I need placeholders in the Textile
for each thing Eruby can parse , so I end up with a lot of <% %>
noise in the templates.
(Doing it the other way round seemed hard to do, since I want to
actually evaluate the dynamic tags at request time, but still preload
the templates).
So, Plan C. Use the ability of RedCloth to create id tags to hook up
to Amrita.
So now we have
request comes in,
fcgi decides what to display,
asks a helper class to get the template,
executes the template using Amrita
and returns the output.
the helper class is responsible for converting the
requested template (textile, (blue|red)cloth, whatever)
to Amrita-ready html.
Problem here is that Amrita doesn't duck type at all, and
it'd really like a Hash. I was hoping to have Cnps just quack
like a Hash, but it isn't going to happen. Arse.
So a few questions:
Is there a way to delay evaluation of tags?
In a 'do the simplest thing' frame of mind, is there a way to
have ruby "" strings evaluated *after* construction?
Amrita seems a bit musty, has it got a successor? Does Kwartz
help me any?
[ This would be my personal preference ]
Is there a facility to extend RedCloth with custom tags?
I could avoid the whole 'do a template from redcloth, then
evaluate template' step that way, and create some kind of unholy
crossbreed of erb/amrita with redcloth?
Something like '>instance.method<' , then these tags get
processed first and return *more* Textile, which finally
gets converted... I'd need to do the evaluation during
to_html(), but that's kind of neat.
Obviously I don't expect anyone else to
put up with that, but if I could add my own tag, that'd be
great. I could see it having lots of uses for others too.
Am I reinventing a wheel here?