[Haskell] Re: [Haskell-cafe] ANNOUNCE: HStringTemplate -- An Elegant, Functional, Nifty Templating Engine for Haskell

[Sterling Clover]

Graham,
   I'm sort of playing fast and loose with referential transparency  
here, as I actually am with stringTemplateFileGroup as well. They  
both use unsafeIO to do what they want, and in corner cases could  
give silly, though not dangerous results (i.e., in the sense of being  
hazardous to your program, these calls are substantially less  
worrisome than head). stringTemplateFileGroup could conceivably be in  
IO, if it was strict in reading all the files in the directory it was  
passed, but cacheSTGroup would force every access to a group to take  
place in IO, which would make the library somewhat of a pain to work  
with. My semi-justification for this is that its referentially- 
transparent-enough for most use cases, in that just about the only  
thing one would be doing with a StringTemplate result would be  
outputting it again in some form anyway. In my experience, being able  
to "hot-edit" templates for a live app is a key benefit of a  
templating system, and forcing everything into IO to get that seems  
like an undue burden on end-users.

On the other hand, I'm also open to implementing an IO API that's  
safe and renaming the current functions to something somewhat  
scarier, or moving both them and their IO equivs to distinct modules  
so that end users could choose which to import. A number of options  
seem reasonable here.

--Sterl.

(btw, I fixed the typo you emailed me about in the repo, and also  
made a few other changes I documented at http:// 
fmapfixreturn.wordpress.com/)

On Jan 16, 2008, at 8:19 PM, Graham Fawcett wrote:

> On Jan 14, 2008 2:47 AM, Sterling Clover <s.clover at gmail.com> wrote:
>> HStringTemplate is a port of Terrence Parr's lovely StringTemplate
>> (http://www.stringtemplate.org) engine to Haskell.
>
> This is very cool.
>
> Your docs describe a function, cacheSTGroup:
>
> cacheSTGroup :: Int -> STGen a -> STGen a
> Given an integral amount of seconds and a group, returns a group
> cached for that span of time. Does not cache "misses."
>
> How does this work without breaking referential transparency?
> Shouldn't it be in the IO monad if it is time-dependent?
>
> Graham