Well, runtime loading is definitely something I want (and the single biggest reason I haven't switched to using Hamlet yet), as I'm currently building my app and then uploading it to my server which is a rather slow process. Being able to upload a new template (or edit it in place) most of the time rather and a completely new binary would be a major time saver. As to your question, I'm not entirely sure, as there needs to be some sort of information about the expected template available at compile time. It ruins the symmetry somewhat, but maybe create a compile time version of the template that essentially specifies what sort of variables are expected, and so long as the variables remain the same at runtime (and there are no syntax errors) the templates should continue to function properly. I'm still pretty new to Haskell, so the kind of reflective black magic necessary to do something like that is a bit beyond me currently, but perhaps some sort of new TH to specify name -> type bindings in the template? Something like: [$runtimehamlet| exampleVar -> SomeType anotherVar -> SomeOtherType |] I'm not sure if it would be better to provide some mechanism to specify which template that's expected to be the args for, or if you should have to define an "argument set" for each template you plan to use. -R. Kyle Murphy -- Curiosity was framed, Ignorance killed the cat. On Sat, Jul 31, 2010 at 15:34, Michael Snoyman wrote:

Hey all,

An often-mentioned "cool feature" for Hamlet would be to support runtime parsing/rendering of templates. Currently, the only supported method is quasi-quoting a template and thereby have it parsed into Haskell code at compile time. In my opinion, this is definitely the preferred way of using Hamlet, as it gives you very solid compile-time guarantees of correct syntax and type safety. Nonetheless, there are some use cases (static site generation via Hakyll, a hamlet-to-html tool, etc) that would really benefit from runtime parsing.

It turns out this is pretty simple to add, except for one thing: I can't figure out a good API for passing in variables for a template. Hamlet templates have essentially four different datatypes they recognize:

* Html * Some URL data type * That same URL datatype along with a [(String, String)] to represent query-string parameters * A Hamlet template

In addition, $forall and $maybe need lists and Maybe values, respectively. Variable lookup is handled by a tree, which allows you to express arbitrary function application. $if requires Bools.

So the question is, what should an API look like? The parse function is fairly straight-forward:

parseHamletRT :: HamletSettings -> String -> Either HamletException HamletRT

However, the render function is more complicated. I've toyed with a few possible ideas:

* Use the data-object package with some complicated HamletData datatype. * Just do the complicated HamletData datatype. * Type lookup functions as parameters. * Disallow most of the more complicated features in Hamlet, like URL and subtemplates, and just allow dollar-sign interpolation with $forall and $if. That tree datatype for variable names could be collapsed into a [String].

Thoughts on the matter are welcome, as well as sample use cases you have for such a function.

Michael

_______________________________________________ web-devel mailing list web-devel@haskell.org http://www.haskell.org/mailman/listinfo/web-devel

Thanks for the input everyone. I've just pushed the new Hamlet runtime code to github[1]. Instead of using data-object, I've created a new datatype: HamletData[2]. Runtime templates support *almost* the entirety of valid Hamlet documents; the only exception is complicated references, such as: $(foo.bar).baz$ I don't think this should be a problem in real life; that syntax was only added to support functions with multiple arguments, and since runtime templates don't deal with functions, it should work just fine. I'd appreciate feedback on this before releasing. I'm planning on making this version 0.4.2. When I *do* make a release, I'll most likely accompany it with a blog post explaining how to use it, and discuss when you should use runtime versus quasi-quoted templates. Michael [1] http://github.com/snoyberg/hamlet http://github.com/snoyberg/hamlet[2] http://github.com/snoyberg/hamlet/blob/master/Text/Hamlet/RT.hs#L21 On Sat, Jul 31, 2010 at 10:34 PM, Michael Snoyman wrote:

Hey all,

An often-mentioned "cool feature" for Hamlet would be to support runtime parsing/rendering of templates. Currently, the only supported method is quasi-quoting a template and thereby have it parsed into Haskell code at compile time. In my opinion, this is definitely the preferred way of using Hamlet, as it gives you very solid compile-time guarantees of correct syntax and type safety. Nonetheless, there are some use cases (static site generation via Hakyll, a hamlet-to-html tool, etc) that would really benefit from runtime parsing.

It turns out this is pretty simple to add, except for one thing: I can't figure out a good API for passing in variables for a template. Hamlet templates have essentially four different datatypes they recognize:

* Html * Some URL data type * That same URL datatype along with a [(String, String)] to represent query-string parameters * A Hamlet template

In addition, $forall and $maybe need lists and Maybe values, respectively. Variable lookup is handled by a tree, which allows you to express arbitrary function application. $if requires Bools.

So the question is, what should an API look like? The parse function is fairly straight-forward:

parseHamletRT :: HamletSettings -> String -> Either HamletException HamletRT

However, the render function is more complicated. I've toyed with a few possible ideas:

* Use the data-object package with some complicated HamletData datatype. * Just do the complicated HamletData datatype. * Type lookup functions as parameters. * Disallow most of the more complicated features in Hamlet, like URL and subtemplates, and just allow dollar-sign interpolation with $forall and $if. That tree datatype for variable names could be collapsed into a [String].

Thoughts on the matter are welcome, as well as sample use cases you have for such a function.

Michael