On Sun, Apr 25, 2010 at 10:54 AM, Jasper Van der Jeugt
So, say you have the regular div tag. This would be, in our library, Text.Blaze.Html.Strict.div. That makes sense, but it seems possible that a user wants to use a div as a leaf node (e.g. `<div />`). So there are two possible signatures for `div`:
div :: Html -- An empty div element. div :: Html -> Html -- The argument is the content of the div element.
Note that the far most common use of empty elements like this is when you later want to populate the element using from JavaScript.
I'm not sure what we want to do in that case. I so far see two major options:
Option 1: We can provide leaf and non-leaf combinators for every tag. I'm not sure if this is overkill or not, but it is, for example not forbidden to have content in an `<img>` tag.
I'm afraid I was wrong when I told you this was the case. The "img" tag (and several others, like "br") must not have any content. Check the HTML spec or try it out in w3's HTML validator. Your problem still remains for tags that may have content, like "div".
I have my doubts with both options, but I would tend to go for (1), because (2) feels more unstable. Anyway, feedback and more ideas would be appreciated :-)
Given that this problem is quite rare I'd just have the user provide some empty contents in the cases where an empty tag is called for div (text "") With OverloadedStrings it's even shorter: div "" If you care about the distinction between "<tag></tag>" and "<tag />" you can postpone outputting the start tag until you produced some content and output the different tag form in case you detect that no content was added. This avoid introducing new operators, type classes or doubling the number of combinators. Cheers, Johan