hiccup.core/h not backward-compatible

[favila]

When hiccup.core/h is used outside macro-evaluation in 2.0.0-alpha2, it's behavior is different from 1.x:

2.x:

(h "&&&")
=> "&&&"

1.x:

(h "&&&")
=> "&&&"

Since hiccup.core/h used to be a simple alias for escape-html and was no documented expectation (that I know of) that it must be inside an html dynamic context, code that would return hiccup fragments for later inclusion into a hiccup.core/html form is going to have its strings not-escaped.

I don't actually see how it's meaningful and safe to mix hiccup.core/h-using code (written with the assumption that it must escape on its own) with hiccup2.core/html macros--something is going to either not get escaped or get double-escaped.

Possibly this could delay the decision to escape by wrapping the to-string in raw instead? This still won't be the same behavior for those using h just to escape things and will still break code that inspects un-evaluated fragments, but it will make it do the right thing when it is eventually evaluated by either hiccup or hiccup2's html.

[weavejester]

The original idea was to allow swapping of hiccup.core/html for hiccup2.core/html without any other changes, but it probably makes more sense to deprecate h and restore its original behaviour.